Network Selection and Setup: Impact, Awin, CJ, Rakuten

Two questions settle the network choice, and publisher count is not one of them. The commercial shape: Awin, CJ, Rakuten and ShareASale take a percentage override on top of every commission you pay, including to partners you recruited yourself and would have paid anyway. Impact and PartnerStack charge a platform fee plus a per-action rate and take no override on your own partners. If most of your first-year revenue will come from relationships you already have — true of nearly every automotive and B2B program we see — the override model charges you a percentage for hosting. If the program will genuinely be built out of a network's existing publisher base, the override buys something real. That is the whole decision, and it takes an afternoon with your numbers rather than a month of demos. What your product forces: Item-level commissioning — different rates by product group, or margin protection on discounted SKUs — is native on Impact and CJ, workable on Awin with the MasterTag and a properly formed conversion object, and awkward elsewhere. Recurring subscription revenue with a declining partner share is what PartnerStack is actually built around. A single-product, single-market merchant with two hundred partners needs none of this and belongs on the cheapest platform that reconciles. Setup, and what we test before go-live: Awin's MasterTag and ConversionPath, CJ's Universal Tag with the Actions API, Impact's tracking with itemised parameters, Rakuten's event tracking, PartnerStack's reward rules — every one of them ships client-side by default, and client-side is where European programs lose conversions. A visitor who declines consent does not fire the tag, so the sale is invisible to the network and plainly visible in your order table. We wire the server-side conversion path first and keep the tag as a secondary signal, then run the same five test transactions we run on every build: full order, partial refund, cancellation, Safari session with ITP active and consent declined, and a conversion delayed three days from the click. Migrations: Moving networks costs you partner history and part of your partner base. Say that out loud before you start. Historical EPC does not transfer, so the new marketplace shows you as untested. Partners must re-apply, and the ones producing revenue need a message from a person they recognise rather than a platform notification. We run both networks in parallel for four to six weeks with a de-duplication rule keyed on order ID so nothing is paid twice, then close the old one after the last validation window expires — not before, or you strand commission that partners are owed and they will remember it for years. Check the notice period first: most network agreements auto-renew annually on sixty to ninety days notice, and clients discover this in month one of a migration more often than not. What we refuse: We take no rebates, referral fees or override kickbacks from any network, which is precisely why we can tell you to stay where you are. We will not configure a network for a merchant who cannot fire a server-side conversion; that is different work and it comes first. We will not run two networks permanently as a hedge, because it doubles the de-duplication surface and partners find the seam. And we will not put a Czech-only merchant on a platform whose annual minimums exceed a realistic first-year commission bill. Not included: We prepare and coach you through the contract, including where the published rate card actually moves, but you are the counterparty and you sign. Creative production, feed engineering beyond the required schema, and partner recruitment are separate. Who this is not for: Anyone who has already chosen a network on the strength of a sales call and wants us to endorse it afterwards. Bring us the decision, not the signature.