Statistics glossary

Plain-language definitions of the statistics terms product managers and marketers meet in dashboards, A/B tests and research.

A

A/B test: A randomised controlled experiment that splits users between a control (A) and one or more variants (B) and compares a metric.
AARRR funnel (pirate metrics): A five-stage framework for the customer lifecycle: Acquisition, Activation, Retention, Referral and Revenue, with one or two metrics per stage to find where growth leaks.
accounts payable: Money the company owes suppliers or providers for what it has already received, such as tutor payouts due at the next payout date.
accounts receivable: Money customers owe for what has already been delivered and invoiced, such as a company that pays for corporate lessons 60 days after the invoice.
accrual accounting: Recording revenue when it is earned and costs when they are incurred, whatever the date money moves. The P&L is built this way; the bank statement isn't.
acqui-hire: An acquisition made mainly to hire the company's team, not to keep its product or customers. The product is often shut down, and most of the value goes to the people through job offers rather than to the owners through the price.
activation: The moment a new user first gets the product's core value, defined by a specific action within a specific time, such as 'added an expense with at least one other member within 3 days'.
advancement: The customer moving to the next concrete step of your buying or sales process, such as a follow-up with goals, an intro or a trial.
advisory flip: Going into a conversation as if you were assessing whether the person could be a good advisor rather than trying to sell to them; it keeps you in control and out of pitch mode.
aha moment: The first moment a new user clearly sees the product's value, such as a shop owner seeing a forgotten regular come back after a reminder. Users who reach it early are far more likely to stay and pay, so onboarding is built to get them there fast.
alternative hypothesis (H1): The claim accepted when H0 is rejected: there is a difference or effect.
analysis of variance (ANOVA): Tests whether the means of three or more groups differ by comparing between-group and within-group variance.
anchoring fluff: Turning a vague answer into a concrete one by asking about a specific past instance, e.g. "When did that last happen? Walk me through it."
annual plan: A subscription paid a year in advance, usually at a discount to twelve monthly payments. It brings cash early, cuts the number of moments when a customer can cancel, and signals strong commitment.
approve: A review verdict that says the changes are fine to merge. Shown as `Approved` with a green check.
ARPA (average revenue per account): MRR divided by the number of paying accounts (customers): how much an average customer pays per month. In B2B products one account often has several users, so ARPA is used instead of revenue per user.
ARPPU: Average revenue per paying user: revenue divided by paying users only. It can rise while total revenue falls.
ARPU: Average revenue per user: revenue in a period divided by all active users, paying or not. Compare with ARPPU.
ARR (annual recurring revenue): Recurring subscription revenue expressed per year, usually MRR × 12. Buyers and investors often quote company value as a multiple of ARR.
attribution: Assigning credit for a new user or purchase to a channel or campaign. On mobile it now relies on privacy-preserving store frameworks with aggregated, delayed data.
average order value (AOV): Revenue divided by the number of orders; usually right-skewed, so outliers matter.

B

bad data: Anything from a conversation that feels like evidence but isn't: compliments, fluff (generic, future and hypothetical answers) and unexamined ideas or feature requests.
balance sheet: A snapshot at one date of what the company has (assets) and how it is financed (liabilities and equity). The two sides are always equal.
bar chart: A chart comparing counts or values across categories with separate bars.
baseline rate: The current value of the metric in control; an input to sample-size calculations.
Bayesian statistics: An approach that combines prior beliefs with new data (Bayes' theorem) to get updated, posterior probabilities.
between-group variance: The part of total variability explained by differences between group means.
blame: A GitHub file view that shows, next to every line, the last commit that changed it: who, when and with what message. It finds the source of a line, not a culprit.
Bonferroni correction: Divide alpha by the number of comparisons, e.g. 0.05 / 3 = 0.017 for three tests.
bootstrapping: Building a business with little or no outside money: the founders' own savings and, as soon as possible, revenue from customers pay for everything. It is also a mindset of spending carefully and growing at a pace the business can afford.
boring technology: Well-known, mature, widely used tools that the team already knows. They carry fewer surprises than new, fashionable ones, which matters when the business itself is already a risk.
bottom-up market sizing: Estimating a market by counting real, nameable potential customers (from maps, directories, supplier lists, your own sales route), keeping only those who fit, and multiplying by your real price. The opposite of starting from an industry total and taking a percentage.
box plot (box-and-whisker plot): A chart showing the median, the IQR as a box, whiskers for the bulk of the data and dots for outliers.
bracket retention: The share of a cohort active at least once within a window, such as days 7–13 or week 4. Fits products used weekly or monthly.
branch: A named, movable line of commits: a parallel version of the project where work happens without touching the main branch.
break-even CAC: The most you can pay per acquired user and still earn it back within a stated horizon. If you count the users they invite, add their value as measured, not assumed: 12-month LTV + invitees per paid user × the invited users' own 12-month LTV. Check the invite credit with a holdout before bidding on it.
break-even conversion: After a price rise of x, the lowest conversion (current ÷ (1 + x)) at which revenue per viewer doesn't fall, if the plan mix stays the same.
break-even point: The level of sales at which revenue exactly covers all costs, so the business neither loses nor makes money. For a subscription product it is usually expressed as a number of paying customers: fixed costs ÷ (revenue per customer − variable cost per customer).
breakage: The part of prepaid lessons or credits that customers never use. It becomes revenue only when the company can reliably expect it or when the right to use it expires.
bubble chart: A scatter plot where point size encodes a third variable.
budget: The plan of revenue, costs and cash a team commits to for a period, used to allocate money and to judge results against.
budget slack: Deliberately low revenue or high cost estimates that make a budget easy to beat. It hides real performance and wastes money.
build vs buy: The decision whether to write a part of the product yourself or pay for a ready-made service. Parts that are not the core of the product, such as login, payments or invoicing, are usually safer to buy.
burn rate: The amount of money a business or founder spends per month. Net burn subtracts the monthly revenue, and it is the number that decides how long the runway lasts.
business case: A short, numbers-backed argument for a decision: the options, the cash flows of each, the key assumptions, the risks and a recommendation.
business objective: A goal the organization wants the product to achieve, such as growing recurring revenue, entering a new market or lowering the cost to serve. Every theme on an outcome-based roadmap should support at least one.
buy-in: Stakeholders' active commitment to the roadmap because they understand its reasoning and had a chance to shape it, as opposed to grudging acceptance of a plan handed to them.

