The Win Board
Measured, not claimed.
UPDATED 2026-08-17
Measured with a script, not claimed. Every number on this page was taken by a script and can be re-checked by anybody, including you, and where I have no data the row says so instead of guessing. The clearest example is my own site: Search Console had flagged 299 pages as discovered but not crawled, and this shows exactly what changed.
This site, rebuilt in the open
The most honest thing I can show you is what I did to my own site, because you can verify every line of it yourself. Google Search Console had flagged 299 pages as discovered but not crawled. Here is what changed and what it moved.
| What was wrong | Before | After |
|---|---|---|
| URLs Google knew about, for a 514-page site | 872 | Consolidated to one canonical each |
| Pages reachable at two live URLs | Every page | 532 forced 301s |
| Analytics on the site | None, on any page | All pages, with conversion events |
| FAQ schema for questions not on the page | 336 of 402 | 0 |
| Titles truncating in search results | 171 | 0 |
| Meta descriptions truncating | 398 | 0 |
| Pages with a keyboard skip link | 0 | All |
| Heading-level skips (accessibility) | 209 | 0 |
| Pages with 2 or fewer inbound links | 185 | 17 |
| Render-blocking third-party font requests | Every page | 0 (self-hosted, 84 KB) |
| Grammar defects in generated copy | 42 | 0 |
| Broken outbound links | 1 | 0 |
Method: every figure came from a script run across all pages, before and after. The scripts are the same ones that now block a release if any of it regresses.
Content depth, measured the hard way
Programmatic pages are easy to make and easy to make badly. The trap is that adding length makes pages more alike, not less. I measured that happening to my own pages and then fixed it properly.
| Page family | Words before → after | Similarity to siblings |
|---|---|---|
| Cost guides by industry | 242 → 654 | 0.47 → 0.45 |
| AI playbooks by industry | 244 → 410 | 0.46 → 0.37 |
| Tool matchups | 212 → 340 | 0.55 → 0.39 |
| Service pages by suburb | 286 → 369 | 0.62 → 0.43 |
Lower similarity is better: it means the pages are genuinely different rather than the same page with the town swapped. A release now fails automatically if any family drifts back above 0.50.
Client work
| Client | What shipped | Result |
|---|---|---|
| North Peak Financial Partners | Full site, accounting & CFO advisory, AccountingService schema | Live. Two 5-star reviews. |
| Perfect Place Remodeling | Full site, quote flow, project galleries | Live. 5-star review. |
| GrandmasterGains | Chess training platform | Live |
| Bast: Nine Lives | Browser game, 147 levels | Live, public beta |
What I don't have yet, and won't invent
No client traffic numbers. No lead counts. No revenue figures. These sites are new, analytics was only installed recently, and a number from three weeks of data is noise dressed as a trend.
When there is a year of real data, it goes in the table above with the same specificity as everything else, including the results that are less flattering. A board that only ever goes up isn't a record, it's an advertisement.
Read the North Peak case study → · See the work · Start yours
Your move