The attribution conversation has gotten complicated since iOS 14 and the broader privacy shifts of the past three years. Most Shopify operators rely on some uneasy combination of platform-reported numbers, GA4 dashboards, and gut feel. The brands that scale efficiently in 2026 have a clearer attribution stack — not because they've solved attribution, but because they've stopped expecting attribution to be solved.
This guide breaks down the three main approaches — MMM, MTA, and GA4 — and how to combine them at each stage of growth.
What each method actually does
Marketing Mix Modeling (MMM) uses statistical models to estimate channel contribution from aggregate spend and revenue data. It doesn't need user-level tracking. It models the relationship between spend in each channel and revenue over time, controlling for seasonality, promotions, and other factors. MMM tells you what each channel contributes on average, with a confidence interval.
Multi-Touch Attribution (MTA) tracks individual user journeys across touchpoints and assigns fractional credit to each touchpoint. Pre-iOS-14, MTA worked reasonably well. Post-iOS-14, the user-level data is so fragmented that MTA outputs are heavily extrapolated.
GA4 attribution uses Google's algorithmic data-driven model to assign credit. It's a flavor of MTA, with Google-specific signal advantages and limitations. Free, accessible, and widely used.
Each method has different strengths. The mistake is picking one and treating it as truth.
When to use which
Under $30K/month total spend
Use GA4 plus platform-reported metrics. Keep it simple. Track MER as your overall efficiency check. Don't pay for attribution platforms or attempt MMM. Your spend is too small for the precision to matter, and you've got bigger leverage points (creative, product, channel selection).
$30K-100K/month spend
Layer in incrementality testing quarterly. GA4 plus platform metrics still your daily reporting layer. Start gathering data for future MMM (clean spend tracking, channel-level breakdowns weekly).
$100K-500K/month spend
MMM becomes worth the investment. Recast, Northbeam, or Prescient typically run $30K-80K/year — meaningful but not absurd at this spend level. Use MMM as your strategic allocation tool; use platform metrics for daily/weekly campaign management. Continue incrementality testing.
$500K+/month spend
Full attribution stack: MMM for strategic allocation, platform metrics for tactical campaign management, incrementality tests for verification, and possibly an MTA tool if your data is clean enough to support it (rare). At this scale, the cost of attribution tooling is small compared to the cost of getting allocation wrong.
What MMM is good at
- Telling you how much incremental revenue each channel is producing on average
- Modeling spend-response curves (saturation points by channel)
- Capturing brand and awareness effects that don't show in click-based attribution
- Working with privacy-respecting aggregate data
- Producing actionable allocation recommendations
What MMM is not good at
- Real-time decisions (the model lags 1-2 weeks)
- Granular insights (campaign-level or creative-level)
- Short-term tests (needs months of data to update)
- Capturing the impact of a single new creative or campaign
- Rapid-iteration environments where tactics shift weekly
What GA4 is good at
- Daily and weekly tactical reporting
- Path analysis (where users come from, what they do next)
- Cross-device journey patterns
- Free and integrated with most other Google products
- Useful for relative channel comparison
What GA4 is not good at
- Absolute revenue attribution to channels (overstates direct, understates assists)
- View-through attribution
- Aggregate brand effects
- Anything where you've blocked third-party cookies (most users)
- Cross-domain or app/web journeys without significant setup work
What MTA is good at (in 2026)
Honestly, less than people claim. MTA tools (Wicked Reports, Triple Whale, etc.) are decent dashboards aggregating multi-platform data. The "attribution" they provide is heavily extrapolated when user-level tracking is incomplete — which is almost always.
Use them for:
- Consolidating data across platforms in one dashboard
- First-party data layer that powers other measurement
- Customer journey visibility for retention and LTV analysis
Don't use them for:
- Channel-level allocation decisions, in isolation. Combine with MMM and incrementality tests.
Building a working stack
For a typical mid-stage Shopify brand at $200K-500K/month spend, the stack we recommend:
Daily / weekly: Platform-reported metrics + GA4. These are your tactical dashboards.
Monthly: MER and CAC trends (blended). Watch the leading indicators that span platforms.
Quarterly: Incrementality test on one channel. MMM model refresh (if you've got one).
Annually: Full channel-mix review using MMM outputs as the primary input. Adjust strategic allocation based on saturation curves.
The principle: short-term decisions on quick data, long-term decisions on slow data. Don't reverse them.
What "MER" actually means
MER (Marketing Efficiency Ratio) is total revenue divided by total ad spend. It's blunt — it doesn't tell you which channel is working. But it's blunt in a useful way: it can't lie the way platform-reported ROAS can.
Track MER weekly. If MER is stable or rising while spend grows, you're scaling efficiently. If MER drops as spend grows, you're hitting saturation regardless of what individual platform dashboards say.
For most brands, target MER:
- Pre-product-market-fit: 1.5-2.5
- Growth stage: 2.5-3.5
- Mature DTC: 3.5-5.0+
Common attribution mistakes
Trusting platform-reported ROAS in isolation. Every platform overstates. The math doesn't work — if Meta says you're at 4x ROAS and Google says 5x ROAS, but blended MER is 2.5x, somebody's lying. Probably both.
Switching attribution models when results are bad. "Maybe last-click is wrong, let me try first-click" is just shopping for a model that supports the answer you want.
Buying expensive attribution software at $50K/month spend. The tooling won't fix the precision problem. Save the money for incrementality tests.
Treating GA4 as ground truth. It's a useful directional tool, not a final answer.
Ignoring incrementality entirely. No matter what your attribution stack looks like, periodic incrementality tests are the only way to ground-truth your assumptions.
What to do this week
Calculate your blended MER for the last 90 days. Plot it weekly. If you don't have a clean MER trend line, that's your starting point — fix the basic measurement before investing in fancier attribution tools.
Then schedule one incrementality test for next quarter. Pick the channel you're most uncertain about. The test will teach you more than any new attribution platform.
For more, see our geo-holdout incrementality testing guide, the first-party data strategy, and the Shopify attribution models explained for foundational background.