C

CAC: Customer acquisition cost: marketing spend divided by the new users (or paying customers) it brought. State which of the two you mean.
CAC (customer acquisition cost): What it costs on average to win one new paying customer: sales and marketing spending for a period ÷ new customers in that period. A healthy business earns back CAC well within the customer's lifetime.
CAC payback period: How many months of gross profit from a new customer it takes to earn back what it cost to acquire them: CAC ÷ (monthly revenue per customer × gross margin).
cannibalization: Sales of a new product or format that only replace sales of an existing one. Its value to the business is the new contribution minus the contribution of the sales it replaces.
cap table: Short for capitalization table: a list of who owns what share of a company (founders, investors, employees with options), and how that changes with each investment, grant or sale.
capital expenditure (capex): Spending on assets that serve for years, such as equipment or a platform built in-house. It hits cash at once but reaches the P&L gradually through depreciation or amortisation.
cash commitment: Putting money or a formal promise of money on the table: a letter of intent, a pre-order or a deposit.
cash conversion cycle: The days between paying for what you deliver and getting paid for it. A negative cycle means customers finance the business.
cash flow: The money actually coming into and going out of the business in a period. It differs from revenue and profit because payments arrive and leave at different times: an annual plan brings a year of cash at once, while the revenue is earned month by month.
cash flow statement: The report of cash in and out over a period, split into operating, investing and financing flows. It links the P&L to the change in the bank balance.
cash gap: A period when planned payments exceed the cash available, even if the business is profitable over the year.
cash plan: A month-by-month forecast of cash in, cash out and the closing balance, with the real timing of payments. It shows a cash gap before it happens.
categorical variable: A variable whose values are labels or groups, e.g. traffic channel, device type or plan tier.
causation (correlation is not causation): A cause-effect link; a correlation alone can also come from reverse causation, a confounder or chance, so causality needs an experiment.
central limit theorem: For large enough samples the sample mean is approximately normally distributed, even if the data are not.
centroid: The centre of a cluster: the mean of its members on every variable.
changelog: A human-readable list of what changed in each version, written for users and colleagues rather than for Git.
channel capacity: How many users a channel can bring at an acceptable cost. Costs usually rise as spend grows, so the best channel at small scale may not be the best at large scale.
Chebyshev distance: The largest absolute difference on any single variable.
churn: The share of paying customers who stop paying during a period, usually a month: customers lost ÷ customers at the start. It includes voluntary cancellations and failed payments that are never recovered.
classification: Assigning observations to known classes, e.g. predicting whether a user will churn.
classification threshold: The predicted probability above which an observation is labelled positive, often 0.5 but tuned to business costs.
click-through rate (CTR): Clicks divided by impressions for an ad, email or link.
clone: Downloading a full copy of a repository, with its history, to a computer. Engineers do it once; this course does everything in the browser instead.
cluster analysis (clustering): Methods that find groups of similar observations without predefined labels, e.g. customer segments.
cluster randomization: Assigning whole groups (a shared-expense group, a city) to a variant instead of single users, when users influence each other. It needs more users for the same precision.
code review: Reading someone else's pull request before it is merged and leaving comments and a verdict: comment, approve or request changes.
coefficient of determination (R²): The share of the outcome's variance explained by the model, from 0 to 1.
coefficient of variation: Standard deviation divided by the mean; lets you compare spread of metrics on different scales.
Cohen's d: A standardised effect size: difference in means divided by the pooled standard deviation.
cohort: A group of customers who started in the same period, for example all shops that began paying in March. Following each cohort separately shows whether newer customers stay longer or leave sooner than older ones.
cohort: A group of users who share a starting event in the same period, e.g. everyone who signed up in March.
cohort analysis (cohort table): A table with one row per acquisition cohort and one column per period since joining, showing retention or revenue. Read rows for a cohort's life, columns to compare cohorts.
cohort ROI: Return on a cohort's acquisition spend: (cumulative gross profit from the cohort − spend) ÷ spend, at a stated age such as day 90. Revenue in a calendar month mixes old and new cohorts and can't replace it.
cohort-based forecast: A forecast of active users built by adding up future cohorts: expected new users per period times their retention curve.
cold outreach: Contacting people who don't know you (calls, emails, LinkedIn messages); its main purpose is to get the first few conversations so you no longer need it.
commit: One saved step in a project's history: a snapshot of the files with an author, a date, a message and a unique id (hash). GitHub shows it as the changes against the previous commit.
commit hash: The unique id of a commit, a long string of letters and digits. GitHub usually shows the first seven characters, such as `a41c09e`.
commit history: The ordered list of commits that led to the current state of a branch or a file. On GitHub it opens from the commit counter or the `History` button.
commit message: The short text an author attaches to a commit to say what changed and why. The first line is the summary shown in lists.
commitment: A sign a customer is serious because they give up something they value: time, reputation or money.
commitment currency: What a customer pays with to show interest: time, reputation risk or cash; compliments are a currency worth nothing.
competitive alternative: Whatever customers would use if your product did not exist: a rival app, but also a paper notebook, a spreadsheet, phone calls, a messenger chat or an employee who does the job by hand.
complainer (vs customer): Someone who grumbles about a problem but has never looked for a solution; if they haven't searched for a fix, they won't look for or buy yours.
compliment: Praise for you or your idea ("Love it!", "Sounds great") that costs the speaker nothing and tells you nothing about real demand.
compound interest: Interest earned on earlier interest as well as on the original amount, so savings grow faster the longer they stay.
concierge MVP: A first version where the founders deliver the result by hand, with spreadsheets, calls and messages, behind a simple front end. It tests whether customers value and pay for the outcome before anything is automated.
confidence (on a roadmap): An explicit estimate of how likely an item is to be delivered within its timeframe, often shown as a percentage per column (e.g. Now 80%, Next 50%, Later 20%) or as a line separating commitments from stretch items.
confidence interval (CI): A range of plausible values for a parameter; a 95% CI is built by a method that captures the true value in 95% of repeated samples.
confidence level: The long-run share of intervals that contain the true value, typically 95%; equals 1 - alpha.
confounding variable (confounder): A third variable that affects both the supposed cause and the outcome, creating a spurious correlation.
confusion matrix: A 2x2 table of true/false positives and negatives for a classifier.
contingency table: A table of counts for combinations of two categorical variables, e.g. variant x converted / not converted.
continuous validation: Regularly re-checking, for example once a quarter, that the product still solves the customers' most important problem, that the problem is still urgent, and that the founders still care about it.
contribution margin: What a sale leaves after the costs that come with it (variable costs such as a tutor's payout, payment fees and video minutes). It is what pays for fixed costs and then becomes profit.
contribution margin ratio: Contribution margin as a share of the price: how many kopiykas of every hryvnia of sales are left to cover fixed costs.
control group: The group that does not get the change and serves as the baseline.
controlled experiment: A study where the researcher assigns the treatment, so a difference between groups can be attributed to it.
conversation prep: A short team session before conversations to agree on the 3 big questions, best guesses about the customer and the commitment to ask for; answer anything researchable online first.
conversation review: Going through notes and quotes with the team after conversations to update beliefs and the 3 big questions, and to discuss which questions worked.
conversion rate: The share of users who complete a target action: conversions / users. A proportion between 0 and 1.
conversion to paid: The share of a stated group, such as paywall viewers or a sign-up cohort, who made a first payment. Always name the denominator.
conversion window: The time a user has to complete the next funnel step for it to count, measured from entering the funnel.
correlation: A statistical association between two variables: when one goes up the other tends to go up (positive) or down (negative).
correlation matrix: A symmetric table of correlation coefficients for every pair of variables, with 1 on the diagonal.
cost allocation: Assigning shared costs (office, finance team, platform) to products, subjects or teams by some key. Useful for pricing and reporting, dangerous for decisions: allocated costs usually don't disappear when a product is dropped.
cost of capital (WACC): The return lenders and owners expect for financing the company, blended by their shares (weighted average cost of capital).
cost of delay: The value a business loses for each unit of time an item is not delivered, e.g. revenue missed per week before a seasonal peak. It makes urgency explicit when comparing items of similar value.
cost of revenue: The costs of delivering what was sold, such as video minutes, payment fees and, for a principal, provider payouts. Revenue minus cost of revenue is gross profit.
covariance: The average product of the deviations of two variables from their means; its sign shows the direction of the relationship.
critical path (customer): The few steps in a customer's journey that a product absolutely must get right for anyone to adopt it. Solving them comes before any other improvement.
critical problem: The problem that matters most to an audience: it is painful, frequent, mandatory, eats time or money, and people already patch it with their own spreadsheets or paper systems. People pay to solve it and cancel other tools first.
critical value: The value of a test statistic beyond which H0 is rejected at the chosen alpha.
currency risk (FX exposure): How much results change when exchange rates move, for example when costs are billed in dollars and revenue arrives in hryvnias.
customer concentration: How much of the revenue depends on a few large customers. High concentration gives those customers power over the product and puts the business at risk if one of them leaves.
customer conversation: An informal, learning-focused talk with a potential or existing customer about their life, problems and behaviour, not a pitch or a scripted survey.
customer development: An approach to building a business by systematically learning from customers before and while building the product; customer conversations are one of its core tools.
customer segment: A group of customers who share the same problem, goals and buying context; if your conversations produce inconsistent problems, the segment is too broad.
customer slicing: Repeatedly splitting a broad segment by who wants the solution most and why, until you know exactly who to talk to and where to find them.
customer-funded business: A business whose growth is paid for by its customers' money rather than by investors: through pre-sales, annual prepayments, deposits or simply revenue that arrives before the costs it covers.
customer/market risk: Uncertainty about whether customers want the solution, will pay for it and exist in large enough numbers; this is what customer conversations test best.

