Conversion Research Sprint
Funnel analysis, session replays, heatmaps, on-site polls and user tests across your top revenue templates. Output is a ranked list of friction points with revenue attached.
Paris
Conversion Rate Optimisation in Paris, France
A test programme planned against a French trading year, where the calendar gives you a limited number of windows and no way to re-run one.
Delivered remotely for brands across Paris and France.
Grow · Paris
Why Paris brands come to us for this
- Test calendar planned around the fixed sale windows, so no experiment straddles a boundary and measures the promotion instead of the change
- Delivery step treated as the primary test surface: pickup-point placement, price framing against home delivery, and estimate wording
- Instalment payment messaging tested by position on product page, cart and checkout, judged on average order value rather than clicks
- French copy tested where it usually fails first, in filter labels, size guidance, delivery estimates and withdrawal wording
- Summer lull used for qualitative research and backlog building instead of underpowered tests nobody can call
Testing in France runs into a scheduling problem before it runs into a statistics problem. The promotional year concentrates into two nationally fixed sale windows, so behaviour during those weeks is not representative of the rest of the year and a test that straddles the boundary is measuring the calendar rather than the change. August distorts things again in the other direction. That means the useful testing weeks are finite and have to be allocated deliberately instead of filled with whatever hypothesis came up in the last meeting.
The friction points themselves are also specific. The delivery step is the single richest test surface on a French store, because pickup-point collection is a mainstream preference rather than a budget option, and where the relais choice sits relative to home delivery visibly changes both conversion and your cost per order. Instalment payment is the second: whether it appears on the product page, in the cart, or only at payment changes average order value in ways worth knowing rather than guessing. Neither of those tests exists on a US store, which is why importing a generic CRO backlog into a French store wastes half of it.
Then there is language, which is a conversion variable and not only a compliance one. Copy that reads as translated is audible to a French buyer in the first two lines, and the places it usually gives itself away are the least glamorous: filter labels, size guidance, delivery estimates, error messages and the withdrawal wording. Rewriting those is often a larger, cheaper win than anything on the homepage, and it is testable.
Local context
Use August for research and the shoulder weeks for tests
For a store doing between a few hundred thousand and a few million euros a year, statistical power is the binding constraint, and a French calendar makes it tighter. So we split the year by what each period is good for. The quiet summer weeks go to research rather than experiments: session replays, on-site polls in French, checkout drop-off analysis, reading support tickets and returns reasons, and building the backlog. The clean trading weeks either side of the sale windows carry the actual tests, one at a time where traffic is thin, with the sample size declared before the test starts and no test allowed to cross a soldes boundary. Inside the sale windows we stop testing and instead instrument heavily, because those weeks generate the highest-volume behavioural data you will get all year and it is worth capturing properly even though it cannot be generalised. And when a hypothesis is genuinely well-evidenced but the traffic will never prove it, we say so and ship it as a considered change rather than pretending a two-week underpowered test settled anything.
Scope
What Conversion Optimisation includes
The same standard of work we run for every client — applied to a Paris brand’s realities.
Full service detailPrioritised Test Roadmap
Every hypothesis scored on impact, confidence and effort, with the projected revenue and required sample size stated up front. You always know what runs next and why.
Experiment Build & QA
Tests built and QA'd across devices, browsers and your app stack, with flicker-free rendering and no measurable hit to Core Web Vitals.
Statistical Analysis
Pre-declared sample sizes, sequential-testing guardrails and segment-level readouts by device, traffic source and new versus returning. No peeking, no calling a test at 80%.
Checkout & Cart Optimisation
Cart drawer, shipping thresholds, payment options, express checkout placement and post-purchase upsell, tested against AOV and revenue per session rather than clicks.
Monthly Programme Report
Tests run, results in full, cumulative revenue impact and what the results taught us about your customers. Losers are reported in the same detail as winners.
Scoped and quoted for your Paris store
We do not work off a rate card. Every Paris engagement starts with a fixed statement of work — named deliverables, named dates, one number — written after we have looked at your store, not before. If a smaller first step would serve you better, we will say so.
Get this scopedHow it runs
From kickoff to results
Measure
We instrument the funnel properly first. Most stores have broken or double-counted events, and you cannot optimise against numbers you cannot trust.
Research
Quant tells us where visitors leave. Qual tells us why. We combine analytics, replays and direct customer input before writing a single hypothesis.
Prioritise
Hypotheses are scored and sequenced so the highest-value, lowest-effort tests run first. The roadmap is shared and you can reorder it.
Test
Two to four concurrent experiments depending on traffic, each run to a pre-declared sample size. No test gets stopped early because it looks good on day three.
