A documented backlog
Every hypothesis, its evidence and its result — including the losers, which stop the same idea being re-proposed every year.
CRO, SPEED, UX AUDITS AND LANDING PAGES
Fix what the traffic is already costing you
Research, hypothesis, test, ship. A continuous programme against the numbers you agreed at the start, not a one-off redesign.
IN SHORT
- Optimization is a continuous programme, not a redesign — we agree the metric before the work starts.
- Winning tests ship permanently in theme code rather than living inside a third-party testing script.
- Stores need roughly 10,000 sessions and a few hundred conversions per variant per month for split testing to reach significance in a sensible window.
- Below that threshold we optimise from research and qualitative evidence instead of pretending a test was conclusive.
- Speed work usually produces the fastest wins, because most Shopify stores carry more third-party JavaScript than first-party.
Where we look first
Four areas, each with its own page.
Conversion Rate Optimization
Research, hypothesis, test, ship — a continuous programme rather than a one-off redesign.
Speed & Core Web Vitals
App bloat stripped, scripts deferred, LCP and CLS measured on real devices rather than a lab score.
UX & Checkout Audit
A senior engineer walks your storefront, theme and checkout and sends back a prioritised list of fixes.
Landing Page Production
Campaign and launch pages built to a section library, so marketing ships without a dev queue.
Why most CRO programmes fail
Not because the ideas are bad. Because nobody agreed what was being measured, so every result becomes an argument. A test lifts add-to-cart and drops revenue per session — was that a win? If you did not decide beforehand, you will decide afterwards, and you will decide in favour of whoever proposed it.
The second failure is stopping early. A variant goes 20% up on day three, someone calls it, and the store ships a change that was noise. Two full business cycles is the minimum for a reason.
The third is testing trivia. Button colours and headline tweaks on a store doing modest traffic will never reach significance, and the team spends a year learning nothing. At that traffic level, research and qualitative evidence beat split testing — and we will say so rather than sell you a testing retainer.
How the programme runs
A four-stage cycle, repeating monthly.
- 01
Research
Analytics, session recordings, support tickets, on-site search and checkout drop-off. Evidence before opinion — most stores already contain the answer to what is wrong, unread.
- 02
Prioritise
Every hypothesis is scored on expected impact against build cost, so the running order is arithmetic rather than whoever argued hardest in the meeting.
- 03
Ship
Changes built in theme code. Where traffic supports a split test we run one; where it does not, we ship the better-evidenced version and measure the before and after honestly.
- 04
Review
Monthly: what shipped, what moved, what was a dead end, what is next. Losers are documented rather than quietly forgotten, because they are the cheapest knowledge you own.
Do you have enough traffic to test?
The honest threshold
Roughly 10,000 sessions and a few hundred conversions per variant per month. That gets a test to significance in two to four weeks, which is a cadence you can actually run a programme on.
Below it, you have two options. Run tests anyway and wait months per result — viable if you only have a handful of questions. Or optimise from research: session recordings, support tickets, on-site search and checkout analytics tell you a great deal, and shipping the better-evidenced version is usually worth more than waiting a quarter to be 95% sure about one button.
What we actually look at
- Checkout and cart drop-off, step by step
- Product page: variant selection, stock messaging, delivery expectations, returns clarity
- Collection pages: filtering, sorting and how people navigate a large catalogue
- On-site search — the most under-read source of customer intent on most stores
- Core Web Vitals in field data, not just a lab score
- Support tickets, which tell you what the store failed to explain
- Mobile behaviour specifically, where most traffic and most friction lives
What you get out of it
Changes in your theme
Winners live in theme code, not in a third-party script that slows the site and expires with the subscription.
Numbers agreed up front
The metric is named before work starts, so there is no argument about what counted afterwards.
Honest reporting
If a quarter produced nothing, the review says so. Optimization programmes that never report a failure are not reporting.
Brands we do this for
Wonderskin
Mighty Jaxx
Wonderskin runs continuous landing page production with us alongside maintenance. Mighty Jaxx we manage end to end. See the full client list.
Optimization questions
How much traffic do we need for A/B testing?
Roughly 10,000 sessions and a few hundred conversions per variant per month to reach statistical significance within two to four weeks. Below that, split testing produces confident-looking numbers that are mostly noise, so we optimise from research and qualitative evidence instead.
How long should a Shopify A/B test run?
At least two full business cycles, usually two to four weeks. Stopping early because a result looks good is the most common way teams convince themselves of something that is not true.
Do you use a testing app?
Only where it earns its performance cost. Testing scripts run before your page renders, so a permanent testing tool on every page is a real speed tax. We prefer tests built in theme code, with the tool used deliberately rather than by default.
What is the difference between CRO and speed work?
CRO changes what people see and decide; speed work changes whether they wait around to see it. Speed is usually the better first move because it lifts every page at once and the work is finite, where CRO is an ongoing programme.
How quickly will we see results?
Speed improvements are measurable within days of shipping. CRO results depend on traffic — with enough of it, the first tests conclude within a month. Anyone promising a specific percentage before looking at your store is guessing.
Related
Start with a free UX and checkout audit if you are not sure where the problem is. If the store is slow, speed optimization usually pays back first. Ongoing work runs as a development retainer.
NEXT STEP
Free store audit
Storefront, theme performance and checkout reviewed by a senior engineer. No pitch deck, no obligation.