D

DAU, WAU, MAU: Daily, weekly and monthly active users: distinct users who did a qualifying action in a day, a 7-day or a 30-day window. What counts as 'active' must be defined.
day-N retention: The share of a cohort that is active exactly on day N after joining. Strict and precise, but noisy for products not used daily.
debt avalanche: Paying minimums on all debts and putting every extra hryvnia into the one with the highest rate. It costs the least interest.
debt financing: Raising money through loans or credit lines that must be repaid with interest on a schedule, whatever the results.
debt snowball: Paying minimums on all debts and putting every extra hryvnia into the smallest balance first. It costs more interest but gives quick wins that help people keep going.
decision maker: The person who says yes to buying and controls the money. In a small business it is often the owner, while the people who feel the problem every day and use the product are employees.
decision maker: The person who controls the budget or signs off on a purchase; in B2B they are often not the person you're talking to, so ask where the money comes from.
decision rule: What the team will do for each possible test result, written before the test starts: ship, don't ship, or iterate, and at what effect size.
deductible: The part of a loss you pay yourself before insurance pays. A higher deductible usually means a cheaper policy.
deferred revenue: Money received for lessons or service not yet delivered. It sits on the balance sheet as a liability and turns into revenue as the company delivers.
deflecting compliments: Politely ignoring praise and steering back to facts, e.g. replying to "Love it!" with "How do you handle this today?"
degrees of freedom (df): The number of values free to vary once the estimated quantities are fixed, e.g. n - 1 for one sample, n1 + n2 - 2 for a two-sample t-test.
dendrogram: A tree diagram of hierarchical clustering; cutting it at a chosen distance gives the clusters.
dependent variable (outcome): The variable a model explains or predicts, e.g. revenue per user.
deploy: Putting a version of the code onto the live servers so users see it. Often happens automatically after a merge into `main`.
deposit guarantee: A state-backed promise to repay depositors up to a limit if a bank fails.
depreciation and amortisation: Spreading the cost of a long-lived asset over the years it is used. A cost in the P&L that moves no cash in that period.
descriptive statistics: Numbers and charts that summarise a dataset (typical value, spread, shape) without generalising beyond it.
design effect: How many times more users a test needs when whole groups are randomized: 1 + (m − 1) × ICC, where m is the average group size and ICC how alike members of one group behave.
desirability, feasibility, viability: Three lenses for judging an idea: do customers want it, can we build it, and can the business make it pay? A simple score (e.g. 1–3 on each, summed) favors ideas that are strong on all three.
deviation: The difference between a value and the mean; deviations always sum to zero.
diff: A view of what changed between two versions: removed lines in red with a minus, added lines in green with a plus.
digging: Asking follow-up questions to uncover the motivation behind a request, opinion or strong emotion: "Why do you want that?", "What would that let you do?"
dilution: The drop in the founders' ownership percentage when the company issues new shares to investors. Each funding round leaves the founders a smaller slice of the company and of any future sale.
dilution: Including users who never saw the change in the analysis. The true effect gets spread over more people and becomes harder to detect.
dimensionality reduction: Replacing many correlated variables with a few summary variables (factors or components).
discount rate: The annual rate used to convert future cash flows into today's money. It reflects what the money could earn elsewhere at similar risk.
discounted payback period: How long it takes for discounted cash flows to repay the investment. Like any payback measure, it ignores everything that happens afterwards.
discriminant analysis: Finds boundaries (discriminant functions) that best separate known groups using predictors; groups minus one functions.
distance metric: A rule for measuring dissimilarity between observations: Euclidean, Manhattan or Chebyshev.
distribution: How often each value (or range of values) of a variable occurs.
diversification: Spreading money across many investments so that one failure can't sink the whole plan.
draft pull request: A pull request marked as work in progress. It cannot be merged until the author clicks `Ready for review`.
driver-based forecast: A forecast built from the operational numbers that cause the financial ones (learners, lessons per learner, price, payout) instead of last year's totals plus a percentage.
DSO (days sales outstanding): The average number of days it takes customers to pay after an invoice: receivables divided by credit sales, times the days in the period.
due diligence: The detailed check a buyer runs before closing a deal to confirm that everything the seller claimed is true: finances, metrics, contracts, code, legal risks. Sellers should also check the buyer.
dunning: The process of recovering failed subscription payments: automatic retries, messages asking the customer to update their card, a grace period and, if nothing works, a pause or cancellation.

E

early adopter: A customer who tries a new product while it is still rough, buys into what it could become, and gives feedback. Early adopters forgive gaps, but expect fast progress in return.
early evangelist: A first customer who has the problem, knows it, has searched for and patched together a solution, and has budget; they commit before it's rational.
earn-out: Part of a company's sale price that is paid only if the business reaches agreed goals after the sale, usually while the founder keeps working in it. The safest approach is to treat it as a possible bonus, not guaranteed money.
EBITDA: Earnings before interest, taxes, depreciation and amortization: operating profit after paying everyone, owners included at a market salary, but before financing, tax and accounting write-downs. Larger companies are often valued as a multiple of it.
effect size: How big a difference or relationship is, independent of sample size, e.g. absolute lift, relative lift or Cohen's d.
eigenvalue: How much of the total variance a factor explains (the sum of its squared loadings).
emergency fund: Money set aside only for a loss of income or an unplanned essential cost, sized in months of essential costs and kept where you can reach it quickly.
entry cohort: The users who entered a funnel or signed up in a date range; their later steps count within the window whatever date they fall on.
equity financing: Raising money by selling a share of the company. There is no interest to pay, but the owners' share and future profits are diluted.
error bars: Lines around a point estimate showing its SD, standard error or confidence interval; always state which one.
essential costs: Spending you can't skip for a month without harm: housing, utilities, food, transport, medicine, minimum debt payments.
Euclidean distance: The straight-line distance: square root of the sum of squared differences across variables.
event: A record that something happened in the product: who did it, what happened and when, plus optional details (properties). Events are the raw material of product analytics.
event property: A detail attached to one event, such as the amount of an expense or how it was entered. It describes that moment, not the user.
event total: How many times an event happened in a period, counting repeats by the same person. Totals grow with heavy users; unique users don't.
exclusions: The users and events a metric leaves out on purpose, such as staff accounts or server-created events, and how they are recognized.
exclusivity: A clause in a letter of intent that bars the seller from talking to other buyers for a set period while the buyer runs due diligence and prepares the final contract. Also called a no-shop clause.
exit: The moment founders turn their ownership into money, usually by selling the company, fully or partly, to another business or investor.
expected frequency: The count a cell would have if the variables were unrelated: row total x column total / grand total.
expense ratio: The share of your money a fund takes every year for running it. It is charged whatever the return.

F

F-statistic (F-ratio): In ANOVA, between-group mean square divided by within-group mean square; large values mean groups differ.
F-test for equality of variances: The ratio of the larger to the smaller sample variance; tests whether two groups differ in spread.
factor (in ANOVA): A categorical explanatory variable whose levels define the groups, e.g. landing page version or traffic channel.
factor analysis: Explains correlations among many observed variables by a few hidden factors; used for surveys and to reduce dimensions.
factor loading: The correlation between an observed variable and a factor.
factor rotation (orthogonal, oblique): Rotating factor axes so each variable loads strongly on one factor; orthogonal keeps factors uncorrelated, oblique allows correlation.
fake door test: A cheap demand test: show a button, page or price for something that does not exist yet and count how many people try to use or buy it before deciding to build it.
false discovery rate (FDR): The expected share of false positives among significant results; controlled by the Benjamini-Hochberg procedure.
false negative: Evidence that makes you drop a good idea, often caused by asking the wrong people or the wrong questions.
false positive: Evidence that makes you believe an idea works when it doesn't, e.g. polite enthusiasm mistaken for demand; more dangerous than no data because it raises confidence.
feature: A specific capability or piece of functionality the team builds to solve a problem. On an outcome-based roadmap, features appear, if at all, under the theme they serve.
feature branch: A short-lived branch for one feature or fix, merged back when the work is done.
feature factory: A team or company that measures success by how many features it ships rather than by the outcomes they produce, and rarely checks whether shipped work changed anything.
feature flag: A switch that turns a feature on or off for some or all users without a new deploy, so unfinished or risky work can be merged and deployed while staying hidden.
feature request: A customer's suggestion of what to build; treat it as a clue to an underlying need to understand, not an item to add to the backlog.
financial buyer: A buyer who acquires a company for its profits and growth and wants it to keep running much as before, often an investment firm with a portfolio of small software businesses.
financial goal: A wish turned into an amount, a date and a monthly saving number.
financial leverage: Using borrowed money to finance the business. It raises the owners' return when things go well and deepens losses when they don't, because interest must be paid either way.
fishing for compliments: Asking for approval rather than information, such as "Do you think it'll work?" or "Do you like it?"; it guarantees kind but useless answers.
fixed cost: A cost that doesn't change with volume within a period and a range of activity, such as salaries of the in-house team or office rent. Fixed doesn't mean unavoidable: many fixed costs are chosen.
flexible spending: Spending you choose month by month and can change quickly, such as eating out, delivery, taxis, clothes and subscriptions. The first place to look when a plan doesn't add up.
fluff: Vague answers that sound like data but describe no real event: generic claims, future promises and hypothetical maybes.
FOP (sole proprietor): A Ukrainian sole proprietor: a person registered to do business in their own name, usually on the simplified tax system.
forecast: A model-based estimate of a future value, ideally with an uncertainty range.
fork: Your own copy of someone else's repository on GitHub, used to propose changes when you cannot create branches in the original.
freemium: A pricing model with a permanent free plan and paid upgrades. It works only if the free plan's limits are hit by customers who get real value, otherwise free users cost money and never convert.
frequency (absolute and relative): The count of observations with a given value (absolute) or its share of the total (relative, in %).
frequency polygon: A line chart that joins the frequencies of consecutive values; useful when there are many values.
frequency table: A table listing each value or bin and how many observations fall into it.
Friedman test: The non-parametric alternative to repeated-measures ANOVA: ranks each subject's measurements and compares rank sums.
fun money: A fixed amount in the spending plan for things you enjoy, spent without justifying it. Keeping it on purpose is what makes a plan last.
functional vs structural model: A functional model links inputs to an outcome (e.g. a regression); a structural model describes the components of the outcome.
funnel: A sequence of steps users should go through, with the share who reach each step. Funnel results depend on step order, the conversion window and who enters.
future promise: A statement like "I would buy that" or "I will use it"; people are consistently over-optimistic about their future behaviour.

