Tags: experimentation ux map

CRO

Date: 2026-08-16

---

Conversion rate optimisation is a loop, not a list of tactics: find where people fail, form a hypothesis about why, test it honestly, and keep what survives. Every stage borrows from a different domain, which is why nobody owns the whole thing.

---

A cross-domain view. Nothing here is filed in /Maps/ — every note lives in its home domain, and this is the order the work actually happens in.

Draws on Analytics, User Experience, Experimentation, Statistics and Commerce & Growth.

The procedural counterpart, following the same order with the arithmetic worked: Guide - Ecommerce CRO.

1. Find where it’s failing

Quantitative first, to locate. Qualitative second, to explain. Doing it in the other order produces confident answers to the wrong question.

• Funnel Analysis — where people leave, and the step-definition choices that move the answer
• Segmentation (analysis) — the group the average was hiding
• Sessionisation — before believing any session-scoped rate
• Session Replay — watching the failure, once you know where to look
• Heatmaps — aggregate attention and interaction
• Form Analytics — field-level abandonment, the highest-yield diagnostic in checkout
• Voice of Customer Data — surveys, support tickets, and what people say is wrong
• Usability Testing — the method that explains why faster than any amount of data
• Triangulation — reconciling all of the above when they disagree
• Guide - Running a Conversion Audit — the full sweep, as a procedure

Before trusting any of it: the symptom list in Analytics MOC and Guide - Auditing a Tracking Plan. A large share of “conversion problems” are measurement problems, and finding that out after three months of testing is the classic waste.

2. Explain it

The diagnosis stage. A finding without a mechanism produces a test that can’t teach you anything.

• Cognitive Load — the budget every interface spends
• Choice Overload — more options, fewer decisions
• Trust Signals — what actually moves credibility
• Loss Aversion · Anchoring · Framing Effects — the three that explain most pricing and offer behaviour
• Error Prevention and Recovery — the form and checkout failures that look like disinterest
• Loading and Perceived Performance — slowness as an abandonment cause, not a technical metric
• Value Propositions — whether the page answers “why this, why you, why now”
• The Offer · Risk Reversal — when the objection is to what’s being sold rather than to the page selling it
• Awareness Stages — whether the page assumes more, or less, than the visitor already knows
• Deceptive Design — the line, and the UK regulatory position on crossing it

3. Prioritise

• Test Prioritisation — ICE, PIE, and their honest limitations
• Metric Sensitivity — whether you have the traffic to detect what you’re proposing
• Minimum Detectable Effect — deciding what’s worth detecting before designing the test
• Win Rate and Expected Value — the arithmetic that beats scoring frameworks: probability × impact × reach ÷ cost
• Performance and Conversion — often the highest-value item on the list, and rarely on it

4. Form the hypothesis

A hypothesis needs a mechanism — the reason the change should work. That reason usually comes from behavioural principles, which is the half of CRO this map would otherwise skip.

• Cognitive Load · Choice Overload · Hick’s Law — friction from too much to process
• Social Proof · Authority · Trust Signals — evidence and credibility
• Scarcity and Urgency · Loss Aversion · Framing Effects — pressure and presentation, and where each stops being honest
• Commitment and Consistency · The Default Effect — sequencing and what’s preselected
• Deceptive Design — the line, and where UK law now sits

Then the hypothesis itself:

• Hypothesis Design — a mechanism and a prediction, not a preference with a metric attached
• Overall Evaluation Criterion — the single primary metric, decided now
• Guardrail Metrics — what must not get worse
• Pre-Registration — writing the analysis plan before the data exists

5. Test it

• Guide - Running an Experiment — the full procedure, and Guide - Statistics for CRO for the arithmetic
• Randomisation Unit — user, not session. Almost always
• Sample Size Calculation — baseline, MDE, power, significance
• Test Duration — full business cycles, not just sample size
• Client-Side vs Server-Side Testing — and the flicker tax if client-side
• A-A Tests — proving the machinery before trusting it
• Sample Ratio Mismatch — the first thing to check, every time
• Experiment QA — before traffic, not after

6. Read the result

The stage where most value is destroyed, by people acting in good faith.

• Guide - Statistics for CRO — the ordered checks, and reading the interval rather than the p-value
• P-Values — what the number is, and the four things it isn’t
• Confidence Intervals — the range, which is the honest output
• Statistical Power — because an underpowered test that “lost” told you nothing
• Peeking — why the result you looked at on day three isn’t the result
• Segmentation (test results) — post-hoc slicing, and why nearly every finding from it is noise
• Winner’s Curse — why shipped winners underdeliver
• Inconclusive Results — the most common outcome, and not a failure

7. Ship and hold

• Feature Flags — shipping the winner without a release
• Post-Test Validation — checking the change delivered what the test promised
• Holdout Groups — measuring cumulative programme effect over a year
• Guide - Running an Experiment — its closing section covers the archive, so the same idea isn’t retested in eighteen months

8. Make the programme compound

Individual tests barely matter. The system that produces them does.

• Experimentation Velocity — tests per period as the real driver
• Win Rate and Expected Value — most tests fail; the arithmetic across all of them is the return
• Institutional Learning — turning results into beliefs, and revising beliefs when results contradict them
• Communicating Uncertainty — presenting ranges to people who want a number
• Experimentation Maturity — the bottleneck at each stage

Where CRO stops

Worth stating, because the discipline gets blamed for these.

• Traffic quality is acquisition’s problem — a page can’t convert badly-matched intent. See Channel Mix and Landing Page Strategy
• Margin is pricing’s problem — conversion rate can rise while contribution margin falls. See Contribution Margin and Discount Impact on Margin
• Retention is lifecycle’s problem — optimising first purchase says nothing about the second. See Repeat Purchase Rate
• Genuine incrementality at channel level needs Incrementality Testing, not on-site testing