When the numbers disagree, we trace the transaction.
From click to revenue — across the tracker, the funnel, the checkout, the processor and the subscription system. Native traffic attribution is our deepest specialism; the rest of the stack is the work that surrounds it.
- CLICK
- REDIRECT
- SESSION
- ORDER
- UPSELL
- REBILL
- REVENUE
Most attribution problems aren’t pixel problems.
They’re state, routing and data problems — and they live between platforms, not inside them. These are the symptoms we get called about, and what’s usually underneath them.
Meta reports far more purchases than the checkout recorded.
The browser pixel and the server event both fire for the same order without a shared event ID, so nothing deduplicates them. Reported ROAS inflates and budget follows the wrong number.
A campaign you know is profitable reports as a loser.
Conversions reach your tracker but never reach the network — or arrive without payout — so the network optimizes against incomplete data and the report understates the truth.
Revenue exists in the checkout but the tracker shows nothing.
The click ID didn't survive a redirect hop, a consent layer or the jump to the checkout domain. The order is real; it simply has no campaign left to attach itself to.
Upsell revenue is missing from campaign reporting.
Upsells post back without a transaction ID, or reuse the parent order's, so downstream systems discard them as duplicates. The most profitable part of the funnel goes uncounted.
Subscription rebills never appear against the original campaign.
Nothing posts back on renewal. LTV-positive acquisition gets judged on first-order ROAS alone, and campaigns that pay back in month two get cut in week one.
Two platforms both claim the same conversion.
No agreed source of truth. Until orders are reconciled against processor records, every dashboard is a differently-wrong guess and no one can say which to trust.
Native traffic is where attribution breaks most — and where we’re deepest.
Native networks don’t behave like Meta. There is no pixel-first SDK quietly doing the measurement for you: identity travels as a URL parameter, conversions only exist if your server sends them back, and every hop between the ad and the order is a place the chain can break. All of it is yours to build — which is precisely why it is worth having built properly.
- No pixel does the work for you — identity arrives as a click-ID macro in the destination URL and exists only if you capture it.
- Redirect chains: network → tracker → presell → lander → checkout. The parameter has to survive every hop, including the cross-domain one.
- Conversions return by server-to-server postback, not a browser event. If your server never fires it, the network never learns the sale happened.
- Postbacks usually need payout and an event name, not just a click ID. Without payout, the network's optimization is bidding blind.
- Section, site and publisher IDs have to reach your tracker — otherwise you cannot see, or block, the placements burning budget.
- Multi-step funnels multiply the hops. Advertorial → lander → checkout → upsell is four chances for state to drop.
- Every network reports on its own schedule with its own attribution window, so cross-network numbers never reconcile by themselves.
- Custom conversion definitions differ per network and have to be mapped deliberately, not left on defaults.
Every hop is a place attribution can die. The engineering job is making each one boring.
Usually with Voluum in the middle, a multi-step funnel in front of CheckoutChamp, and subscription rebills that need attributing back to the click that started them. We are an engineering partner to teams buying native at scale — not a media buyer, and not an ads agency.
Perfect attribution isn’t available. Reaching the limit is.
Browsers, consent layers and network policies put a hard ceiling on what can be tracked, and anyone promising you total certainty is selling something. The engineering job is to reach that ceiling — then reconcile everything above it at the order level, against the processor, which is the one record that can’t be argued with.
What we build and repair.
ARCHTracking architecture
Design of the full measurement layer: which platform is the source of truth for spend, clicks, orders and revenue — and how data moves between them.
VOLUUMVoluum implementation
Campaign, flow and path setup, S2S postback URLs, custom conversions, upsell and rebill tracking via transaction IDs, API-based reporting.
S2SS2S postbacks & conversion APIs
Server-to-server conversion delivery to Meta CAPI, native networks and trackers — with deduplication against browser events.
EDGEServer-side endpoints
Custom tracking endpoints and Cloudflare Workers that capture, enrich and route events independently of the browser.
PERSISTClick-ID & cross-domain persistence
Click IDs carried across redirect chains, landers, checkout domains, upsell steps and into the subscription record.
RECONRevenue reconciliation
Order-level comparison between tracker, checkout, store and processor — locating exactly where numbers diverge and why.
DEBUGEvent debugging
Duplicate and missing conversion diagnosis, webhook processing, event pipeline repair, reporting reconciliation between platforms.
Reporting doesn't add up?
A tracking audit traces real transactions across your stack, documents where the data breaks, and specifies the fix — before anything gets rebuilt.