Crafte Sessions analysis 2026 09 29 2026 10 29

sessions_analysis-2026-09-29-2026-10-29.csv · 2,246 rows · 15 columns · definition v3 saved 29 Oct 2026 14:08

Answer
The biggest lever is landingPage. If everyone had /product/75-christmas-cake-boxed, the model expects 11.3% to have ordered instead of 3.3% today (+8.0 points).
Association, not cause
These are the model's associations, not measured causes: people who arrive via one source may differ in ways the data doesn't hold. Test the top lever before rolling it out.
Suggested test
To check landingPage = /product/75-christmas-cake-boxed: about 489 people in each group (change vs no change) to detect a move from 3.3% to 7.3% (half the model's estimate, 95% confidence, 80% power).

Levers, biggest first

Model's average prediction for 2,232 snapshots
The biggest lever is landingPage. If everyone had /product/75-christmas-cake-boxed, the model expects 11.3% to have ordered instead of 3.3% today (+8.0 points).
LeverWorst to bestRangeBest value
other (1.6%) to /product/75-christmas-cake-boxed (11.3%)+9.7 points/product/75-christmas-cake-boxed → 11.3%

What can you change?

Tick what your business could actually change: where people come from, which page they land on, an offer, who handles a case. Things you can't change (device, tenure) can't be levers.
Model
Using Logistic (all inputs) from model run 30 Oct 2026 14:32 (feature set v9, at the start of a session).