Work

Three projects, with the numbers attached.

Client names are withheld where the contract asks us to. Every figure below is measured against the same period a year earlier, not against the worst week we could find.

+31%

Checkout completion, home goods store

Rankings lost across a 4,000-product migration

9 hrs

Weekly admin removed by one supplier feed

Case 01 Home goods Shopify 12 weeks

Two thirds of their checkout traffic never finished

The situation: A home goods store doing about €1.8m a year had spent two quarters increasing ad spend with almost no revenue movement. They assumed the problem was traffic quality. The funnel said otherwise: people were reaching checkout in normal numbers and abandoning at the payment step.
What we found: Card only, no local wallets, in three markets where most people pay by bank transfer or Klarna. Shipping cost appeared for the first time on the final step. The address form asked for eleven fields including a company VAT number that only 4% of buyers had.
What we changed: Added the two local payment methods and one wallet, moved the delivery promise and cost above the cart, cut the form to seven fields and made the VAT field optional behind a toggle. Four releases over six weeks, each measured on its own before the next went out.
The result: Checkout completion moved from 38% to 50% — a 31% lift — with no change in ad spend. The payment methods alone accounted for roughly half of it.

Checkout funnelBefore → after

Reached checkout4.2% → 4.3%

Entered payment3.1% → 3.6%

Completed order1.6% → 2.1%

Form fields11 → 7

Traffic was flat across the period. The lift came entirely from people who were already at the payment step and leaving.

Case 02 Speciality retail Woo → Shopify 9 weeks

4,000 products moved without losing a ranking

The situation: A speciality retailer on an ageing WooCommerce build wanted out. Roughly 60% of their revenue arrived through organic search, so the board's only real question was whether the move would cost them their positions.
What we found: Eleven years of URL history: three previous URL structures, redirect chains four hops deep, and about 900 live product URLs that no longer appeared anywhere in the site's navigation but still pulled traffic.
What we changed: Built the URL map from a live crawl plus twelve months of search console data rather than from the product export. Flattened every redirect chain to a single hop. Rebuilt the theme on the new platform, tested the map against staging, and launched on a Sunday night with us watching logs for 48 hours.
The result: No measurable ranking loss on tracked terms after eight weeks. Organic sessions dipped 6% in week one and recovered by week three. Page load on product templates came down by just over a second as a side effect.

MigrationStore B

Products moved4,120

Redirects mapped6,884

Redirect chains left0

Tracked terms lost0

Trading hours missed0

The 900 orphaned URLs would have 404'd on a product-export migration. They were worth about 8% of organic revenue.

Case 03 Trade supplier BigCommerce 4 weeks

The morning CSV that ate nine hours a week

The situation: A trade supplier with 2,600 SKUs across four wholesalers. Every morning someone downloaded four stock files, merged them by hand and uploaded the result. Overselling was routine, and the fix was always a phone call.
What we found: The spreadsheet step was not the real problem — the mismatched supplier codes underneath it were. Two suppliers used their own SKUs, one changed its column order without warning, and nobody was alerted when a file simply failed to arrive.
What we changed: Built a mapping table between supplier codes and the store catalogue, then a scheduled job that pulls all four feeds, validates the shape of each file before it touches stock, and writes levels straight to the store. Failures raise an alert in the team's chat within the hour.
The result: Nine hours a week returned to the team, stock accurate to within the hour instead of the day, and oversell incidents down from several a week to two in the following quarter.

OperationsStore C

Suppliers connected4

SKUs synced2,600

Stock refreshDaily → hourly

Admin hours / week9 → 0

Oversells / quarter~40 → 2

Validation before write is the part that matters. A malformed feed now fails loudly instead of quietly zeroing stock.

The fix list was the useful part. Half of it we did ourselves in a fortnight, and we'd have never known those items existed.
Operations director · home goods retailer

How we report

What we count, and what we don't

Baselines

Same period, previous year

Ecommerce is seasonal. Comparing a November to the October before it makes any project look like a triumph, so we don't do it.

Attribution

One change at a time

Where traffic supports a test, we test. Where it doesn't, we sequence releases so each change has a clean window either side of it.

Honesty

Failed changes stay in the report

Roughly a third of what we ship does nothing measurable. It stays in the monthly readout, because knowing what doesn't work has a value too.

Access

You own the data

Everything runs in your analytics and your accounts. When an engagement ends, nothing has to be handed back — it was always yours.