G

generic claim: A statement about what someone "always", "usually" or "never" does; it describes who they want to be, not what they actually did.
Git: The version control system most software teams use. It runs on each developer's computer and records the project's history as commits.
git flow: An older, release-oriented branching model with long-lived `main` and `develop` branches plus `feature/*`, `release/*` and `hotfix/*` branches.
GitHub: A website that hosts Git repositories and adds teamwork tools on top: pull requests, reviews, issues, project boards and releases.
GitHub flow: A simple team workflow: `main` is always deployable; every change goes on a short-lived branch, through a pull request and review, and is merged and deployed soon after.
GMV (gross merchandise value): The total value of everything sold through a platform, including the part that goes to sellers or tutors. A volume metric, not revenue.
government bonds (OVDP): Debt of the state sold to investors for a fixed period at a fixed yield; in Ukraine individuals can buy them through banks and apps.
grace period: The time during which a credit card purchase costs no interest if the whole balance is repaid by the due date.
grandfathering: Letting existing customers keep their old price or plan after a price increase, for a limited time or forever, while new customers pay the new price.
gross margin: The share of revenue left after the direct costs of delivering the product (for SaaS: hosting, third-party services used per customer, payment fees and customer support): (revenue − cost of revenue) ÷ revenue.
gross profit: Revenue minus cost of revenue: what is left to pay for the team, marketing and everything else.
gross revenue retention (GRR): Share of recurring revenue from the customers you had at the start of a period that you still have at the end, ignoring upgrades; never above 100%.
guardrail metric: A metric the change must not worsen beyond an agreed margin, checked by its interval rather than its point estimate.
guardrail metric: A metric that must not get worse (e.g. refunds, page speed) even if the primary metric improves.

H

help center: A collection of short, searchable articles, usually with screenshots, that answer customers' common questions, so support sends a link instead of typing the same answer again. A good habit is to write an article the second time a question comes in.
hierarchical clustering: Repeatedly merges the most similar observations or clusters until one cluster remains; the number of groups need not be known.
histogram: A chart of a quantitative variable's distribution: adjacent bars show how many values fall into each bin.
holdout group: A small group kept without a feature or campaign for a long time to measure its cumulative effect after launch.
hotfix: An urgent fix for a problem in the live product, shipped outside the normal release rhythm.
hurdle rate: The minimum return a project must beat to be worth doing, usually the cost of capital plus a premium for extra risk.
hypothesis testing: A procedure that decides whether data give enough evidence against a null hypothesis.
hypothetical question: A question about an imagined situation ("Would you…?", "Could you see yourself…?"); the answers are guesses, not evidence.

I

implications of a problem: What a problem actually costs the customer in money, time or risk; it separates problems people will pay to solve from annoyances they live with.
important question: A question whose answer could change or kill your business plan; if a surprising answer wouldn't change anything, the question wasn't important.
incentivized sign-ups: New users who joined because a partner or promo paid them something for signing up, such as a bank bonus. Many sign up for the reward, not the product, and retain worse than organic users.
income smoothing: Paying yourself the same amount every month from a buffer account that absorbs good and bad months.
incrementality: The part of a result that happened only because of a campaign or channel, measured against a holdout or a geo test. Attributed conversions are often not incremental.
independent samples: Groups made of different, unrelated observations, e.g. users in variant A vs users in variant B.
independent variable (predictor): A variable used to explain or predict the outcome, e.g. ad spend or number of sessions.
index fund: A fund that holds all the shares or bonds of a market index instead of picking winners, usually at a low fee.
inflation: The rise in the general level of prices, which means each hryvnia buys less over time.
information criterion (AIC, BIC): A model-comparison score that rewards fit and penalises the number of parameters; lower is better.
input metric: A metric a team can move directly that feeds a higher-level metric, such as the share of new groups with a second member.
installment plan: Buying now and paying in equal parts. Often advertised as zero percent, but it can hide a monthly fee or a higher price than paying at once.
interaction effect: When the effect of one factor depends on the level of another, e.g. a discount works on mobile but not on desktop; shows as non-parallel lines.
intercept: b0: the predicted outcome when all predictors equal zero.
internal traffic: Activity from the team's own accounts: QA, staff, test devices. It must be excluded from quality and volume metrics, and the flag that marks it is often incomplete.
interquartile range (IQR): Q3 minus Q1: the spread of the middle 50% of the data; robust to outliers.
involuntary churn: Customers lost because their payment failed and was never recovered (an expired, blocked or empty card), not because they chose to cancel.
iron triangle: The trade-off between schedule, scope and resources, with quality caught in the middle: when work runs late, one of the three must give, or quality quietly pays the price.
IRR (internal rate of return): The discount rate at which a project's NPV is zero. Useful as a summary, misleading when ranking projects of different size or timing.
irregular income: Income that changes a lot from month to month, as for freelancers, tutors and sole proprietors. It is planned over a year, not a month.
issue: A GitHub item for a bug, task or idea, with a discussion, labels, assignees and a state (open, closed as completed, closed as not planned).

J

job story: A need statement in the form "When [situation], I want to [motivation], so I can [expected result]". It focuses on context and motivation rather than on a user persona.

K

k-means clustering: Splits data into k clusters by alternately assigning points to the nearest centroid and moving centroids to cluster means.
Kaiser criterion: Keep only factors with an eigenvalue of at least 1.
Kano model: A way to classify product attributes by how they affect satisfaction: basic expectations cause dissatisfaction when missing, performance attributes satisfy in proportion to how well they work, and delighters please without being expected. Over time delighters turn into expectations.
keeping it casual: Running early customer conversations as quick, relaxed chats rather than scheduled formal interviews; people speak more honestly and you learn faster.
key result: A measurable indicator with a baseline and a target that shows progress towards an objective, e.g. "raise the 60-day subscription renewal rate from 42% to 50%". A good key result describes an outcome, not a task.
key-person risk: The danger that the business stalls when one person, often the founder, is sick, on holiday or gone, because only they know how things work. Documentation, automation and hiring reduce it.
Kruskal-Wallis H test: The non-parametric alternative to one-way ANOVA, based on rank sums of three or more groups.

L

label: A coloured tag on issues and pull requests, such as `bug` or `checkout`, used to sort and filter work.
latent factor: An unobserved trait inferred from correlated indicators, e.g. 'price sensitivity' behind several survey answers.
leading question: A question that signals the answer you want ("Don't you hate it when…?"); it manufactures agreement instead of revealing the truth.
learning bottleneck: When customer insight lives in one person's head instead of being shared with the team, so "the customer said so" becomes an unchallengeable order.
least squares method: Fits a regression line by choosing coefficients that minimise the sum of squared residuals.
legacy plan: An old plan or price that new customers can no longer choose. Keeping existing subscribers on it is called grandfathering; the fair version is time-limited, with clear notice before they move to the current price.
letter of intent (LOI): A mostly non-binding document in which a buyer sets out the main terms of a planned acquisition, such as price, payment structure and timeline, before full due diligence. It is the point to involve a lawyer.
letter of intent (LOI): A non-binding written statement that a customer intends to buy under certain conditions; weaker than payment but far stronger than words.
lifetime deal: A one-time payment for permanent access to a subscription product, often sold through deal websites. It brings quick cash but turns future recurring revenue into a single payment, while the cost of serving the customer continues.
lift (absolute and relative): The change of a metric in the variant vs control: absolute lift in percentage points, relative lift in % of control.
Likert scale: A survey rating scale such as 1 = strongly disagree to 5 = strongly agree; an ordinal variable.
linear regression: A model that predicts an outcome as a straight-line function of a predictor: y = b0 + b1 x.
liquidity: How quickly and cheaply you can turn savings into cash you can spend, without losing interest or value.
list of 3 (big questions): The three most important things you want to learn from a given type of person, chosen with your team before the conversation and updated as you learn.
log-odds (logit): The natural logarithm of the odds; positive when p > 0.5, negative when p < 0.5.
logistic regression: A model for a yes/no outcome that predicts the log-odds as a linear function of predictors and returns a probability.
LTV (customer lifetime value): The total revenue (or profit) a business expects from one customer over the whole time they pay. A rough estimate: average monthly revenue per customer ÷ monthly churn.
LTV curve: Cumulative revenue (or gross profit) per user of a cohort by day or month since joining. Its shape shows when a cohort pays back and whether it's still growing.
LTV:CAC ratio: Customer lifetime value divided by customer acquisition cost: how many times a typical customer pays back what it cost to win them. A common rule of thumb for healthy subscription businesses is about 3 or more.

