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
You've got a draft deck or summary for management, maybe written by a colleague or by another AI tool, and it's full of confident numbers. Before you stand up and present it, you want to know which of those numbers actually match your marketing mix model. That's a job most marketing managers end up doing by hand, one dashboard tab at a time.
Check these statements against the MMM for Q3 2026 (2026-07-01 to 2026-09-30), all markets. For each one, say whether it's correct, show the actual figure and explain any difference: 1) Media drove about 25% of total sales in Q3. 2) Paid Social was our most efficient media group, with an ROI above 5. 3) TV took about 10% of media spend. 4) Cross-network ROI improved from last year. 5) Total media ROI was higher than in Q3 2025.
Why is media's incremental sales a slightly different number in the first check than in the ROI table?
The assistant makes two queries. For the share-of-sales claim, it splits Q3 total sales into base, promotions, weather and media. For the rest, it pulls spend, incremental sales and ROI by media group for Q3 2026, with Q3 2025 as the comparison. It states the window, the markets and the comparison period up front, so nobody has to guess what was checked.
Then it lays the result out as a claim, the figure, and a verdict:
| Claim | What the model says (illustrative demo data, USD) | Verdict |
|---|---|---|
| Media drove about 25% of total sales | Media: 14 467 870 of 35 914 642 total sales, 40.3% | Wrong |
| Paid Social was the most efficient media group, ROI above 5 | Paid Social ROI 6.51, the highest of all media groups (next: Paid Video 5.04) | Correct |
| TV took about 10% of media spend | TV spend 420 431 of 3 587 891, about 11.7% | Roughly correct |
| Cross-network ROI improved from last year | Down from 5.84 to 4.88 | Wrong |
| Total media ROI was higher than in Q3 2025 | Up from 3.51 to 4.11 (+17.05%) | Correct |
Two of five are wrong, and one is close enough to keep with a small edit. The Cross-network one is worth a second look before you fix it: its spend went up 28.43% year on year, but incremental sales only rose 7.34%, so the ROI fell. A deck that says Cross-network "improved" is probably looking at sales, not efficiency.
The follow-up catches something that trips up a lot of reports. The share-of-sales check uses same-period uplift, where media's sales are counted in the week they happen, so the pieces add up to total sales. The ROI table uses backtraced uplift, where the carryover from a week's spend is credited back to that week. That's the right basis for ROI, but it gives a slightly different total. In this run that's 14 467 870 versus 14 759 230 for the same quarter. Neither is wrong. They answer different questions, and the assistant says which one each figure comes from.
The fastest way to lose a room is a number someone can disprove on their laptop. Checking claim by claim, with the actual figure and the period next to each verdict, gives you a deck you can defend, and a list of fixes rather than a vague "some of this looks off".
It also helps to be explicit about scope. If your text doesn't say which brand, market or period it means, the assistant has to pick defaults, and a verdict can flip with the scope. So paste the claims with the period and markets written out, and ask for the dashboard view behind each number so you can confirm it yourself. One limit to know: the assistant can only check against Sellforte. If a figure came from another BI tool, it can explain likely reasons for a gap (period, geography, metric definition), but it can't open that tool.
Have a deck you'd like to check against your own model? Book a demo, or see how to connect Sellforte MCP to your LLM client. We add new use cases to this library regularly.