Proof
Case studies that read like investigations.
4 investigations
How we actually work through a problem
- EcommerceWorked example12 June 2026 · 6 min
Twelve thousand product pages that could not rank for their own model names
A canonical rule added during a duplicate-content cleanup pointed every product URL at its parent category. The catalogue was invisible for its highest-intent queries and nobody had connected the two facts.
Read the investigation
- Product URLs canonicalised away
- 11,902
- Indexed product URLs at baseline
- 0
- Days from release to detection
- 94
- Effort to fix the rule
- 1 sprint
- EcommerceWorked example3 July 2026 · 5 min
A migration redirect map that covered 96% of URLs and 41% of the revenue
The redirect map was built from a CMS export, so it covered almost every URL by count. Scored by traffic, conversions and backlinks, it missed most of what the old site actually earned.
Read the investigation
- URL coverage by count
- 96%
- Coverage by scored value
- 41%
- High-value URLs with no destination
- 2,140
- Weeks before launch the review starts
- 6
- B2B servicesWorked example24 July 2026 · 5 min
GA4 and the CRM disagreed by 38%, and the marketing budget was moved on it
Four separate causes, each measurable. The team had been arguing about which number was right instead of decomposing the gap into its parts.
Read the investigation
- Reported gap, GA4 vs CRM
- 38%
- Distinct causes found
- 4
- Explained after decomposition
- 34 of 38 pts
- Days to first significant finding
- 2
- SaaSWorked example14 August 2026 · 6 min
AI assistants were quoting two-year-old pricing and citing a directory
A fixed prompt set, run across four surfaces, found the company mentioned in a third of answers, recommended in almost none, and described with pricing it had abandoned two years earlier.
Read the investigation
- Prompts in the frozen set
- 40
- Mention rate at baseline
- 33%
- Recommendation rate at baseline
- 8%
- Factual errors found
- 6
How we handle proof
We publish results only when they are real, measured, and the client has agreed to them being published. Until then we would rather show you how we think than show you numbers we invented.
What you can check today: the method, the evidence standard, the priority system, and the fact that every finding on this site is written in the same Problem → Evidence → Impact → Root cause → Fix → Measurement format we use in delivery.