The CRO glossary, plainly written.

Two versions of a page, shown to two random user groups, to find out which converts better.

The default experiment shape in CRO. Half of traffic sees variant A (the control), half sees variant B (the challenger). After a sufficient sample, you can claim with confidence whether B is genuinely better, worse, or indistinguishable from A. What makes A/B data unusually reliable is that it is double-blind: visitors do not know they are in a test, so they act naturally rather than performing for an audience. This separates it from surveys and focus groups, where the act of being observed changes the behaviour being measured. Most A/B tests do not produce a winning variant — a win rate of around one in three is normal in well-run programmes. The learning from non-winning tests is still real: a flat or negative result confirms the hypothesis was wrong or the change was not the right solution, which narrows the problem space for the next test.

The part of the page a user sees before scrolling.

A holdover from print, still useful as a forcing function. The above-the-fold area has to answer three questions in five seconds: what is this, who is it for, and why should I care. If it doesn't, the rest of the page is mostly decoration.

How you assign credit to channels and touchpoints for a conversion.

CRO needs attribution to know whether a test won. The simplest models — last-click, first-click — are usually wrong in interesting ways. Multi-touch attribution is closer to truth but harder to instrument. Pick a model, document it, and don't silently change it mid-experiment.

The first number a user sees becomes the yardstick they judge every other number against.

When the page shows ₹2,000 crossed out next to ₹1,200, the user isn't really judging ₹1,200 — they're judging the gap. Anchoring is why "list price + discount" still works. It also fires the wrong way: a free tier with a generous limit anchors the paid tier as "expensive" before the user reads what's in it.

A statistical approach that updates probability as data comes in.

An alternative to frequentist (p-value-based) testing. Bayesian tools give you "the probability that B is better than A" rather than "the probability we'd see this if A and B were the same." Often more intuitive. Adopted by VWO, Optimizely, and others as the default mode.

The share of sessions that end on the first page with no interaction.

A noisy metric — high bounce can be good (people got the answer fast) or bad (people didn't find what they wanted). In CRO, treat bounce as a question, not an answer. Pair it with scroll depth and time-on-page.

The button, link, or form that asks the user to do the thing.

The single most-tested element on most landing pages, and usually the wrong place to start. Copy and contrast matter, but a CTA only works if everything above it has earned the click. Fix the page's argument first; tune the CTA last.

The range your true lift is likely to sit inside.

A test reporting "+15% lift" without a confidence interval is reporting half the story. A real result reads "+15%, 95% CI [+8%, +22%]." If the interval crosses zero, you don't actually have a winner.

The percentage of visitors who completed a defined action.

The currency of CRO. Always defined relative to a specific event — sign-ups per session, trials per click, paid per trial. Whenever someone reports "our conversion rate," ask: from what, to what, measured over what window.

The practice of using evidence from real visitors to improve how well a website achieves its goal.

CRO is not a software tool or a single technique — it's a way of making decisions. Instead of changing a website based on opinion or preference, CRO practitioners observe what visitors actually do, form a hypothesis about why something isn't working, test a change, and measure the result. The loop repeats indefinitely. Done well, CRO shifts a team's culture from 'I think we should change X' to 'here's what the data says, and here's how we'll test it.'

Too many options on the same screen, and the user defaults to none.

Past about five comparable choices, people start delaying or bailing. Hick's law puts a number on it — decision time grows with the log of options. On pricing pages this shows up as "three tiers convert, four tiers stall." On homepage heroes it shows up as a primary CTA plus three secondaries diluting all four.

We see the data that agrees with what we already think.

What kills a CRO program isn't lack of data — it's the team finding the slice of data that flatters their existing belief. A test that "won" because we cherry-picked the segment where it worked is confirmation bias dressed as analytics. The defense is pre-registering the hypothesis and the success metric before the test runs.

Add a clearly worse third option and the option next to it suddenly looks like a deal.

The classic "small ₹120, medium ₹250, large ₹260" — the medium is the decoy that makes the large feel obvious. On SaaS pricing pages, the decoy is usually a middle tier deliberately under-spec'd to push buyers up to the premium tier. Powerful, also obvious to a sophisticated buyer — use sparingly or you look like a casino.

An organizational habit of making decisions through structured tests rather than authority or instinct.

A team with an experimentation culture doesn't argue about what will work — they agree on a hypothesis and run a test. This sounds simple but requires two things most organizations resist: the willingness to be wrong publicly, and the patience to wait for a result instead of acting on a hunch. Building this culture is arguably harder than learning any CRO technique — and more valuable.

