Revenue Share Programme Managed Against a Holdout

Across the last four programmes I inherited — two apparel retailers, an app subscription business and a donor programme — between 31 and 46 percent of what the network reported as affiliate-driven revenue survived a suppression holdout. The rest was coupon extensions and loyalty toolbars claiming a click on a basket the shopper had already filled. I manage revenue share against the measured number rather than the network dashboard, so the first thing you receive is the size of your own gap. It usually changes the commission conversation before it changes the partner list. The first six weeks: I take twelve months of transaction-level data out of Awin, Impact.com, Partnerize or Everflow, join it to your order table on order ID, and sort every partner by where its click sat relative to the first add-to-cart. Then the holdout: coupon and cashback partners suppressed for a randomised share of sessions, or across Ireland and one matched market, run long enough to read a delta on order count rather than on revenue. Four to six weeks is normal. A shorter read flatters the partners you were hoping to keep, which is why short reads are popular. How partners get paid after that: The single flat rate goes. New and returning customers separate, and the returning rate is the one your finance team will argue about, so I bring the repeat-purchase data to that meeting rather than an opinion. A content partner whose review ranks and holds gets a rate that makes the placement worth defending internally. An extension firing at checkout gets a de-duplication rule and, if it still earns one, a fixed fee instead of a percentage. On iOS the arithmetic changes: with AdAttributionKit and SKAN postbacks you cannot pay a share of a lifetime value you cannot see per partner, so those partners move to a fixed amount per trial-to-paid conversion, reconciled monthly against AppsFlyer or Adjust cohorts. For donor programmes I will not pay on a first gift — a bought first gift churns — the payment attaches to the second monthly debit. What arrives each month: A partner-level P&L with media value, commission, network fee and the holdout-adjusted contribution in the same table. A named list of partners I am ending and the reason. Rate changes I intend to make next month, sent before I make them, because a partner manager who learns of a rate cut from a dashboard becomes an expensive enemy. Voucher-code leakage checks against your own brand terms, and a note when a partner is bidding on them. Not included: Paid search, paid social and influencer whitelisting. Network migration, though I will tell you plainly whether yours is worth leaving and what the transition costs in lost partner links. Creative production. Recruitment measured in partner count — I will bring in partners where the model says there is margin for both sides, and some months that is two. Who should not hire me: A brand whose finance function will not release order-level data. Nothing here works from platform exports alone, and I would rather say that now than three weeks in. A programme whose actual goal is a larger partner number in a board pack. And any team that has already decided the coupon partner stays for relationship reasons — a defensible decision, but it makes the measurement I sell unreadable, and you would be paying me for a report you cannot act on.