How we optimize

A loop that shows what to keep, change or stop.

Standing policies refresh as data changes. Randomized holdouts show which ones caused revenue. The team revises the policy on that evidence, and the loop runs again.

  • Timing from actual purchasing

    Reorder timing comes from your customers' real purchase patterns. A customer enters the audience when they become eligible, not when someone next runs a report.

  • Audiences that stay current

    Eligibility is re-evaluated as data changes. Customers enter and leave automatically when they purchase, cancel, lapse, change consent or submit a quiz.

  • Lift decides what continues

    Treated-versus-holdout results show which campaigns caused revenue. That tells the team whether a campaign should run, who should receive it and whether it created economic value.

  • Every change is reviewable

    Each condition has a readable definition, and each customer record shows why they qualified or did not. Marketing can review, approve, revise or audit the policy.

Policy dials

Three values change what a policy does.

Every campaign policy is a readable set of conditions. Three of them are dials: the minimum historical value, the contact-fatigue window and the holdout share. The rest stay explicit and auditable.

overdue_reorder_policyIllustrative
  • completed_orders>=2
  • days_past_expected_reorder>0
  • historical_value>=X
  • email_consent=valid
  • days_since_relevant_campaign>Y
  • active_exclusions=none
  • holdout_share=Z%
  • XMinimum historical value
  • YContact-fatigue window, in days
  • ZShare of eligible customers held out

Start with an audit of the revenue you are already leaving behind.

Request an audit