Anything that slows the user down, in their head or on the page.

Friction comes in four flavors: cognitive (too much to read), interactive (too many clicks), trust (why should I believe you), and technical (slow, broken, weird). Most CRO work is friction removal — and most "redesigns" add more of it than they take away.

The ordered set of steps a user takes from arrival to conversion.

Mapping the funnel is the first move of any CRO engagement. You can't improve what you can't measure stepwise. A typical SaaS funnel: landing → signup form → email confirm → activation → first value. Each step has its own rate, its own drop-off, its own audit.

A test that shows users a page and asks where they'd click first to complete a task.

Participants are shown a design and given a specific task — 'where would you go to start a free trial?' The test records what they click, how long they took, and whether they hit the intended target. High miss rates reveal CTA placement or labeling problems before a single line of A/B test code is written. One of the cheapest ways to validate whether a page's visual hierarchy is doing what you intend.

A design is shown for five seconds, then hidden — participants answer questions about what they saw.

Tests first impressions and whether a page communicates its core message in the time most visitors actually spend before deciding to stay or leave. Standard questions: 'What does this company do?', 'What is the main offer?', 'What do you remember most?' If most participants cannot answer these after five seconds, the headline and hero are not doing their job. Widely used for landing pages, homepage heroes, and ad creative.

The same fact, said differently, converts differently.

"Save 10% if you pay annually" and "Pay 10% more for monthly" describe the same offer. The first frame consistently outperforms the second. Framing is the single most under-used lever in CRO because it's "just words" — but the words are the page.

The closer you feel to the finish, the harder you push to get there.

A loyalty card with two boxes pre-stamped converts higher than the same card with zero stamps, even though both require the same total purchases. On signup flows this is why "Step 2 of 3" outperforms "Step 2" — the visible finish line moves the body. Works in onboarding, multi-step forms, content series.

A visualization of where users click, move, or scroll on a page.

Useful for spotting two things: where attention concentrates that shouldn't (rage clicks on non-clickable elements) and where attention misses what should be central. A bad heatmap reader sees patterns where none exist; a good one cross-references with funnel data.

The top section of a landing page — headline, subhead, primary CTA, visual.

The load-bearing wall of any page. Most landing-page lift comes from rewriting the hero, not redesigning everything below it. If the hero is wrong, every other change is rearranging the same furniture.

The more choices you give someone, the longer it takes them to decide — and the more likely they are to decide nothing.

Named after psychologist William Hick, who demonstrated in the 1950s that decision time increases predictably as the number of available options grows. In CRO, this shows up on every page that has too many competing CTAs, an overloaded navigation, or a pricing structure with too many tiers. The fix is not fewer pages or less information — it is fewer simultaneous choices. One prominent CTA. A clear visual hierarchy that makes primary and secondary actions instantly distinguishable. A pricing page that recommends one tier rather than expecting every visitor to self-select. Managing the number of choices a visitor faces at any one moment is one of the most direct levers for reducing friction and increasing conversion rate.

A decision made by the most senior person in the room rather than by evidence.

HiPPO decisions feel safe because they come with authority. They're not. When a senior stakeholder says 'I don't like the homepage' and the team redesigns it without data, that's a HiPPO in action. The term isn't about blaming leaders — it's about recognizing a structural pattern where seniority overrides evidence. CRO teams use shared data and pre-agreed hypotheses to make it easier to push back: not 'you're wrong,' but 'let's test it.'

A written, testable claim that links a specific change to a specific outcome.

The unit of work in CRO. The simplest reliable format: 'If I [make this specific change], I expect [this specific outcome], as measured by [this specific metric].' The template forces three things in one sentence: a commitment to what you're actually changing, a prediction instead of a hope, and a named metric so there's no argument about what winning looks like. Without a hypothesis in this form, you're not running an experiment — you're running a guess with extra steps. One useful side-effect: writing the hypothesis often reveals that a vague idea ('refresh the website') contains three or four genuinely distinct testable ideas.

A small reward today beats a bigger reward later, every time.

A free 14-day trial today wins over "50% off if you commit for a year." Humans dramatically over-discount future benefit. On landing pages, the lever is moving the user's first reward earlier — "see results in 5 minutes" outperforms "see results in 30 days" even when the 30-day version is objectively better.

A prioritization scoring of Impact × Confidence × Ease.

A quick-and-dirty way to rank your hypothesis backlog. Score each from 1–10, multiply, sort. ICE is biased toward "ease" — high-confidence small wins float to the top, and big bets get under-ranked. Use it as a starting point, not as the answer.

