Question 1

How do you recognize it?

Accepted Answer

Quarterly client reviews still open with a platform-reported ROAS slide, and nobody in the room can say what share of that revenue would have happened without the spend. Budget reallocation decisions get made on 7-day click windows while the client's actual sales cycle runs 45 to 90 days, so upper-funnel channels get cut first and pipeline dries up two quarters later. Two dashboards disagree by 30 percent or more (ad platform versus CRM closed-won) and the account team quietly picks whichever number makes the month look better. Incrementality tests were run once during onboarding, then never repeated, so the agency is defending a 2024 baseline against 2026 auction dynamics. Clients start asking for holdout tests or geo experiments in procurement conversations, which means the measurement gap has already reached the buying committee.

Question 2

Why does it happen?

Accepted Answer

Last-click is the default in every ad platform UI, so it wins by inertia. Multi-touch attribution and media mix modeling require someone to own a data model, and most retainer scopes never budgeted that role. Agency compensation is tied to reported performance, which creates a structural disincentive to adopt measurement that might shrink the number the client sees. Admitting 40 percent of attributed conversions were organic is a hard sell mid-contract. Black-box model outputs get presented without methodology notes, so when a client's in-house analyst challenges the numbers the agency has no defensible answer and the credibility loss outlasts the reporting error. Data plumbing is treated as a one-time setup. Server-side tracking, CRM joins, and offline conversion imports decay as the client changes forms, adds a chat widget, or migrates their CRM.

Question 3

How do you fix it?

Accepted Answer

Run one geo holdout or spend-level test on the client's largest channel this month and report the incremental lift number next to the platform-attributed number. SegmentStream and Measured both support experiment design calibrated against modeled ROI. Publish a one-page methodology appendix with every attribution deliverable: model type, lookback window, which touchpoints are excluded, and known blind spots. Ruler Analytics and Prescient AI both expose model inputs that can be documented rather than hidden. Reconcile ad platform revenue to CRM closed-won for the trailing 90 days and present the variance openly. Heeet and Cometly both sit inside Salesforce and HubSpot, which makes the pipeline join auditable instead of asserted. Put measurement maintenance in the retainer scope as a named line item with hours attached. If the client will not fund it, downgrade the reporting promise in writing rather than absorbing the decay silently.

More for Attribution Analytics