Attribution Platform Setup Calibrated to Your Order Ledger

The first deliverable is a coverage number: the share of the revenue in your order table that the attribution tool can see and match to a click. On a first pass it usually lands between 70 and 92 percent, and every ROAS the platform prints afterwards is scaled by whatever that gap is. Most installs never measure it. This one measures it first and puts the figure in writing before a budget decision is hung on the tool. Which of the three: Triple Whale fits a Shopify-native business with a short consideration window and creative iterating in-house. Northbeam fits spend heavy enough across four or more channels that the live argument is incrementality between them rather than tracking accuracy. Hyros fits long, human-touched sales where the conversion is a call booked now and closed six weeks later. If you have already bought one, I work with what you bought. If you have not, I will say which of the three your data can actually feed — and for some businesses the answer is none of them, because an order-level export into a warehouse with a stated last-non-direct rule will be more honest and cost a fraction of the license. The install: Server-side collection through a container you control, with the browser tag as a secondary signal rather than the only path. Order events taken from cart webhooks — created, updated, refunded — not from a thank-you page that a third of mobile sessions never render. Click identifiers captured and written to the order record server-side, including gclid, fbclid, ttclid and the Yahoo! JAPAN Ads and LINE equivalents, so a purchase that finishes in a different browser three days later still carries its origin. Identity joined on order id first, hashed email second, platform customer id last, in that order of trust. Consent state stored beside every event, so a later audit can tell suppressed from missing. What calibration means here: Japanese carts settle late. Konbini payment and daibiki create an order the platform counts on the spot and a payment that arrives days later or never; on the wellness accounts I see, non-settlement across those two methods is large enough to move reported ROAS by a fifth. Retail energy has the same shape at a longer interval — the click is in March and the switch confirms through the back office in April. Calibration is the written set of rules deciding when an order counts, what a refund or cancellation does to a day already reported, and which margin and currency figure the tool receives. Agreed with finance, then implemented in the feed rather than argued about in the dashboard. Not included: Media buying and creative. Consent platform selection or any opinion on lawful basis — bring counsel and I implement what they specify. Rebuilding a checkout that does not emit order events. Warehouse modelling and reporting, which are separate engagements. The platform license, which you hold directly. Who should not buy this: Anyone who needs the tool to report a higher number than finance does. A correct install almost always produces a lower and later ROAS than the ad platforms show, and where that outcome is politically impossible the install gets quietly discredited inside two quarters. Also wrong for businesses too early to have a sample worth attributing. Handover: The event contract, the calibration rules, the coverage number with the method used to reach it, and a working session with whoever owns the tool next. I do not stay on as the only person who understands the setup.