We over-value things we helped build.

The user who configured their own dashboard, picked their own theme, named their own workspace — they retain at scale. Onboarding flows that ask for any tiny customisation upfront beat flows that auto-pick everything. The same principle is why "Build your custom plan" outperforms "Choose a pre-built plan" at the top of the funnel.

A page built to do one job — convert a specific audience on a specific offer.

Not the same as the homepage. A landing page has one audience, one ask, no distractions. The classic mistake is treating it like a homepage with a banner — leaving every nav link, every secondary CTA, every "while you're here" intact.

The relative change in conversion rate between the control and the variant.

If control is 2.0% and variant is 2.4%, that's a +20% lift, not a +0.4-point lift. The distinction matters because compounding is multiplicative; a 20% lift on the next test compounds with the last.

Losing ₹100 hurts about twice as much as gaining ₹100 feels good.

Loss-framed copy outconverts gain-framed copy for most actions that require effort. "Don't miss your free audit" beats "Get your free audit." "Avoid the GA4 migration mistakes" beats "Learn GA4 migration best practices." Use sparingly — over-using it tips into FUD and the page starts feeling cheap.

The main outcome you care about — a purchase, a trial, a qualified lead.

As opposed to micro-conversions (newsletter sign-up, video play). CRO programs that only optimize macros miss insight; programs that only optimize micros optimize for vanity. Track both, but never confuse them.

The smallest lift your test can reliably detect given your traffic and conversion rate.

The single most useful number to compute before running a test. If your MDE is 30% and your hypothesized lift is 10%, you cannot validate this test in any reasonable time — and you shouldn't run it. MDE forces honesty about what your sample can actually prove.

A smaller, on-page action that signals interest — clicks, scrolls, video plays.

Useful as leading indicators when macro-conversion volume is too low to test on. The trap is optimizing for micros that don't correlate with macros — celebrating a 30% lift in "scroll past the fold" while signups stay flat.

The tiny text on buttons, form labels, error messages, tooltips.

Punches well above its weight. Changing a CTA from "Submit" to "Get my free audit" routinely moves conversion 10–20%. Microcopy is the cheapest CRO lever in the toolkit — and the one most teams ignore.

Familiar things feel more trustworthy, even without new evidence.

This is why retargeting works even when the ads aren't great — the user has now seen your logo five times and "feels like they've heard of you." For CRO, mere exposure is why the third visit to a pricing page converts higher than the first. The implication: design your landing pages to be revisited, not just visited.

The tendency to mistake the excitement a new idea generates internally for evidence that it will work with visitors.

New designs, bold creative, and unexpected layouts generate genuine enthusiasm in review meetings. That enthusiasm feels like a signal. It isn't. Your visitors weren't in the brainstorm — they arrive at a page with no shared context, no investment in the process, and no reason to respond well to something just because it's new. Novelty bias is especially dangerous in redesigns: the team's excitement about what changed is often inversely correlated with the visitor's appreciation of it.

The probability of seeing your result if there were no real difference.

In frequentist A/B testing, the standard cutoff is p < 0.05 — but the cutoff is a convention, not a law. The bigger trap is "p-hacking": peeking at results, stopping the test when you like the number, and reporting that as significant. Pre-commit to a sample size; resist the urge to peek.

A prioritization framework — Potential, Importance, Ease.

An older sibling of ICE. Same idea, slightly different vocabulary. Use whichever your team already knows; the framework matters less than the consistency.

A test that asks participants which of two designs they prefer.

Participants are shown two designs and asked which they like more. The result is a preference — not a conversion prediction. Because people's aesthetic preferences rarely align with what actually makes them take action, preference tests are a poor guide to conversion decisions. They are occasionally useful for early-stage directional work or stakeholder communication, but should not be used as a substitute for tests that measure behaviour. If you want to know which design converts better, run an A/B test. If you want to know which design people prefer, a preference test will tell you — just don't confuse the two.

People remember the most intense moment and the final moment — not the average.

A user's memory of an experience is anchored on its emotional peak (good or bad) and how it ended. On a SaaS trial, the *last* day matters disproportionately — a great moment in the final 24 hours can save a trial that was mostly forgettable. On onboarding, the final confirmation screen carries more weight than the middle steps.

Direct user observation — interviews, usability tests, session replays.

Quantitative tells you what; qualitative tells you why. Five well-run user interviews will sharpen a hypothesis more than another week of dashboard staring. Indispensable when traffic is too low to A/B test.

