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
Marketing mix models don't stand still. New data comes in, a broken connector gets fixed, someone tweaks a modelling choice, and suddenly there's a fresh model run waiting to replace the numbers everyone is already using. Before you hit release, you want to know what changed and why. If you're a data scientist, a customer success manager or an in-house analyst juggling more than one environment, you know this moment: it's a go / no-go call, usually on a deadline, usually with a very big spreadsheet.
Compare the new development model run with production for 2026-07-01 to 2026-09-30: spend, incremental sales and ROI by media group and month. Which months and channels moved, and why?
Break that down by month for Cross-network, Paid Social, Direct Mail and TV.
Hook your AI assistant up to Sellforte MCP and it works through the comparison much like a careful analyst would:
Here's what one side looks like, by media group for Q3 2026 (illustrative demo data, USD):
| Media group | Spend | Incremental sales | ROI |
|---|---|---|---|
| Cross-network | 1 004 759 | 4 903 695 | 4.88 |
| Paid Social | 745 448 | 4 853 759 | 6.51 |
| Direct Mail | 502 857 | 2 106 266 | 4.19 |
| 425 705 | 621 827 | 1.46 | |
| TV | 420 431 | 556 924 | 1.32 |
| All 12 media groups | 3 587 891 | 14 759 230 | 4.11 |
The monthly follow-up is where things get interesting. In the demo data, Direct Mail spend drops from 160 000 in July to 14 286 in August, then jumps to 328 571 in September, with ROI going from 3.51 to 1.34 to 4.64. TV only shows spend in September. These are exactly the patterns you want to check on both sides before release. If the development run has them and production doesn't, it's usually a data delivery or a modelling change, and you'll want to know which one before anyone puts the numbers on a slide.
One more handy detail: Sellforte MCP tells the assistant which uplift basis each figure uses. ROI is based on backtraced uplift, which credits carryover to the period the money was spent, while the sales decomposition uses same-period uplift. Knowing that saves you from a false alarm when two views of the same window don't quite match.
A refreshed model that quietly shuffles ROI between channels can undo months of trust in one meeting. This check is really about the go / no-go decision: release the run, hold it, or release it with the explanation ready. And because the assistant reads the model's own figures rather than platform reports, the explanation is about the model, not about two reports that never agreed in the first place.
The same side-by-side read is also a quick way to confirm that a fix actually landed. Added a missing ad-platform feed or a new conversion source? Check the affected channel on both tenants and you'll see whether the development results pick it up, before you tell anyone it's sorted.
Two quick tips: name both tenants and one fixed time window in your prompt, and split very long or multi-country pulls by year or by country so each query stays manageable.
Curious how this looks on your own model? Book a demo. Already a customer? Here's how to connect Sellforte MCP to your AI assistant. And keep an eye on this library: a new use case lands here regularly.