Implement & Compound
Winners get hard-coded into the theme, losers get documented, and every result feeds the next round of hypotheses. The programme gets smarter each month.
Proof
Conversion Optimisation results
Anonymised under NDA. Figures pulled from the client’s own analytics.
Premium Skincare (DTC)
8-figure DTC brand, ~140 SKUs, US + CA · Shopify Plus (migrated from Magento 2)
Magento 2 cost $9k a month in hosting, extensions and emergency dev before a single feature shipped. Mobile LCP sat at 5.4 seconds and mobile converted at less than half the desktop rate across 68% of sessions. The named constraint: 11,412 indexed URLs and a top-3 organic position on their hero category, so a migration that lost rankings would cost more than the platform ever saved.
+41%mobile conversion rate1.62% to 2.28% over the first 90 days post-launch
5.4s → 1.8smobile LCP75th-percentile CrUX field data, not lab
+27%organic sessionssix months post-migration vs. pre-migration baseline, branded queries excluded
$108k/yrplatform cost removedhosting, extension licences and the standing emergency dev retainer
Performance Apparel (DTC)
~$6M/yr DTC, 900+ SKUs across size and colour variants, US · Shopify Plus
Returns ran at 31% and refund cost consumed the entire paid media margin. One size chart image served 40 different fits, and 62% of add-to-carts started on a collection page that never showed variant availability. The named constraint: no new product photography budget, so every fix had to come out of the existing asset library and the review corpus.
+29%sitewide conversion rate1.71% to 2.21%, five-month average
+14%average order value$84 to $96 once the threshold bar and cross-sell shipped
31% → 22%return ratenine-point drop, roughly $310k/yr in recovered margin
1.9x → 2.4xblended MERwith paid spend held flat throughout
In their words
Clients on this work
CVR 1.9% → 2.7%
“We'd been running our own tests for a year and calling every 3% swing a win. The first thing they did was tell us our sample sizes were garbage, which was not a fun call to sit through. Six months later we're getting wins that actually hold up when you re-measure them. Sitewide CVR went from 1.9% to 2.7%.”
Founder & CEOApparel DTC brand, ~$12M/yr · Los Angeles, CA
Verified client, 2025FAQ
Conversion Optimisation in Paris — your questions
We say so, and change the method rather than dress up an underpowered test. Below the volume where significance is reachable in a sensible timeframe, the work becomes research-led: funnel instrumentation, session replay, on-site polling, returns-reason analysis and checkout drop-off, then considered changes shipped with before-and-after monitoring and an honest label on the certainty. That is a legitimate programme. What is not legitimate is running a two-week test on three hundred orders and announcing a winner.
Test, no. Measure, absolutely. Behaviour in those weeks is not representative of the other ten months, so a result from them tells you about a discount period rather than about your store. What we do instead is instrument the windows properly, capture the drop-off and delivery-choice data at the highest traffic volume of the year, and use it to build hypotheses for the clean weeks that follow.
In our experience it moves both, which is why the test is worth running properly rather than assuming. Presented as an equal option with its price difference visible, it converts buyers who will not commit to being home for a delivery, and it lowers your cost per order and your failed-delivery rate at the same time. Buried under home delivery it barely gets selected and you carry the integration cost for nothing. The variable being tested is placement and framing, not whether to offer it.
On Shopify Plus, yes, through checkout UI extensions and Functions, which is where the delivery-method and payment-presentation tests live. On standard Shopify the checkout is locked, so the programme concentrates on the cart, the delivery step as far as it is configurable, express payment placement, free-shipping thresholds in euros and post-purchase offers. That is where most of the recoverable revenue sits on a store this size in any case.
Around 40,000 sessions and 800 orders a month is where a testing programme becomes statistically viable. Below that, tests take months to reach significance and you are better served by research-led redesign work and analytics fixes. We will tell you honestly which bucket you are in.
The first test goes live in week three, after research and instrumentation. Meaningful cumulative impact typically shows around month four, once six to ten tests have run. CRO is a compounding programme, not a one-month fix, and anyone promising otherwise is selling best practices.
Some do, and that is normal. A loser is still information: it rules out a hypothesis and sharpens the next one. We report losses in the same detail as wins because a programme that only ever produces winners is one that is not being measured properly.
Next step
Conversion Optimisation for your Paris brand.
Thirty minutes with the strategist who would actually run your account. We screen-share your store, read your data live, and tell you the three highest-value things we can see from the outside.
Shopify or Shopify Plus stores doing $150k/mo or moreFounder, CEO or eCommerce lead on the callNo deck and no pitch — we open your store instead
Prefer to write it out? [email protected] gets a real reply the same business day, Mon-Fri, 9am-6pm MT.