What We See Across 1,000 Accounts: Attribution Patterns

Admaxxer is a DTC analytics platform with built-in Meta + Google ad ops. When you look across enough accounts, a handful of attribution patterns keep repeating — not because every brand is identical, but because every platform's attribution model has the same structural incentives. This post names six of the patterns we see most often. ## The claim - **Meta systematically over-credits top-of-funnel**; prospecting ROAS is usually inflated by 20–40%. - **Google systematically over-credits bottom-of-funnel**; branded search takes credit for traffic the rest of the stack created. - **Organic picks up the leak** between platform-estimated conversions and Shopify's truth. - **Email is under-credited nearly everywhere** — last-click gives credit to paid; MTA models often lack email as a touchpoint. - **CAPI lifts add 15–30%** to matched conversion signal when implemented correctly. - **Ad-level LTV cohorts diverge sharply after ~30 days**, which changes the "winning ad" definition. ## Why platforms over- and under-credit Every platform's attribution model is built to make the platform look good. Meta's default 7-day-click-1-day-view window includes view-through conversions where the user saw an ad but clicked something else to convert. Google's last-click default gives branded search disproportionate credit. Neither is "wrong" — they're just optimizing for different things than you are. The fix isn't to pick the one true attribution model. It's to **hold blended MER as the honest anchor** and use platform views as directional inputs. See our [blended vs multi-touch deep dive](/blog/blended-vs-multi-touch-attribution). ## Pattern 1: Meta over-credits prospecting Across accounts we benchmark, the sum of Meta-reported conversions from prospecting campaigns typically exceeds the number of genuinely new customers Shopify sees by **20–40%**. The gap comes from view-through double counting (a user sees a prospecting ad, later clicks retargeting, Meta credits both), and from attribution to users who were already on the customer list. The practical effect: if you optimize prospecting ad sets purely on Meta-reported ROAS, you will over-invest in creative that doesn't actually find new people. The cleaner test: **new-customer MER**, calculated as (new-customer revenue in Shopify) / (prospecting spend), over a rolling 7 or 14-day window. ## Pattern 2: Google over-credits branded search If branded search is running at 8× ROAS, congratulations — it's correctly reflecting demand that already existed. Google will happily take credit for every branded click, but most of that intent was created by Meta prospecting, organic content, email, and word of mouth. In accounts we benchmark, branded-search ROAS is typically **inflated by 40–60%** when compared against incrementality tests. This doesn't mean turn branded off (that usually costs you conversions to competitors bidding on your brand). It means: don't scale other Google campaigns based on branded ROAS mixed into the same account view. ## Pattern 3: Organic picks up the leak When platform-reported paid conversions don't match Shopify total revenue, the delta usually shows up in the "direct" or "organic" bucket. Across accounts we benchmark, organic traffic typically grows 10–20% as paid traffic grows — not because SEO is improving, but because users who saw paid ads return via direct navigation, branded search, or organic social. This is the single biggest reason to anchor on blended MER: it captures the leak automatically. ## Pattern 4: Email is chronically under-credited Every platform last-click attribution model strips email out of the credit path. If a user clicks a Meta ad, adds to cart, abandons, receives a Klaviyo flow, clicks back and converts — Klaviyo gets last-click, but most multi-touch models downstream of Meta only see a Meta session re-entry and call it paid. When we add email properly to a blended model, it typically takes **8–18% of total credit**, almost all of which was previously being absorbed by either paid or "organic." ## Pattern 5: CAPI match rate changes the signal Brands without clean CAPI typically run at **55–65% event match rate**. Brands with a properly deployed CAPI pipeline — `event_id` deduplication, server-side hashed PII, `fbp`/`fbc` passthrough — typically run at **75–88%**. That 15–30 point lift is the difference between a model that sees most of your conversions and one that is guessing. The real kicker: the lift doesn't just show up in reporting. Meta's bidding model uses the same signal, so match-rate improvements usually translate to better prospecting efficiency within 7–14 days. See [our CAPI match rate feature](/features/capi-match-rate). ## Pattern 6: Ad-level LTV cohorts diverge after 30 days The ad that wins at 7-day ROAS is often not the ad that wins at 90-day revenue per acquired customer. Across accounts we benchmark, the top-ROAS-at-day-7 ad and the top-LTV-at-day-90 ad are the same ad only **about 40% of the time**. Discount-heavy creatives win early and lose late; brand-led creatives lose early and win late. This is why we show [ad-level LTV](/features/ad-level-ltv) cohorts (7/30/90d) in the creative grid — so you can see the divergence before you double down on the wrong winner. ## What to do about it 1. Anchor on **blended MER**, not per-platform ROAS. 2. Compute **new-customer MER** weekly and compare against Meta-reported prospecting ROAS. 3. Audit CAPI quarterly; fix match-rate gaps first before touching bid strategies. 4. Add **ad-level LTV** into your weekly creative review, not just 7-day ROAS. ## Caveats These patterns are archetypal — they hold across the bulk of DTC accounts we see, but not every one. Brands with very short consideration cycles (consumables, impulse buys) see less multi-touch leakage. Brands selling high-AOV considered goods see more. Apply the patterns with judgment, not as rules. ## FAQs **Q: Should I turn off branded search because it's over-credited?** A: Usually no — competitors will bid on your brand. Just don't use branded-search ROAS as a scaling signal. **Q: How do I compute new-customer MER?** A: (New customer revenue in Shopify, from the `customers.first_order_date` cohort) / (prospecting spend across Meta + Google), on a rolling 7 or 14-day window. **Q: Is CAPI worth the engineering time?** A: For almost every brand spending over $20k/month on Meta, yes. The lift is typically larger than most CRO experiments, and it compounds. --- **TRIAL_LINE:** Start your 7-day free trial — no credit card required. [See Admaxxer pricing](/pricing).