Services / Rebuild
On a live store, the research is the deliverable. The changes are the consequence.
A store that already sells has something a new one does not: evidence. The work is therefore not design but diagnosis, and the most valuable thing we produce is usually the ranked list of what is actually costing money, not the redesign that follows it.
The canon
The principles this rests on, named.
Each with who established it, so the claim can be checked rather than taken on our word.
Test the big things. Small tests need traffic almost nobody has.
Statistical power; see Georgi Georgiev, Statistical Methods in Online A/B Testing
Detecting a 2% relative lift needs sample sizes far beyond a typical Indian D2C store. Which means most 'A/B testing' being sold is measuring noise.
Never stop a test because it reached significance.
The peeking problem; Johari et al., sequential testing
Repeatedly checking an unadjusted test inflates the false positive rate dramatically. Fixed horizon set in advance, or a properly sequential design. There is no third option.
Watch what people do before asking them what they think.
Jakob Nielsen, usability engineering; and the general unreliability of stated preference
Session recordings and funnel data say where. Interviews say why. Reversing that order produces expensive redesigns of things that were never broken.
Fitts's law: time to a target is a function of distance and size.
Paul Fitts (1954)
Concrete and measurable. It is why the primary action's position and size on a mobile product page is an arithmetic question, not a taste one.
Hick's law: choice time rises with the number of options.
Hick (1952), Hyman (1953)
The argument for cutting a nineteen-item navigation, and against adding a fourth upsell to the cart drawer.
Speed is a conversion variable with a measurable coefficient.
Core Web Vitals field data; consistent across published retail studies
On a live store it is often the largest single available gain, and unlike most changes it needs no design approval and cannot be argued about.
The funnel names the constraint. Fix that, not the thing you find ugly.
Theory of constraints, applied to conversion
A 40% improvement to a step that only 3% of sessions reach is worth almost nothing, and it is the most common way a CRO budget is wasted.
Method
What actually happens, in order.
Check that the existing analytics are telling the truth first. A surprising share of audits begin by discovering the data was wrong.
Where sessions are lost, by device and by traffic source. Ranked by absolute lost revenue, not by percentage drop.
Recordings and form analytics at the specific step that costs most. Not a general browse.
Post-purchase and exit questions, plus a handful of real conversations. Applied only to the leaks already identified quantitatively.
Estimated impact against effort, with the reasoning shown, so the client can disagree with the ranking rather than just receive it.
Where traffic supports a test, test properly with a fixed horizon. Where it does not, say so, ship on judgement, and measure the before and after without pretending it was an experiment.
What goes wrong
The failure modes nobody advertises.
The single most common dishonesty in CRO. A test with a fifth of the sample it needs will still produce a number, and that number is noise.
Checking daily and stopping on the first significant reading. It inflates false positives to a level where the whole programme is decorative.
A sticky add-to-cart bar installed on a store whose actual leak is a shipping cost revealed at checkout.
The homepage gets rebuilt because it is the most visible page, while the checkout that loses 60% of carts is untouched.
Audited on a desktop on office broadband. The actual customer is on a mid-range Android on a congested network.
What we refuse
A studio is defined as much by this list.
- We do not report a test result without its sample size and power.
- We do not stop a test early because it is winning.
- We do not recommend a change we cannot connect to something in the data.
- We do not install dark patterns, fake scarcity, or pre-ticked add-ons.
- We do not claim credit for a lift we cannot separate from seasonality.
Measured, not felt
How anyone can tell whether it worked.
Revenue per session, which moves for the right reasons where conversion rate alone can mislead.
Step-by-step funnel completion, by device.
Core Web Vitals field data, not lab scores.
For every shipped change: the before, the after, and an honest statement of what else changed at the same time.