M

main branch: The branch that holds the accepted, current state of the project, usually called `main` (older projects: `master`). In GitHub flow it is what gets deployed.
main effect: The average effect of one factor across all levels of the other factors.
Manhattan distance: The sum of absolute differences across variables, like walking along a street grid; less sensitive to outliers.
Mann-Whitney U test: A non-parametric test comparing two independent groups by pooling, ranking and comparing rank sums.
margin of error: Half the width of a confidence interval, e.g. about 1.96 x SE for 95%.
margin of safety: How far sales can fall before the business reaches break-even, as units or as a share of current sales.
marginal CAC: Extra spend divided by the extra users it brought. It rises as a channel scales, while the average CAC hides the rise.
mathematical model: A simplified formal description of a system that lets you explore scenarios without real experiments.
mean (arithmetic mean): The sum of all values divided by their count; sensitive to outliers.
measure of central tendency: A single number describing the typical value of a variable: the mean, median or mode.
measure of variability (spread): A number describing how much values differ from each other: range, IQR, variance or standard deviation.
median: The middle value of the sorted data (the average of the two middle values when n is even); robust to outliers.
meetings anti-pattern: The habit of turning every chance to talk to a customer into a scheduled meeting; it wastes time, raises expectations and kills spontaneous learning.
merge: Combining the commits of one branch into another, for example a finished feature branch into `main`.
merge commit: A commit with two parents that joins two branches. GitHub creates one when a pull request is merged with `Create a merge commit`.
merge conflict: A situation where two branches changed the same lines differently, so Git cannot combine them automatically and a person has to choose the result.
metric: A quantitative measure tracked for a product or campaign, such as conversion rate, AOV, retention or CTR.
metric definition: The exact rule for computing a metric: which events count, which users are excluded, the time window and the time zone. Two people with the same definition get the same number.
metric tree: A breakdown of a top metric into the input metrics that drive it, so each team can see which lever it moves.
microloan: A small, short loan, usually online, with a very high real annual cost. Rolling it over is how small debts turn into large ones.
milestone: A group of issues and pull requests with an optional due date, used to track progress towards a release or goal.
minimum detectable effect (MDE): The smallest change a test is sized to detect with the chosen significance level and power. Pick it from what would matter to the product, not from what is easy to detect.
minimum detectable effect (MDE): The smallest effect a test is designed to detect with the planned power; smaller MDE needs a larger sample.
minimum payment: The smallest amount a card requires each month. Paying only this keeps most of the debt, and its interest, alive for years.
mission statement: A statement of what the organization does now, and for whom, that moves it towards its vision. A good mission creates value for others, inspires, is plausible and is specific to the business.
mix shift: A change in an average or ratio caused by a change in who is in the group, not by a change in behavior. Example: expenses per active user fall because a wave of new users arrived.
mix-adjusted rate: A rate for one period recomputed with another period's segment mix, so that a comparison shows changes in behavior rather than changes in who arrived.
mode: The most frequent value in the data; the only measure of central tendency that works for categories.
money milestone: A point where a young product's recurring income covers one more real bill: the first payment, the tools, one founder's living costs, both founders, proper salaries. Progress is measured by these sums, not by months or effort, and each one changes what the founders should work on.
money plan: A one-page summary of goals, the monthly plan, the cushion, debts, savings and investments, reviewed once a year.
MoSCoW method: A way to communicate release scope by sorting requirements into Must have, Should have, Could have and Won't have (this time). It labels priorities that have already been decided; it does not decide them.
MRR (monthly recurring revenue): The subscription revenue a business can expect every month from its current customers. Annual plans count as their price divided by 12; one-off payments are left out.
multi-sided market: A business serving several distinct customer groups at once (e.g. hosts and guests, users and advertisers); each side needs its own conversations.
multicollinearity: Strong correlation between predictors that makes regression coefficients unstable and hard to interpret.
multiple comparisons problem: Running many tests (metrics, segments, variants) raises the chance that at least one is a false positive.
multiple regression: Regression with several predictors; each coefficient is the effect of its predictor holding the others constant.
MVP (minimum viable product): The smallest product that solves the core problem well enough that early customers will pay for it. For a subscription app that usually means: sign up, get the main job done, and pay; everything else waits.

N

natural usage rhythm: How often a satisfied user needs the product: daily for messaging, weekly or monthly for splitting bills. Metrics and retention should be read at this rhythm.
negative churn: When extra revenue from customers who stay (upgrades, more locations, add-ons) is larger than the revenue lost to cancellations and downgrades, so MRR grows even without new customers.
Net Promoter Score (NPS): Share of promoters (9-10) minus share of detractors (0-6) on a 0-10 'would you recommend' question; ranges from -100 to 100.
net revenue (agent vs principal): Revenue reported as the company's own share when it acts as an agent that connects buyers and providers; a principal that controls the service reports the full price and the provider's share as a cost.
net revenue retention (NRR): Recurring revenue from the customers you had at the start of a period, including their upgrades, as a share of what they paid at the start; can exceed 100%.
net worth: Everything you own minus everything you owe; tracked once a year, it shows whether the plan is working.
network effect: When the product's value for one user depends on others using it. In tests it means one user's variant can leak into another's experience.
next steps: The concrete, agreed actions that follow a meeting (who does what, by when); if you don't know what happens next, the meeting was pointless.
niche: A narrow, clearly defined group of customers who share a job, problems and vocabulary and talk to each other, for example independent car repair shops. Small enough to reach and understand, big enough to support the business.
nominal rate: The rate a bank or bond states, before inflation and usually before tax.
non-parametric test: A test that makes no normality assumption, usually by working with ranks; robust to outliers and skew.
normal distribution: The symmetric bell-shaped distribution; about 68% of values lie within 1 SD of the mean and about 95% within 2 SD.
north star metric: The one metric that best captures the value users get from the product and predicts long-term success. Teams use it to align decisions, not as the only number they watch.
note-taking symbols: Short marks added to notes to tag signals such as emotion, pain, goal, obstacle, workaround, context, feature request, money, a named person and a follow-up, so notes can be sorted later.
novelty effect: A temporary change in behaviour because something is new; it fades, so short tests can overstate the effect.
Now-Next-Later roadmap: A roadmap format with three columns instead of dates: what the team is working on now, what comes next and what is planned for later, with certainty and detail decreasing from left to right.
NPV (net present value): The sum of all of a project's cash flows, each converted to today's money, minus the investment. A positive NPV means the bet earns more than the discount rate.
NRR (net revenue retention): The share of recurring revenue from existing customers that is still there after a period, counting expansion and subtracting downgrades and churn: (starting MRR + expansion − contraction − churned MRR) ÷ starting MRR. Above 100% means existing customers grow revenue on their own.
null hypothesis (H0): The default claim of no difference or no effect, e.g. 'the new checkout does not change conversion'.

O

observation: One measured unit in a dataset (one user, order or session); usually one row of a table.
observational comparison: Comparing groups the team didn't assign at random, such as before vs after a release. It can show a difference but can't prove the change caused it.
obstacle: Something that stops a customer from solving a problem they want to solve (budget rules, IT policy, approvals); your product will likely have to deal with it too.
odds: Probability of an event divided by the probability of it not happening: p / (1 - p).
odds ratio: How many times the odds change per unit of a predictor; exp of a logistic regression coefficient.
OKR (objectives and key results): A goal-setting method that pairs a qualitative objective with a few quantitative key results showing whether it has been reached, usually over a quarter.
one-tailed vs two-tailed test: A two-tailed test detects a difference in either direction; a one-tailed test only in a direction fixed in advance.
operating cash flow: Cash generated or used by the core business in a period: profit adjusted for non-cash items and for changes in working capital.
operating expenses (opex): The costs of running the company that aren't part of delivering a specific sale: salaries, marketing, office, tools.
operating leverage: How strongly operating profit reacts to a change in sales: contribution margin divided by operating profit. The more fixed costs, the stronger the swing in both directions.
operating profit (EBIT): Gross profit minus operating expenses, before interest and taxes: what the business itself earns.
opinion question: A question asking what someone thinks of your idea or a price ("Is it a good idea?", "How much would you pay?"); only the market can answer it.
opportunity cost: The value of the best option you give up by choosing another, such as the contribution of the lessons a tutor could have taught instead.
opportunity solution tree: A visual map that starts from one desired outcome, branches into customer opportunities (needs, pains, desires), then into possible solutions and the experiments that test them.
ordinal variable: A categorical variable whose values have a natural order but uneven gaps, e.g. a 1-5 satisfaction rating.
Osborne effect: A drop in current sales caused by announcing a future product too early, because buyers wait for the new version. It is one reason to be careful about what an external roadmap reveals.
outcome: A measurable change in customer behavior or business results that a piece of work is meant to cause, such as fewer abandoned carts or more subscription renewals.
outcome vs output: The distinction between what a team produces (output: features, releases, documents) and the difference those things make for customers and the business (outcome: changed behavior, solved problems, moved metrics).
outcome-based roadmap: A roadmap organized around the changes the team wants to achieve for customers and the business (themes tied to objectives and key results) rather than around a list of features with dates.
outlier: An observation far from the rest of the data (a common rule: beyond 1.5 IQR from the quartiles); can distort the mean and SD.
output: Anything a team ships or produces: a feature, a release, a campaign, a document. Output is necessary, but on its own it says nothing about whether anything improved.
overfitting: A model that fits noise in its sample and predicts poorly on new data; checked with holdout data, limited with AIC/BIC.
owner earnings: A small business's yearly profit before paying its owners anything: revenue minus all costs except the founders' own pay and perks. Also called seller's discretionary earnings (SDE); buyers of small businesses often price them as a multiple of it.

