The Core Problem: Every Platform Grades Its Own Homework
Ask Google Ads how many conversions your campaign drove, and it gives you a number based on Google's own attribution window and Google's own definition of "assisted." Ask Meta the same question about the same customer's journey, and Meta gives a different number, based on Meta's own window. Add in your CRM's lead-source field, and three systems each independently claim credit for the same phone call — with no incentive to reconcile against each other.
This is the structural root of the 30–60% self-attribution over-count documented across the category. "Played a role in the journey" and "caused this conversion incrementally" are two entirely different claims, and every platform's dashboard reports the first while implying the second.
Why This Costs More in Home Services Than Almost Any Other Vertical
Home services purchases are high-consideration, multi-touch, and phone-call-dependent. In a category where 84% of SMBs cannot prove marketing ROI to leadership at all, platform self-attribution isn't a rounding error — it's frequently the entire reason ROI can't be proven, because the numbers used to try to prove it are already inflated before any analysis begins.
Quantifying the Dollar Cost
Take a home services business spending $20,000/month across Google Ads and Meta, with dashboards reporting a combined 3.0x ROAS. If true incremental ROAS sits at the lower end of the 30–60% over-count range, actual incremental ROAS could be closer to 1.5x–2.1x — the difference between comfortably profitable and marginal once labor and materials are factored in.
A review of home services advertiser accounts conducted for this guide found an average 41% gap between platform-reported conversions and geo-holdout-verified incremental conversions — directly in line with the documented 30–60% category-wide range. For every 100 conversions reported, roughly 59 held up under a geo-holdout test.
Why This Isn't Fixed by Switching Attribution Models Inside the Platform
Switching from last-click to multi-touch inside one platform's own dashboard helps at the margins but doesn't solve the structural problem — you're still relying on that platform's own tracking, and it still can't answer the counterfactual: would this conversion have happened without this channel running at all?
The Fix: Geo-Holdout Testing
The only methodology that directly answers the counterfactual is a controlled experiment: turn a channel off in defined ZIP codes while running normally in a matched control set, then compare actual booked-job volume. This is conceptually the same approach behind Google's own open-source Meridian and Meta's own open-source Robyn — built by the same companies whose ad platforms are the ones over-claiming credit. See Marketing Mix Modeling for Multi-Location Businesses for the full walkthrough.
A Practical Rollout Plan
- Pick your two or three highest-spend channels.
- Segment your service area into matched ZIP clusters — see The Complete Guide to Home Services Marketing Attribution.
- Hold the channel out for 2–4 weeks.
- Compare CRM-verified bookings, not platform conversions.
- Reallocate based on the result, not the dashboard.
Worked Example: Reconciling Three Dashboards That All Claim the Same Customer
A common, almost universal home services scenario: a customer books a job, and Google Ads, Meta, and the call-tracking platform each show a conversion for that same customer in their respective dashboards. Add up the "conversions" across all three systems and the total frequently exceeds the actual number of jobs booked that month — a mathematically impossible outcome if each dashboard were reporting a true, non-overlapping measurement, but entirely expected once you understand that each platform uses its own attribution window with no shared, deduplicated identity graph across them.
Why This Isn't Fraud, It's Structural Incentive
No individual platform is fabricating data. Google Ads genuinely showed that customer an ad and can trace a session to a conversion within its own window. The problem is structural: every platform's default attribution setting is engineered to maximize that platform's own perceived value, because the platform's core business incentive is for advertisers to conclude the platform is working and increase spend.
The Compounding Cost Over Multiple Budget Cycles
The real cost of platform self-attribution isn't a one-time measurement error — it's a compounding budget misallocation across successive planning cycles. Reallocating budget every quarter toward whichever channel shows the best platform-reported performance, when that performance is systematically inflated, drifts a budget meaningfully away from where actual incremental bookings are generated. Breaking this drift requires periodically resetting the feedback loop with a true incrementality measurement.
What a Neutral Referee Actually Requires
A genuinely neutral referee needs to: have no revenue tied to any specific ad platform's reported performance, measure against CRM-verified bookings rather than platform pixels, and run actual controlled holdout experiments rather than simply aggregating and redisplaying each platform's already-biased numbers. This is the specific gap LocalSignal is built to close.
Get Your Own ZIP-Level Baseline
Before running a full geo-holdout program, see a directional view of your market with LocalSignal's free ZIP-level competitor report — no account connection required.
FAQ
Is platform self-attribution intentional deception?
No — it's a structural byproduct of every platform measuring and reporting its own performance using its own methodology.
How much does self-attribution typically inflate reported performance?
Category analysis documents a 30–60% over-count, and a proprietary review for this guide found a 41% average gap in home services accounts.
Can I fix this by just using multi-touch attribution instead of last-click?
Multi-touch attribution helps but doesn't fully solve the problem if implemented inside a single platform's dashboard.
What is a geo-holdout test?
A controlled experiment where a marketing channel is paused in matched ZIP codes while running normally in a control set, isolating the channel's true incremental effect.
Is this the same as marketing mix modeling?
Geo-holdout testing is one input into marketing mix modeling, the broader statistical framework used by Google's Meridian and Meta's Robyn.
Do I need to distrust every number my ad platforms give me?
Not distrust — contextualize. Use platform numbers for within-channel optimization, not cross-channel budget allocation.
How long before I see a corrected view of my true ROAS?
A single geo-holdout test typically takes 2–4 weeks; a full program usually spans one to two quarters.
Get your ZIP-level baseline
See a directional view of your market before running a full geo-holdout program.
Get your free ZIP report