90 days post-launch, 41 pilot catalogs
Across apparel, electronics, home
Suggested, not blindly applied
Median R² on held-out weeks
Timeline
- 2025 Q3Thesis
Two operators, one spreadsheet, and 40,000 rows of order lines from a friend's apparel brand. A third of the catalog turned out to be price-insensitive.
- 2026 Q1Prototype
First hierarchical elasticity model. Backtests on eleven catalogs recover held-out demand within 0.84 median R².
- 2026 Q2Pilots
Forty-one catalogs under model across apparel, electronics and home. Median contribution-margin lift of 6.4% at flat units.
- 2026 Q4Cohort 03
Onboarding twenty-five brands at a time so every account gets a human reading its first suggestion set.
How we decide things
Evidence over opinion
01Every recommendation ships with its confidence and the data behind it. If we can't defend a number, we mark it 'hold'.
Margin over volume
02Revenue is vanity when the discount was unnecessary. We optimise contribution margin and report units honestly.
Bounded autonomy
03The model proposes; your guardrails dispose. Floors, ceilings and daily caps are yours, not ours.
Legible by default
04Monospace numbers, visible assumptions, exportable history. Pricing that can't be audited won't be trusted.
Talk to the people building it
Pre-launch, every conversation is with a founder. Send us a catalog export and we'll tell you honestly whether elasticity modelling will move your numbers.