P

p-value: The probability of getting a result at least as extreme as the observed one if H0 were true. It is not the probability that H0 is true.
P&L (income statement): The report of revenue, costs and profit for a period, built on the accrual principle: it shows what was earned, not what arrived in the bank.
painkiller vs vitamin: A test of urgency: a painkiller solves a problem people need fixed now; a vitamin is a nice-to-have they'll postpone.
paired (related) samples: Measurements that come in matched pairs, typically the same users before and after a change.
paired t-test: Tests whether the mean of before-after differences is zero: t = mean difference / (SD of differences / sqrt(n)).
parameter: A true (usually unknown) property of the population, such as the population mean.
parametric test: A test that assumes a distribution (usually normal) and works with means and variances, e.g. the t-test or ANOVA.
past specifics: Concrete facts about something that already happened ("Tell me about the last time…"); the most reliable kind of answer in a customer conversation.
Pathos Problem: When you visibly expose your hopes or ego ("I quit my job for this, be honest"), people protect your feelings instead of telling you the truth.
pay cycle: The time from one payday to the next. Tracking and planning by pay cycle, not by calendar month, shows whether the money lasts.
pay yourself first: Moving a set amount to savings automatically on payday, before any spending, instead of saving what is left at the end of the month.
payback period: How long until a cohort's cumulative gross profit per user covers its acquisition cost.
paywall: The screen that asks a user to pay or start a trial to use a feature. Its conversion depends on what triggered it as much as on its design.
Pearson correlation coefficient (r): Measures the strength of a linear relationship from -1 to +1; sensitive to outliers.
Pearson's chi-square test: Tests association between categorical variables by summing (observed - expected)^2 / expected over all cells.
peeking (optional stopping): Checking results repeatedly and stopping as soon as p < 0.05; inflates the false-positive rate far above alpha.
perceived vs real problem: The difference between what a customer says is wrong ("we need better messaging") and the underlying need that asking "why?" reveals ("we need everyone on the latest file").
percentage point (pp): The unit of an absolute difference between two percentages: 4% to 5% is +1 pp but +25% relative.
percentile: The value below which a given percentage of observations falls, e.g. the 90th percentile of page load time.
pie chart: A circle split into sectors proportional to shares of the whole; readable only with a few categories.
pitching: Slipping into selling or defending your idea during a learning conversation; the moment you pitch, they stop talking about their problems.
placebo: A dummy treatment that looks real; separates the effect of the change from the effect of expecting a change.
polynomial (non-linear) regression: Regression that includes squared or higher-power terms to fit curved relationships, e.g. diminishing returns on ad spend.
Ponzi scheme: A fraud that pays earlier investors with money from new ones and promises high, steady, "guaranteed" returns until it collapses.
population: The whole group we want conclusions about, e.g. all current and future visitors of the shop.
positioning: How a product is placed in customers' minds: compared with which alternatives, for whom, and why it is better for them. Good positioning names the category and the difference in the customer's own words.
post hoc test: A pairwise comparison run after a significant ANOVA with control of multiple comparisons (Tukey, Bonferroni, Games-Howell).
practical significance: Whether an effect is large enough to matter for the business, judged by effect size and cost, not by p-value.
pre-order: Paying for a product before it exists (as on Kickstarter); the clearest proof that "I would buy it" is true.
pre-release: A test version published before the final one, such as `2.0.0-beta.1`. On GitHub a release can be marked as a pre-release, and then it is never shown as the latest release.
pre-sale: Selling a product before it is built: the customer pays now, the product starts on a stated date, and the money is returned in full if it does not. The strongest test of a solution, because it costs the customer something real.
precision and recall: Precision: share of predicted positives that are correct. Recall: share of actual positives that were found.
premature zoom: Jumping into details of a specific problem before checking that the person cares about that area at all; it produces plausible but misleading answers.
present value: What a future amount is worth today: the amount divided by (1 + discount rate) raised to the number of years.
price anchoring: Presenting a reference number before or next to your price so customers judge the price against it: for example, what the problem costs them today, or the monthly price next to a cheaper yearly one.
price, volume and mix variances: A split of a revenue or margin change into the part caused by different prices, by selling more or fewer units, and by a shift between products with different margins.
pricing tiers: Several versions of a product at different prices, usually two to four, each aimed at a different kind of customer. Tiers let small customers start cheaply and larger ones pay more for more value.
primary metric: The one metric, chosen before launch, that a test's decision depends on.
primary metric (OEC): The single metric, chosen before launch, that decides whether a test wins.
principal component analysis (PCA): Rotates the data to new uncorrelated axes ordered by explained variance; a common way to reduce dimensions.
prior and posterior probability: The prior is what we believe before seeing the data; the posterior is the belief after updating on the data.
prioritization framework: An explicit, repeatable method for ranking themes or features against agreed criteria (value, effort, risk, confidence), so priority decisions can be explained and challenged instead of being made on gut feeling.
product analytics: The practice of measuring what people actually do inside a product, from the events they trigger, and turning those numbers into decisions about what to build, fix, price and grow.
product backlog: An ordered list of concrete work items (user stories, bugs, tasks) for the development team. It is an execution tool, far more detailed and short-lived than a roadmap.
product life cycle: The stages a product passes through, from launch through growth and maturity to decline or end of life. The stage shapes what the roadmap emphasizes and how far ahead it can sensibly plan.
product risk: Uncertainty about whether you can build the product and make it grow; conversations alone can't remove it, e.g. "I'll pay if you bring me customers".
product roadmap: A strategic communication tool that shows how a product will move towards its vision: which customer problems and business objectives the team will tackle, in roughly what order, and why. It states intent and direction, not a delivery contract.
product strategy: The chosen path from the product vision to results: which customers and problems to focus on, which bets to make and which objectives will show progress. The roadmap is how that strategy is sequenced and communicated.
product vision: A short statement of the future the product aims to create: who benefits, from what change, and why it matters to them and to the company. It is the anchor every roadmap item should trace back to.
product-market fit: The state in which a product satisfies a clearly defined market so well that customers buy, stay and recommend it with little pushing. It shows up in the numbers: sales get easier, churn falls and new customers arrive by word of mouth.
product-market fit: The state where a product satisfies a real market's need well enough that users keep using it and recommend it without being pushed. Seen in retention, not in sign-ups.
project board: A GitHub Projects view that shows issues and pull requests as cards in columns or as a table, for planning and tracking work.
proportion: The share of observations with a yes/no property; its standard error is sqrt(p(1 - p) / n).
pull request: A GitHub page that proposes merging one branch into another. It gathers the description, commits, changed files, automated checks and review discussion in one place.
purchasing criteria: The factors a customer actually uses to choose and pay for a solution; far more useful than a list of wished-for features.
push: Sending new local commits from a computer to the repository on GitHub. When you edit in the browser, GitHub saves the commit there directly.

Q

qualifying action: The action that makes a user count in a metric, such as opening the app or adding an expense. 'Any event' is rarely the right choice.
quality metric: A metric that shows how well the product works for each user or group it already has, such as retention, activation or expenses per active group. It should not move just because more users arrived.
quantitative (continuous) variable: A variable measured on a numeric scale where arithmetic makes sense, e.g. revenue or time on page.
quartile: One of three cut points (Q1, Q2 = median, Q3) that split sorted data into four equal parts.

