On this page
Playbook
Dylan Karaitiana 19 Apr 2026 4 min read Attribution drift is a tax operators pay when nobody is watching where leads actually come from. A 30% gap between “what the ad platform reports” and “what the bank account shows” is the norm, not the exception. The tax is real money — operators over-spend on channels that look efficient and under-spend on channels that don’t get full credit.
We close the gap with a playbook that doesn’t require a consultancy. It also doesn’t require a $50K/year attribution platform. It does require discipline, which is harder to sell than software.
The three sources of drift
Most operators we work with have attribution gaps from three places stacked on top of each other.
Browser-side measurement decay
Safari ITP, Firefox ETP, ad-blockers, iOS link decoration stripping. The platform tag fires from the browser; the browser blocks or attenuates the signal; the platform reports fewer conversions than actually happened. Net effect: paid platforms under-report, operator concludes their spend is less efficient than it is.
Last-click attribution defaults
Most platforms still default to last-click. The operator runs a YouTube brand campaign that drives high-funnel attention; the buyer comes back via a Google search a week later; Google takes credit for the close. YouTube looks dead, gets cut. The brand part of the operating system underperforms.
System-of-record drift
The CRM has different revenue numbers from the finance system. The finance system has different revenue from the bank. Reconciliation lives in someone’s head. When attribution rolls up, every report uses a slightly different denominator, and the gap compounds.
The three moves we make
We run the same playbook on every project that touches the visibility part of the operating system. It’s not glamorous. It works.
Move one — Server-side container
Move conversion tracking server-side. We use GTM server container by default; for operators on Shopify Plus we use the Shopify Custom Pixels with a server endpoint. The point is that the conversion event leaves the operator’s server, not the customer’s browser. ITP and ad-blockers stop attenuating the signal. Most operators recover 15–25% of conversion volume in week one.
Move two — Clean attribution model
Server-side data goes into one analytics destination (we default to GA4, swap to a warehouse-driven model when volume justifies). Attribution model gets specified explicitly — we use a simple time-decay default with a 30-day window for most operators. When volume is high enough, we move to data-driven attribution; when it’s not, we stay with time-decay because the model needs enough data to learn from.
Move three — Reconciliation cadence
Weekly snapshot. Three columns: platform-reported revenue, GA4-reported revenue, finance-system revenue. Run for the prior 7 days, run again at month-end against the prior month for trend.
If the gap between any two columns is over 5%, we triage. Could be a platform-tag misfire, a server-event drop, a CRM stage that didn’t auto-progress. The triage is the value — it’s the human attention that catches drift before it compounds.
The artifact
The deliverable isn’t the dashboard, it’s the reconciliation report. We send the operator a one-page weekly with the three numbers, the gap, and a note on anything we triaged. The operator reads it in two minutes, files it, has the receipt for the CFO conversation.
Operators who’ve used reporting platforms before tell us this is the part that surprised them most — the value isn’t in seeing more data, it’s in the gap being acknowledged in writing each week. When the gap is 2% three weeks running, the CFO trusts the marketing budget. When it’s 28% one week unexplained, the CFO wants the conversation. Either way, the operator and the CFO are looking at the same number.
What we don’t recommend
A few things we deliberately don’t do.
We don’t recommend a separate attribution platform unless the operator’s volume genuinely justifies it. For most projects in the $500K–$10M band, GA4 + a weekly reconciliation does the job. Adding a $30K/year platform on top is the consultancy tax we’re trying to remove.
We don’t recommend last-click as a default. Even if the data is too sparse for a model, time-decay with a 30-day window beats last-click in 95% of the projects we’ve audited.
We don’t recommend “tighter attribution” as a goal in itself. The goal is reconciliation — finance, marketing, and the platform agreeing on the same number. Tighter attribution is a means to that, not the end.
The summary
Server-side, modelled, reconciled. Three moves, weekly cadence, a one-page artifact. No consultancy, no platform tax, no ongoing service fee. The playbook is the work; the discipline is in actually running it every week.