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.
A structured experimentation programme for Amsterdam stores where the Dutch numbers and the cross-border numbers are two different businesses.
Delivered remotely for brands across Amsterdam and Netherlands.
The first thing we do on a Dutch account is stop looking at the blended conversion rate. A store selling into the Netherlands, Belgium, Germany and France has four conversion rates, four average order values and four returns rates, and averaging them produces a number that describes none of them. The domestic rate is usually strong and driven by familiarity, iDEAL and a delivery promise the buyer trusts. The German rate is often half of it, and the reason is almost never the product. It is a missing payment method, a delivery estimate the buyer does not believe, or a returns path that reads as expensive.
The second thing is treating returns as part of the conversion equation instead of an operations problem downstream. The EU withdrawal right gives every consumer fourteen days to change their mind, and in denim and apparel that makes the returns rate the largest single drain on contribution margin in this market. A test that lifts add-to-cart by six percent and lifts returns by nine is a loss, and you will not see it in a standard CRO report. So our primary metric on apparel accounts is revenue per session net of returns, read by market and by size, which changes which tests are worth running.
The third is the checkout experience itself, where the Dutch specifics are unusually concrete. Payment order, because iDEAL presented below card fields costs domestic conversion measurably. Delivery choice, because a named day, an evening slot and a pickup point are what buyers expect to see rather than a single generic shipping line. And the point at which shipping cost appears, which in a market with high free-shipping expectations is a moment of truth that belongs earlier on the page than most stores put it.
Amsterdam brands run into a specific statistical trap. Domestic traffic is the biggest and the best-converting segment, so any test read on the blended number gets decided by Dutch buyers, and the German or French experience never gets fixed because it never gets seen. We set market as a first-class segment before the first experiment goes live, and where a secondary market lacks the volume to reach significance on its own, we say so and treat it with research rather than pretending a test read it. The other local input is seasonality with an early peak. Sinterklaas on 5 December compresses the Dutch gifting window and pulls it forward of the rest of Europe, which means a test running through late November is measuring a different shopper on either side of the date. We freeze the roadmap through the peak, use it to gather research rather than to run experiments, and resume in January with a better set of hypotheses than we went in with.
The same standard of work we run for every client — applied to a Amsterdam brand’s realities.
Full service detailFunnel 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.
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.
Tests built and QA'd across devices, browsers and your app stack, with flicker-free rendering and no measurable hit to Core Web Vitals.
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%.
Cart drawer, shipping thresholds, payment options, express checkout placement and post-purchase upsell, tested against AOV and revenue per session rather than clicks.
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.
We do not work off a rate card. Every Amsterdam 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 scopedWe instrument the funnel properly first. Most stores have broken or double-counted events, and you cannot optimise against numbers you cannot trust.
Quant tells us where visitors leave. Qual tells us why. We combine analytics, replays and direct customer input before writing a single hypothesis.
Hypotheses are scored and sequenced so the highest-value, lowest-effort tests run first. The roadmap is shared and you can reorder it.
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.
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.
Anonymised under NDA. Figures pulled from the client’s own analytics.
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.
~$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.
“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%.”
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.
Prefer to write it out? [email protected] gets a real reply the same business day, Mon-Fri, 9am-6pm MT.