Ask Sellforte MCP: Since when do GA4 and our MMM disagree?
Welcome to the Sellforte MCP Use Case Library. Each post takes one question that marketing teams really ask their AI assistant through Sellforte MCP, which is now in general availability and accessible to all marketers worldwide. It strips out anything customer-specific and walks through what the assistant does with it, step by step. We run every question on our demo account so the numbers you see are illustrative, not anyone's real results. The idea is simple: steal the prompts, try them on your own model, and see what you get back. We hope you find this series inspiring!
-Juha, CEO at Sellforte
When GA4, the ad platform and the MMM stop agreeing
Here's a familiar one. Your GA4 report says a channel is suddenly doing much better, the ad platform says roughly the same as before, and the marketing mix model says something else again. Before anyone moves budget, the analyst who reconciles these sources needs to know two things: is the gap real, and when did it start?
What you ask
We see a gap in ROI and sales between the MMM, Meta Ads and GA4 for Meta. Do you see it too, and since when? Look at the last 12 months by month.
Show me the weekly trend from the start of September and tell me which week the sources start to diverge.
What Sellforte MCP does
Lines up the three sources month by month
The assistant pulls Meta Ads spend with three views of return side by side: incremental sales and ROI from the MMM, attributed sales and ROI from GA4, and revenue and ROAS as reported by the platform. It first resolves "Meta" to the right ad platform name in the account, then widens the window to whole months and labels November 2026 as partial, since the data ends on 2026-11-01.
On the demo account the three sources sit on very different scales, so the assistant compares how each one moves rather than their levels. Over the 12 months, Meta spend went down 57.2% from November 2025 to October 2026. MMM incremental sales fell 30.0% and platform revenue 72.5%. GA4 attributed sales went the other way, up 171.2%.
The month that stands out is October 2026. Compared with September, MMM ROI fell from 6.53 to 5.58 and platform ROAS from 71.98 to 63.23, while GA4 ROI more than doubled, from 0.461 to 0.994. So yes, there's a gap, and it's GA4 that's moving differently.
Narrows it down to the week
The follow-up switches to weeks (the demo account's weeks start on Sunday):
| Week starting (illustrative demo data, USD) | Spend | MMM ROI | GA4 ROI | Platform ROAS | GA4 attributed sales |
|---|---|---|---|---|---|
| 2026-09-06 | 64 744 | 7.51 | 0.30 | 78.77 | 19 449 |
| 2026-09-13 | 59 355 | 8.33 | 0.452 | 88.24 | 26 842 |
| 2026-09-20 | 49 895 | 4.17 | 0.624 | 55.49 | 31 127 |
| 2026-09-27 | 44 634 | 5.83 | 0.924 | 68.63 | 41 244 |
| 2026-10-04 | 48 211 | 5.79 | 1.15 | 70.11 | 55 569 |
| 2026-10-11 | 50 808 | 6.59 | 0.992 | 67.40 | 50 416 |
| 2026-10-18 | 47 873 | 5.17 | 1.08 | 61.80 | 51 615 |
| 2026-10-25 | 49 303 | 4.83 | 0.762 | 54.83 | 37 570 |
Up to the week of 2026-09-13, all three sources move together. In the week of 2026-09-20, spend drops, and MMM ROI and platform ROAS both fall with it. GA4 doesn't. Its attributed sales keep climbing for three more weeks, from 26 842 to 55 569, while spend stays lower than in early September. That puts the break in the two weeks starting 2026-09-13 and 2026-09-20.
Says what it can't tell you
The assistant is upfront about two limits. It sees the GA4 attribution held in Sellforte, not your own GA4 reports. And it can find when the sources diverge, not why. A GA4 jump with no matching move in the model or the platform usually points to something on the measurement side, like a tracking change, consent settings or attribution settings, so it suggests checking those for the weeks around 2026-09-20.
Why it matters
A tracking change can look exactly like a performance change. If you'd only had GA4 here, October would look like Meta's best month of the year. The useful way to read it: treat the platform or GA4 number as the hypothesis, and the model as the place you test it. When the model and the platform hold steady and one source jumps on its own, that's a measurement question before it's a budget question.
One practical tip: read the last few days before the data date with care. The newest days can still be filling in, and ask again after the next refresh if a gap shows up right at the end.
Try these follow-ups
- Show daily values for the three weeks before and after 2026-09-20.
- Is the same GA4 jump visible on Google Ads, or only on Meta?
- Split the gap by country. Does it start in one market first?
- When were the MMM results last refreshed, and which recent days should I treat as incomplete?
Related
- Sellforte MCP
- Model validation
- Data in Sellforte
- Sellforte dashboard
- Ask Sellforte MCP: Did yesterday look normal?
Try it yourself
Want to run this check on your own channels? Book a demo, or read how to connect Sellforte MCP to your LLM client. Keep an eye on this library, we add new use cases regularly.