R

ramen profitability: The point where a young business earns just enough to cover the founders' basic living costs. It does not make anyone rich, but it removes the deadline set by the runway.
randomisation: Assigning units to groups by chance, so groups are comparable and confounders are balanced on average.
randomisation unit: The entity assigned to a variant (user, session, device, shop); analysis should be done at the same level.
range: The difference between the largest and the smallest value; very sensitive to outliers.
rank: The position of a value in the sorted data; rank-based tests compare positions instead of raw values.
README: A file (usually `README.md`) that GitHub shows under the file list on the repository page. It explains what the project is and how to work with it.
real annual cost of credit: The yearly cost of a loan with all interest, fees and commissions included. It is the only fair way to compare loans; Ukrainian lenders must disclose it.
real return: A return after inflation: what your savings gain in purchasing power, (1 + nominal) ÷ (1 + inflation) − 1.
rebase: Replaying a branch's commits on top of another branch as new commits, so history stays a straight line (`Rebase and merge` on GitHub).
referral program: A system that rewards customers for bringing in new customers, often with a discount or free month for one or both sides. It works best when sharing the product makes the customer look good or helps them, not when it helps their competitors.
regular investing (dollar-cost averaging): Investing the same amount at regular intervals whatever the price, which removes the need to guess the right moment.
release: A published version of the product. On GitHub a release is a page built on a tag, with notes about what changed.
release plan: A tactical plan of which solutions ship when, with scope, dates and capacity. It answers "how and when" for work the roadmap has justified with "what and why".
relevant cost: A future cost that differs between the options being compared. Only relevant costs and revenues belong in a decision.
repeated-measures ANOVA: ANOVA for three or more measurements on the same subjects; removes between-subject variability from the error.
reporting time zone: The time zone that decides which calendar day an event belongs to. Logs are usually in UTC; a user's local day can differ, which changes daily metrics.
repository: A project folder together with its complete history of changes. On GitHub it has its own page with files, commits, branches and pull requests.
representative sample: A sample whose composition mirrors the population, so results generalise; random sampling helps achieve it.
reputation risk: A commitment where the customer stakes their name on you: introductions to peers or decision makers, a public testimonial or a case study.
request changes: A review verdict that asks the author to change something before merging. Shown as `Changes requested` in red.
residual: Observed value minus the value predicted by the model.
retention: The share of customers (or revenue) that stays over a period; for one month it equals 1 − churn. High retention is what makes a subscription business grow instead of refilling a leaking bucket.
retention plateau: The level where a cohort's retention curve stops falling: the users who stay for good. A plateau is a strong product-market-fit signal; a curve that keeps falling to zero is a warning.
retention rate: The share of users who are still active after a given period, e.g. day-30 retention.
revenue: What the company earned from its own sales in a period, recognised when it delivers what was paid for. For a marketplace acting as an agent this is its commission, not the whole amount the customer paid.
revenue churn: The share of MRR lost in a period to cancellations and downgrades: lost MRR ÷ MRR at the start. It can differ a lot from customer churn when customers pay different amounts.
revenue recognition: The rules for when money from a sale becomes revenue: when the promised lesson or service period is delivered, not when the customer pays.
revert: A new commit that undoes an earlier one. History keeps both, so nothing is erased. GitHub offers a `Revert` button on merged pull requests.
RICE scoring: A prioritization formula that scores each idea as Reach × Impact × Confidence ÷ Effort, giving a comparable number for how much value it delivers per unit of work.
right censoring: Users who entered recently haven't had the full window to convert yet, so their conversion looks lower than it will be. Recent cohorts must be excluded or marked incomplete.
roadmap alignment: A shared understanding among stakeholders of where the product is going, why, and what their part is. Aligned people can disagree on details yet act in the same direction; alignment does not require consensus.
roadmap co-creation workshop: A structured session in which key stakeholders build or finalize the roadmap together, typically covering hopes and fears, vision and goals, working back from the desired future, and sizing and prioritizing themes.
roadmap disclaimer: An explicit note that the roadmap reflects current plans and may change without notice. It protects the team from accusations of broken promises and warns customers not to base purchases on future items.
rolling forecast: A forecast updated every month or quarter that always looks the same distance ahead, instead of a fixed annual budget that goes stale.
rolling retention: The share of a cohort active on day N or any later day. It is always higher than day-N retention and keeps changing as new data arrives.
rules for honest answers: A set of rules for asking questions so that even someone who wants to please you (a friend or relative, say) can't give you misleading answers: talk about their life, ask about past specifics, listen more than you talk.
runway: How many months the founders or the business can keep going before the money runs out, at the current level of spending and income. Usually calculated as available cash divided by the net monthly burn.

S

SaaS (software as a service): Software that customers use online and pay for with a regular subscription instead of buying and installing it once.
SAM (serviceable available market): The part of the total market that your product and sales channels can actually serve, for example only shops in regions and segments you can reach.
sample: The subset of the population that we actually observe or measure.
sample ratio mismatch (SRM): Variant sizes that differ from the planned split more than chance allows. It usually means a bug in assignment or logging, and the test result can't be trusted until it's explained.
sample ratio mismatch (SRM): When the observed split between groups differs significantly from the planned one (e.g. 50/50); signals a broken experiment.
sample size: The number of observations (per group in a test); planned in advance from alpha, power, baseline and MDE.
sampling bias: A systematic difference between sample and population caused by how observations were selected.
sampling distribution: The distribution of a statistic (e.g. the mean) across many hypothetical samples of the same size.
saturation: The point where new conversations with a segment stop producing new information; often after 3–5 talks with a well-defined segment.
scary question: An important question you instinctively avoid because the answer might hurt the idea (price, budget, legal blockers); every conversation should include at least one.
scatter plot: A chart with one point per observation, placed by the values of two quantitative variables.
scenario analysis: Building several consistent versions of the future (for example base, weak and strong), each with its own set of assumptions, and comparing the results.
scree plot: A plot of eigenvalues by factor number; factors after the 'elbow' are dropped (Cattell's criterion).
seasonality: A repeating pattern tied to the calendar, such as weekday effects or holiday peaks.
secondary metric: A metric read to explain a test result, never to decide it.
segmentation: Splitting users or customers into groups that behave differently, so each can be targeted differently.
selection bias: A difference between groups created by who ends up in each group, not by the thing being compared. 'Users who use receipt scanning retain better' is a classic case.
self-funding: Paying for a new business with the founders' own money, such as savings or a personal loan, instead of selling a share to investors.
semantic versioning: A version format MAJOR.MINOR.PATCH: breaking changes raise MAJOR, new features raise MINOR, fixes raise PATCH.
sensitivity analysis: Changing one assumption at a time to see how much the result moves, to find the assumptions a decision really depends on.
sequential testing: Test designs with adjusted thresholds that allow valid interim looks at the data.
server-side event: An event logged by the backend rather than by the app on the user's device, for example a recurring bill created automatically. It can happen without the user doing anything.
settle-up: A payment that clears a debt between two members of a Halves group. The payer records it and the receiver confirms it, so one settle-up can appear as two events.
shuttle diplomacy: Building alignment by meeting stakeholders one-on-one before any group session: learning each person's goals and constraints, scoring their ideas against shared objectives and bringing a pre-negotiated draft to the room.
sign test: Compares the number of positive and negative before-after changes, ignoring their size.
significance level (alpha): The threshold for rejecting H0, fixed before the test, usually 0.05; equals the accepted false-positive rate.
Simpson's paradox: A trend that appears in every subgroup reverses when the groups are combined, because group sizes differ.
single tax: The simplified tax a Ukrainian sole proprietor pays instead of income tax: a fixed monthly amount in some groups or a share of income in others.
sinking fund: Saving a little every month for a known future cost, such as a holiday or car repair, so it doesn't land on a card.
skewness (skewed distribution): Lack of symmetry in a distribution; with a long right tail (typical for revenue) the mean is above the median.
slope (regression coefficient): b1: how much the predicted outcome changes when the predictor increases by one unit.
social engineering fraud: Fraud that tricks people into giving codes, card details or money, for example a fake call from "the bank's security service".
solution validation: Checking, before building a product, that a proposed way of solving the problem fits the customers' real workflow, beats what they use now and does not create new problems.
SOM (serviceable obtainable market): The share of the serviceable market you can realistically win in the next few years, given competition and your own capacity. It is the number to compare with your revenue goal.
SOP (standard operating procedure): A written, step-by-step description of a recurring task, with the reason behind it and known pitfalls, detailed enough that someone new can do it without asking the founder.
Spearman rank correlation (rho): Pearson's r computed on ranks; captures monotonic relationships and is robust to outliers.
spending plan: A plan for the month that splits income into essentials, saving and flexible spending, including money for things you enjoy. Unlike a strict budget, it is meant to be lived with.
spike investigation: Explaining a sudden jump or drop in a metric by listing possible causes (tracking, traffic, product, outside events) and ruling them out one by one with the data already available.
squash merge: Merging a pull request as one new commit on the target branch that contains all its changes (`Squash and merge` on GitHub).
stage of development: A label showing how far a roadmap item has progressed, e.g. discovery, design, in development, beta, released. It tells stakeholders what is still an open problem and what is already a concrete solution.
stakeholder: Anyone who influences whether your product gets bought or used, such as users, buyers, IT, legal or partners; missing one can quietly block a deal.
stakeholder (roadmap stakeholder): Anyone inside or outside the company who shapes, funds, builds, sells, supports or depends on the product and so has a stake in the roadmap: executives, sales, marketing, support, engineering, finance, partners and customers.
stalling (brush-off): A polite way of ending the conversation without committing: "Keep me posted", "Let me know when it launches", "Don't call us, we'll call you".
standard deviation (SD): The square root of the variance; spread expressed in the same units as the data.
standard error (SE): The standard deviation of a statistic's sampling distribution; for a mean it is SD / sqrt(n).
standardisation (scaling): Converting variables to z-scores so that variables on large scales do not dominate distances.
state pension: The pension paid from the solidarity system to people who reach retirement age with enough years of insurance record.
static vs dynamic model: A static model describes a state at one moment; a dynamic model describes change over time.
statistic (estimate): A number computed from a sample that estimates a population parameter, e.g. the sample mean.
statistical power: The probability that a test detects a real effect of a given size; 1 - beta, usually planned at 0.8.
statistical significance: A result is significant when p < alpha; it says the effect is unlikely to be pure noise, not that it is large or important.
status checks: Automated jobs (tests, builds, linters) that run on a pull request's commits and report passed, failed, pending or skipped.
step conversion: Users who reached a funnel step divided by users who reached the previous step. Multiply step conversions to get the whole funnel.
step cost: A cost that stays flat over a range of volume and then jumps to a new level, such as one more support coordinator for every few hundred learners.
stepwise regression: Builds a model by adding or removing predictors one at a time based on their contribution.
stickiness: DAU divided by MAU (or WAU by MAU): the share of monthly users who show up on a typical day or week. Only meaningful at the product’s natural usage rhythm.
strategic buyer: A buyer, often a larger company in a related market, who acquires a business for its customers, technology or position and is likely to change the product and ask the founders to stay on.
Student's t-test (independent samples): Compares the means of two independent groups: t = (mean1 - mean2) / SE of the difference.
subscription creep: Small recurring payments that pile up unnoticed until together they cost more than one noticeable purchase.
subtheme: A narrower need within a theme. Subthemes split a broad problem into parts the team can explore separately, and they are also where a known or probable solution can be noted without losing the "why".
suggested change: A review comment that proposes exact replacement text for a line. The author can apply it with one click (`Commit suggestion`).
sunk cost: Money already spent that no decision can bring back. It must not influence the choice about what to do next.
synthesis: Combining notes from several conversations into patterns and decisions, rather than acting on any single customer's words.

