What Snapshot Picked
// gate one, continued
Snapshot CRO scores every page in that table and recommends what to audit next. Here is its answer, and here is exactly how it got there.
| Page | Sessions | Weakness | Opportunity | Priority | Confidence |
|---|---|---|---|---|---|
| Alpine Down Parka | 42 | 92 | 74 | 83 | 100% |
| Trail Runner 2 | 1,204 | 44 | 100 | 72 | 100% |
| Base Layer Crew | 318 | 15 | 97 | 56 | 100% |
| How to Fit Trail Shoes | 274 | 51 | 60 | 56 | 100% |
| Homepage | 1,876 | 4 | 100 | 52 | 100% |
| Summit Jacket | 742 | 0 | 100 | 50 | 100% |
| Shipping & Returns | 461 | 34 | 60 | 47 | 100% |
| New Arrivals | 986 | 12 | 75 | 44 | 100% |
the model, in full
No black box. Weakness adds up how far each metric falls short of a target, weighted by how much that metric matters:
| scroll < 60% | up to 24 points |
| click-through < 25% | up to 24 points |
| exit > 35% | up to 16 points |
| zombie > 20% | up to 16 points |
| bot > 10% | up to 12 points |
| product pages add | add-to-cart under 10% (18 points) and conversion under 2.5% (14 points) |
Opportunity is traffic plus money plus intent:
| traffic | min(60, sessions × 3) |
| revenue | min(25, revenue / 40) |
| intent | min(15, searches × 2 + orders × 5 + 5 if any add-to-cart) |
And then priority = (opportunity + weakness) / 2, top eight wins.
now look again
Three things in that table should bother you.
The top recommendation has 42 sessions. Alpine Down Parka really is the most broken page in the store — 29% scroll, 41% zombies, no conversions. The weakness score is honest. But min(60, sessions × 3) maxes out at twenty sessions, so a page with 42 visitors scores full marks on traffic — the same as the homepage with 1,876. Every page in this table maxes it. Traffic stops being a signal the moment you have twenty visitors.
Revenue saturates too, at $1,000. Summit Jacket's $5,120 and Base Layer Crew's $890 score within three points of each other. Once a page clears twenty sessions and a thousand dollars, opportunity is effectively a constant — and priority quietly collapses into "weakness, halved". The model is ranking friction and calling it priority.
Every page reports 100% confidence. That number is sessions / 30, capped at 100. Thirty human sessions buys you a hundred percent. It is a progress bar for data collection wearing the word "confidence", and it is why gate three exists later today.
Two more, once you're looking: a blog post about shoe fit ties a product page making $890 a month. And Shipping & Returns gets marked down for a 61% exit rate — on a page where reading the return policy and leaving is the page doing its job. The model has no idea what any page is for.
so what is it good for
Shortlisting. It read eight pages of behavioural data in a second and put two pages worth looking at in front of you, which beats scrolling a spreadsheet and guessing. Deciding which of those two deserves a month is your job, and it isn't close: 1,204 sessions against 42.
That is the whole relationship. The tool narrows, you decide. Shipping the top row without reading it means letting an arithmetic mean run your quarter.
// your read on the model
How much would you trust this ranking to pick next month's work?
1 = I'd ignore it10 = I'd ship the top row