--- title: Fix Duplicate Attribution Across Channels | Guide | RGM® url: https://realgrowthmatters.com/learn/measurement/deduplicating-multichannel-attribution/ updated: 2026-06-10 source_html: https://realgrowthmatters.com/learn/measurement/deduplicating-multichannel-attribution/ --- One sale. *Three trophies.* # Fixing Duplicate Attribution Across Multi-Channel Campaigns Run three channels and each one claims the same sale. Here is the dedupe ladder that turns three competing dashboards into one honest scoreboard. Duplicate attribution happens because every platform measures only its own touchpoints, applies its own windows, and takes full credit by default. You fix it with a ladder: one referee system for orders, shared rules, event IDs for deduplication, blended metrics as the scoreboard, and [holdout tests](/glossary/incrementality-testing/) as the tiebreaker. By David Schaefer · [LinkedIn](https://www.linkedin.com/in/daschaefer/) · Updated June 2026 ## The anatomy of a double-counted sale A buyer sees your TikTok ad, clicks your Google ad, then opens your email and purchases. TikTok logs a view-through conversion. Google logs a click conversion. The email platform logs one too. One order — three "conversions." Every dashboard is technically telling its own truth. The duplication is structural, not dishonest. Each platform is a walled garden: it sees its own touches, sets its own attribution window, and defaults to taking full credit for any conversion it touched. None of them can see the others. **Claim:** Observational, platform-side attribution can mis-state true ad lift by 3x or more versus randomized experiments. **Source:** [Gordon et al., Marketing Science (2019)](https://pubsonline.informs.org/doi/10.1287/mksc.2018.1135). **Context:** duplication plus self-grading is how a brand "earns" 4x ROAS on every dashboard while the bank account stays flat. ## The four mechanics behind the double count Four mechanics drive almost all duplication: overlapping windows, view-through credit, last-touch-per-platform defaults, and double-fired events. Name them and the fixes become obvious. Where duplicate credit comes from — and the matching fix | Mechanic | What happens | The fix | | Overlapping windows | A 30-day window on one platform overlaps a 7-day window on another; both claim the same order | Set comparable windows everywhere; document them | | View-through credit | An impression nobody clicked claims the sale | Discount view-through unless a lift test backs it | | Last-touch per platform | Each platform runs its own "last touch" — so every platform is the last touch | One referee system decides; platforms become claims | | Double-fired events | Pixel and server both send the same conversion | Shared event IDs so the platform deduplicates | > "Segment or die." > > — Avinash Kaushik, *Web Analytics 2.0* · [on honest data](/glossary/incrementality/) ## The dedupe ladder — five rungs, in order Work the ladder top to bottom. Each rung removes one class of double counting, and each is cheaper than the confusion it prevents. 1. **Appoint one referee.** Pick a single source of truth for orders — your analytics or your commerce backend. From this moment, platform numbers are claims to verify, never the scoreboard. This one decision ends most attribution meetings. 2. **Align the rules.** Same conversion definition and comparable attribution windows on every platform. Write the settings down. Mismatched windows are the quietest source of double counting in multi-channel accounts. 3. **Deduplicate events.** Send conversions from the browser and the server with a shared event ID, so the platform keeps exactly one copy. [Meta's Conversions API dedup](https://www.facebook.com/business/help/823677331451951) and [Google's enhanced conversions](https://support.google.com/google-ads/answer/9888656) both support this natively. 4. **Score with blended metrics.** [Blended CAC](/glossary/blended-cac/) and [MER](/glossary/marketing-efficiency-ratio-mer/) divide real revenue by total spend. No single platform can inflate them — which is exactly why finance teams trust them. 5. **Break ties with holdouts.** When two platforms claim the same sales, pause one in matched geographies and watch what actually disappears. The eBay experiments are the famous worked example: brand-keyword ads claimed sales that arrived organically anyway ([Blake, Nosko & Tadelis, 2015](https://www.nber.org/papers/w20171)). ## The math, end to end A worked example makes the inflation visible. Watch what happens when platform claims meet the referee. **Worked example.** A brand spends $54,000/mo across three channels. The dashboards report: paid social 120 conversions, paid search 95, email 45 — a claimed total of 260. The commerce backend (the referee) shows 180 actual orders. The inflation ratio is 260 ÷ 180 = 1.44 — about 44% over-claiming, typical for a three-channel mix. The honest scoreboard: blended CAC = $54,000 ÷ 180 = $300 per order. Now every platform's claimed CAC can be compared against the blended truth, and the gap between 260 and 180 becomes a weekly agenda item instead of an invisible tax. *(Illustrative model — RGM analysis.)* 260 Conversions the three dashboards claim 180 Orders the referee system actually recorded 1.44x Inflation ratio — claims ÷ reality $300 Blended CAC — the number that cannot lie ## What good looks like Healthy multi-channel reporting has one order count, platform claims that sum to more than it (overlap is normal — uncorrected overlap is not), a blended CAC trend everyone watches, and a testing calendar that settles disputes. The platforms argue. The referee decides. The holdout settles it. That is the entire governance model, and it fits on an index card. For the full measurement stack — attribution, incrementality, and media mix modeling working together — read the companion guide: [how to measure performance marketing](/learn/measurement/how-to-measure-performance-marketing/). ## The mistakes that undo it **Letting every platform keep its defaults.** Default windows maximize each platform's credit, not your clarity. Set them deliberately and write them down. **Banking view-through conversions at face value.** They are the most over-claimed conversion type. Discount them until a lift test earns them back. **Reporting platform totals to the board.** The day finance notices conversions exceed orders, every marketing number you have ever shown loses credibility. Lead with blended. ## Quick answers What is duplicate attribution? : Two or more platforms each claiming credit for the same sale. Add up the dashboards and you get more conversions than real orders. What is the fastest fix? : Appoint one referee system for orders and demote platform numbers to claims. Then work the ladder: shared rules, event IDs, blended scoreboard, holdouts. What is the single best metric for multi-channel truth? : Blended CAC or MER — real revenue divided by total spend. No single platform can inflate it. ## Frequently asked Why do my platform conversions add up to more than my orders? Each platform measures only its own touchpoints with its own windows and defaults to full credit. Overlapping journeys get counted once per platform — so the totals always exceed reality. Are view-through conversions real? Sometimes — but they are the most over-claimed conversion type. Discount them unless an incrementality test supports the channel's claim. Does GA4 solve deduplication by itself? It helps as a referee because it applies one attribution model across channels. But it has its own blind spots — view-through activity and walled-garden touches it cannot see — so pair it with blended metrics and holdouts. How do event IDs deduplicate conversions? When the browser pixel and the server both send the same conversion, a shared event ID tells the platform they are one event, and it keeps a single copy. Meta and Google both document this pattern. ### Keep reading ## Sources 1. [Gordon, Zettelmeyer, Bhargava & Chapsky — Marketing Science (2019)](https://pubsonline.informs.org/doi/10.1287/mksc.2018.1135) 2. [Blake, Nosko & Tadelis — Consumer Heterogeneity and Paid Search Effectiveness, Econometrica (2015)](https://www.nber.org/papers/w20171) 3. [Meta — Conversions API event deduplication documentation](https://www.facebook.com/business/help/823677331451951) 4. [Google — Enhanced conversions documentation](https://support.google.com/google-ads/answer/9888656) 5. [Google — GA4 attribution and attribution settings documentation](https://support.google.com/analytics/answer/10596866)