T

T-shirt sizing: A quick, relative effort estimate using sizes XS, S, M, L and XL (often mapped to 1–5) instead of days or story points. It is precise enough for prioritizing without implying a schedule.
tag: A permanent name attached to one commit, most often a version number such as `v2.4.0`.
TAM (total addressable market): The total yearly revenue available if every potential customer of the kind you serve bought your product: number of potential customers × yearly price.
term deposit: Money placed with a bank for a fixed period at a fixed rate; taking it out early usually costs part of the interest.
term life insurance: Insurance that pays a sum to your dependants if you die within the term. It protects the family and is kept separate from saving.
test assumptions: Conditions a test needs to be valid, e.g. independent observations, approximate normality, equal variances, no extreme outliers.
test statistic: A number computed from the data (t, U, F, chi-square...) that is compared with its distribution under H0.
theme (roadmap theme): A high-level customer need, problem or job that groups related work on the roadmap, phrased as a result rather than a solution, e.g. "make reordering effortless for regular buyers".
time commitment: Investing real time in your solution: a follow-up meeting with clear goals, reviewing wireframes, or using a prototype for a meaningful period.
time series: Observations of a metric ordered in time, e.g. daily orders; often has trend and seasonality.
time to value: How long it takes a new user to reach activation. Shorter is usually better, and a median tells more than a mean here.
time value of money: A hryvnia today is worth more than a hryvnia next year, because it can be invested or used now and because future money is uncertain.
time window: The period a metric covers, such as one calendar day or seven days ending yesterday, together with the time zone that decides where each day starts.
timeframe (time horizon): The broad time bucket a roadmap item sits in: a quarter, a half-year, or Now / Next / Later. Broad buckets show sequence and urgency without promising exact ship dates.
tracking plan: A document that lists the questions a feature must answer, the events and properties needed to answer them, where each is logged and how it will be checked before launch.
traffic mix: The shares of new users from each source (invites, store search, paid ads, partner promos). When the mix changes, averages change even if nothing in the product did.
train/test split (holdout): Fitting a model on one part of the data and evaluating it on unseen data to estimate real performance.
treatment (experimental) group: The group that gets the change being tested (the variant).
trial-to-paid conversion: The share of users who start a free trial and then make a first payment.
triggered enrollment: Counting a user in a test only when they reach the point where the variants differ, such as opening the paywall. It removes users the change can't affect.
trimmed mean: The mean computed after dropping a fixed share (e.g. 5-10%) of the smallest and largest values.
truncated axis: A value axis that does not start at zero; it exaggerates small differences in bar charts.
trunk-based development: A way of working where everyone merges small changes into the main branch at least once a day, with very short-lived branches or none at all. Unfinished work is hidden behind feature flags.
Tukey's HSD test: A common post hoc test comparing all pairs of groups when variances are equal.
two-proportion z-test: Compares two conversion rates using the normal approximation: z = (p1 - p2) / SE of the difference.
two-way (multifactor) ANOVA: ANOVA with two or more factors that tests each main effect and their interaction.
type I error (false positive): Rejecting a true H0: declaring an effect that does not exist. Its probability is alpha.
type II error (false negative): Failing to reject a false H0: missing a real effect. Its probability is beta.

U

unified social contribution (ESV): The mandatory social contribution that funds the state pension and counts toward a person's insurance record.
unique users: The number of distinct people who did something in a period, however many times each did it. Compare with the event total.
unit counted: What a metric counts: distinct users, distinct groups, events or settlements. Changing the unit changes the number even on the same data.
unit economics: The revenue and costs of one unit of the business, usually one customer: what it costs to win them, what they pay, what it costs to serve them and how long they stay. It shows whether each new customer makes the business richer or poorer.
user churn: Users who stop using the product. For free products it is inferred from inactivity, so its definition (how long inactive) must be stated.
user journey map: A step-by-step picture of how a person solves a problem today, from noticing it to moving on, often with their feelings at each step. Painful steps are candidates for roadmap themes.
user property: A detail about the user that holds across events, such as sign-up date, platform or acquisition channel. Used to segment users.

V

valuation multiple: The number by which a financial figure, such as ARR or yearly profit, is multiplied to estimate a company's price. It goes up with growth, low churn and independence from the founder, and down with risks.
value metric: The unit that grows when a customer gets more value from the product, such as bookings or cars served per month. Tying prices or plan limits to it lets revenue grow together with customer success.
value proposition: A concise statement of who the product is for, which need it meets, what benefit it delivers and how it differs from the alternatives. It is often written with a fill-in template (For… who… the… is a… that… unlike…).
value vs effort scoring (ROI scorecard): A scoring method that rates each item's contribution to customer needs and business objectives, divides the total by estimated effort and often multiplies by confidence: priority = value ÷ effort × confidence.
vanity metrics: Numbers that look impressive and usually only go up, such as total downloads or sign-ups, but don’t tell the team whether users get value or what to do next.
variable: A measurable property of each observation, such as order value, session length or whether a user converted.
variable cost: A cost that grows with every unit sold or delivered, such as a tutor's payout per lesson or a payment fee per purchase.
variance: The average squared deviation from the mean; the sample version divides by n - 1. Measured in squared units.
variance analysis: Comparing actual results with the plan and splitting the difference into its causes, such as price, volume, mix and currency.
venture capital: Money from professional investors who buy a share of a young company and expect very fast growth, because they count on a few big winners to cover many failed bets.
verbatim quote: A customer's exact words written down in quotation marks; useful as evidence for the team and as language for marketing.
version control: A way of keeping every saved state of a project, with who changed what, when and why, so any earlier state can be viewed or restored.
vesting: Earning your shares in a company gradually over time, usually over three or four years, instead of owning them all on day one. If a founder or employee leaves early, the shares they haven't earned yet go back to the company.
viral coefficient (k-factor): The number of new users each user brings on average: invitations per user × share of invitations that turn into new users. Below 1, virality amplifies other channels but can't drive growth alone.
viral cycle time: The time from a user joining to the users they invited joining. Shorter cycles make the same k grow much faster.
Vision, Framing, Weakness, Pedestal, Ask: A five-part way to request a meeting: the problem you want to solve (without your idea), your stage and that you're not selling, where you're stuck, why this person can help, and a clear ask.
volatility: How much an investment's value swings up and down. Higher expected returns come with bigger swings.
volume metric: A metric that counts how much: users, groups, expenses or revenue in total. It can rise because of ad spend or a season while the product itself gets no better.
voluntary pension fund: A private fund where a person or employer adds to retirement savings on top of the state pension.

W

warm intro: An introduction through someone both sides trust, which gives you instant credibility; the main source of good conversations after the first few.
weekly active groups: Distinct groups in which at least two members added an expense (not server-created) or recorded a settle-up in the week; the north star metric of Halves in this course.
weighted mean: A mean in which each value counts in proportion to its weight, e.g. average conversion across segments weighted by traffic.
Welch's t-test: A version of the two-sample t-test that does not assume equal variances; a safe default.
who-where pair: A segment described as a specific group plus a place (physical or online) where you can reliably find them; if you can't name the "where", keep slicing.
Wilcoxon signed-rank test: The non-parametric alternative to the paired t-test: ranks absolute differences and compares sums for positive and negative shifts.
within-group variance: The variability of observations around their own group mean; the 'noise' in ANOVA.
word of mouth: Customers telling other potential customers about a product on their own, in conversation, chats or communities. In a close-knit niche it is often the cheapest and most trusted way new customers arrive.
workaround: A makeshift way a customer already copes with a problem (spreadsheets, interns, manual steps); a strong sign the problem is real and a benchmark for price.
working capital: Money tied up in day-to-day operations: what customers owe you plus other current assets, minus what you owe suppliers and customers who prepaid. Prepayments can make it negative, which funds growth.
WSJF (weighted shortest job first): A sequencing rule that divides cost of delay by job size and takes the highest ratio first, so short, urgent, valuable items ship before long items of similar value.

Z

z-score (standard score): How many standard deviations a value is from the mean: (x - mean) / SD.
zombie lead: A prospect who keeps taking meetings and saying nice things but never commits or says no; they keep you in the friend zone and waste your time.