Repeated clicks on the same spot in a short window — a sign of broken expectations.

Surfaces in tools like Hotjar or FullStory. Useful as a leading indicator: rage clicks on a non-clickable element usually mean the design implies clickability the page doesn't deliver. Worth investigating every time.

Give something useful first, and the user feels they owe a small response.

A free audit, a free checklist, a free Loom — none of these are charity. They trigger the social instinct to reciprocate. The trick is the gift has to be *useful*, not a bait-and-switch. A real 6-page audit doc lands more discovery calls than a 30-page "ebook" that's actually a sales pitch.

How many users a test needs per variant before you can read it.

Determined by your baseline conversion rate, your desired MDE, and your significance threshold. There are calculators for this. Pre-commit to a number before launching — otherwise you'll peek, find what you want to see, and call it done.

A heat map variant showing what percentage of visitors reach each point on a page.

Scroll maps answer one question: how far are visitors actually reading? Each horizontal band of the page is shaded by the percentage of sessions that reached it — typically warm colours at the top fading to cold as you scroll down. The most useful finding is often not where people stop, but the gap between where the team assumed people would stop and where they actually do. A proof section that 80% of visitors never see is not a proof section — it is decoration. Before rewriting content that sits below the fold, always check whether the real fix is moving it up.

A video-like reconstruction of an individual user's session.

A close cousin of the heatmap. Replays are gold for diagnosing weird funnel drop-offs. The trap is watching 50 of them and concluding the world from anecdotes; pair every replay session with quantitative data.

A measure of how unlikely your result is to be random chance.

A 95% significance threshold means: if there were no real difference between variants, you'd only see this result 5% of the time by chance. A test that's "significant" is one that has cleared this bar — usually p < 0.05 in frequentist terms.

Things that feel limited feel more valuable.

Real scarcity (3 cohort seats, 1 client at a time, waitlist closing Friday) converts. Fake scarcity (rotating "12 people viewing this page right now") corrodes trust the moment it's spotted — and most buyers spot it. Treat scarcity as a signal you can only spend if it's actually true.

We use other people's behaviour as a shortcut to decide what to do.

The strongest CRO lever after the headline itself. Specific outperforms generic — "847 founders use this" beats "trusted by founders." Named users outperform anonymous counts. And testimonials work when they match the visitor's exact objection, not the brand's preferred narrative.

The prioritised list of hypotheses a CRO team is working through.

Every idea that surfaces — from data, from colleagues, from customers, from the team's own observation — goes on the list. The backlog does two jobs: it's a resource to pull from when deciding what to work on next, and it's a professional buffer against ad-hoc requests. When someone walks in with a competitor's website and says 'why aren't we doing this?' the answer is 'great idea — it's on the backlog.' The idea gets a fair process rather than either being dismissed or jumping the queue. A well-maintained backlog has every item written as a hypothesis (not just a vague idea), scored on ICE, and annotated with whatever supporting evidence exists.

Anything on the page that reduces the user's "why should I believe you" friction.

Testimonials, logos, awards, security badges, customer counts, case-study quotes, real names with real photos. The bar is "does this proof point at the specific claim in the hero." A logo wall with no story is decoration; a single customer quote that matches the headline is a trust signal.

The craft of writing the words inside an interface.

Related to microcopy, broader in scope. UX writing covers labels, error states, empty states, onboarding flows, system messages. CRO and UX writing overlap heavily — most "design fixes" in CRO are actually writing fixes.

A version of the page being tested against the control.

In an A/B test there are two: A (control) and B (variant). In an A/B/n test there are more — B, C, D, each tested against A. More variants need more traffic for the same MDE; resist the urge to test five things at once unless your numbers support it.

Direct feedback from real customers — what they said, not what you inferred from their behaviour.

The umbrella term for any research method that captures the customer's own words: post-purchase surveys, user interviews, on-site feedback widgets, live chat transcripts, review mining. The most useful VoC question in CRO is 'what almost stopped you?' asked immediately after conversion — it captures the real barriers real visitors encountered, rather than the imagined barriers the team debated. VoC is qualitative by nature: it explains the why behind the what that analytics reveals.

An unfinished task occupies the mind more than a finished one.

We remember and return to incomplete things. On product onboarding this is why "3 of 5 setup steps complete" pulls users back to finish — the unchecked boxes pester them. On content sites, this is why a "Part 1 of a 3-part series" article gets more return visits than a self-contained equivalent.