Methodology
How we run CRO — in more detail than most agencies put on a website.
The three-layer measurement framework, the change log, and why Class B uses heuristic CRO while only Class A runs A/B tests.
Three-layer measurement
No single source of truth. Three, triangulated.
Shopify, GA4, and Clarity each answer a different question. Together they tell you what happened, where, and why. Missing one of the three means guessing at the answer to something the other two can't see.
Revenue truth
Shopify Analytics
Orders, AOV, conversion rate at the storefront. The authoritative record of what actually paid.
Traffic + funnel
GA4
Source, medium, landing, exit. Where sessions come from and where in the funnel they die.
Behaviour
Microsoft Clarity
Session recordings, heatmaps, rage-clicks, dead-clicks. Why the behaviour is what it is, not just that it is.
Heuristic CRO vs. A/B testing
Traffic decides which tool applies.
Running A/B tests under 25,000 monthly sessions means waiting months for any one test to reach significance — by which point the test is measuring a different market than the one it started in. So we don't. Instead, we run heuristic CRO until the traffic supports the statistical tool.
10,000+ monthly sessions
Heuristic CRO (Class B)
We observe real user behaviour in Clarity, cross-reference funnel drop-offs in GA4, and ship targeted fixes against the highest-leverage leaks first. No split traffic — every fix ships to 100% of visitors. Measurement is before/after against the 60-day baseline, with a 90-day post-deployment window.
- →No statistical-significance claims — we don't split traffic at this tier
- →Fixes ship weekly, measured monthly against baseline
- →Reverted if the next month's data shows a regression
25,000+ monthly sessions
Structured A/B testing (Class A)
At 25K+ monthly sessions, structured A/B testing starts producing statistically meaningful results within a reasonable time window. We run tests via GrowthBook with pre-registered sample sizes, significance thresholds agreed before the test starts, and outcome documentation regardless of win or loss.
- →Sample size + significance threshold set before any test runs
- →Primary metric declared before the test — no post-hoc reframing
- →Losers and inconclusives documented as carefully as winners
Measurement windows
60 days of baseline. 90 days of measurement. Everything anchored.
60-day baseline
The window before any fix is deployed — the thing every subsequent measurement is compared against.
Fix deployed
Change goes live with rationale, before/after screenshots, and the primary metric it's targeting logged in the dashboard.
90-day post-window
The window we measure the fix across. Shorter windows can be noisy on small stores — we default to 90 days for stability.
If baseline shifts ±25%
Traffic mix changes break the apples-to-apples comparison. We pause the comparison, re-baseline, and flag it in the change log.
The change log
Every change, every rationale, in one place.
The change log is the institutional memory of your store. Every row is a fix we shipped, a test we ran, or a decision we made not to ship. It lives in the Client Dashboard forever.
Example entry
Date2026-03-28
Page/products/[handle]
ChangeAdded trust strip (returns, shipping, support) above add-to-cart.
RationaleClarity recordings showed 34% of visitors scrolling past ATC without clicking. GA4 cross-referenced with a 22% PDP→cart drop. Trust strip addresses the dominant hesitation pattern observed in recordings (return policy ambiguity).
Result+0.34 CVR (90-day window, baseline 2.76%)
Documentation discipline
The rules we hold ourselves to.
Every decision logged.
Even decisions not to ship. The archive is the institutional memory of your store.
Every result measured.
Win, loss, inconclusive — all three are documented with the same rigour. Nothing gets to hide.
Every page-scope isolated.
If you make UX changes to pages we're working on, we pause the measurement for those pages and re-baseline. Attribution stays clean.
Ready to grow?
Want to see this methodology applied to your store?
We'll look at your site, find where it's losing revenue, and tell you exactly what needs to change — before any engagement starts.
No pitch. No pricing. Just the finding.