PerkFit

I. The audience

Example A: late-20s Brooklyn

AgeGender

Fleabag
Phoebe Bridgers
Glossier

These chips are exactly what we send to Qloo. No personal data.

II. The budget

Monthly budget ASSUMPTION

Members ASSUMPTION

Usage by taste match ASSUMPTION

HighMidLow

Assumed share of members using the perk. Per-use ranking does not depend on usage assumptions; only fixed-fee perks are sensitive to them.

Domain weights (0 to 2) ASSUMPTION

PlacesTV showsBrandsBooks

Also ask an LLM alone for a baseline bundle

III. Candidates

candidate perks, with price, effort and taste match

Example A, cached

Which perks should this audience get?

Pick partner perks for a customer segment by taste, inside a budget. Qloo ranks how well each perk fits the segment; costs and usage are editable assumptions.

Same 20 perks, two rulers: popularity cannot tell them apart, taste match can.

In the best bundleSkipped, low fitOther candidates

Agent trace

No agent steps in this run. Sample mode runs no agent.