Work

The findings clients did not want to hear

Every engagement below started with an audit of what the numbers were really saying. Sectors and figures are real; client names are withheld.

Selected work

Four engagements, start to finish

Health & wellness ecommerce · Diagnostic, then advisory

Meta said scale. The evidence said stop.

A collapsing ad account diagnosed through measurement, budget moved to the channel actually converting, and an email audience cut to the people who could legally be emailed.

~10xCPC inflation found on Meta (11p to £1+)
2.55Blended ROAS on the channel quietly carrying the account
47%Of the claimed email audience found unsendable
Home & garden ecommerce · Audit, rebuild, advisory

When platform ROAS lies, run the business on profit

Retailers undercutting RRP made revenue-based ROAS actively misleading. The measurement was rebuilt and decisions moved to profit on ad spend.

0.27ROAS on a broad match campaign found and stopped
387Collections audited and remapped for search visibility
1,400+Contacts protected from an unreviewed automated send
Independent music retail · Forensic measurement rebuild

Attribution rebuilt after a headless migration broke it

A cross-domain storefront broke attribution end to end. One deterministic join key stitched it back together.

3 → 1Competing identity systems reduced to one join key
60 vs 22Inflated checkout starts against actual purchases, before dedupe
~6 hrsOf forensic debugging to isolate a rogue pixel
B2B data & AI consultancy · 6 months

The MQL target hit for the first time in over a year

Marketing and sales reconnected through one rebuilt measurement layer. The MQL target fell inside six months.

~100%Tracking accuracy on the rebuilt setup
12+ monthsOf missed MQL targets, ended
3Enterprise meetings in week one of launch
One to two weeks, fixed fee

Find out which of your numbers you can actually trust