Conclusion
// the backlog
Your backlog, in the order you'd work it.
// nothing here yet. write your three hypotheses and score them first.
take it with you
This list only lives in this browser. Copy it into your own Google Sheet (or download it) to keep it as the start of your program.
four sessions, one sentence each
- one — customers arrive with a question, and the page either answers it or doesn't
- two — the data tells you where they stopped, never why
- three — PACE is how you find the why, one stage at a time
- four — a program is a queue, a model, and a loop that keeps turning
what to do on monday
- run a store snapshot on your own store and read gate one honestly — including which pages the model over-ranks
- pick the page with traffic and friction, not the one with the worst score
- run the sample-size numbers before you plan a single test. Most of your pages will route to "ship it", and that's the finding
- write down what you expect before you ship it. Every time
The whole thing collapses to this: narrow with data, decide with judgment, measure as honestly as your traffic allows, and keep the loop turning. Nobody gets the last part for free.