Decide — Next-best incentive decisioning
    Decide

    Recommend the next-best incentive for the real customer state.

    The Decide layer produces an Incentive Decision Record for every recommendation — mechanism, cap, audience, assumptions, and how it will be measured.

    Illustrative view of a next-best incentive recommendation with confidence and assumptions
    Customer state model
    Recency, frequency, margin, and intent — reasoned about together, not in silos.
    Mechanism selection
    Value credit, service, timing, or advocacy — chosen against a merchant policy.
    Confidence + assumptions
    Every recommendation labels its confidence and states what it assumes.
    Auditable decision log
    Every decision, override, and outcome is written line-by-line for review.
    What Decide does not do

    It does not ship anything unattended.

    Every recommendation is a proposal. Merchant approval — or a pre-agreed autonomous window with a hard cap — is required before it reaches a customer.

    Decisions can be reversed and cohorts can be rolled back. That is the point.

    • No blind autonomy — approvals or capped windows
    • Reversible: every decision can be paused and rolled back
    • Confidence is disclosed, not hidden

    See a decision recommended live.

    Pick one customer behavior. See the recommendation, the assumptions, and the proof plan before anything ships.