Most measurement tools tell you what already happened. The harder question is what to do with that on Monday morning.
In this 45-minute session, Sellforte CEO Juha Nuutinen and Senior Marketing Manager Daria Alén walk through a single workflow that connects three things most teams keep separate: measuring the campaign changes your team already made, planning the next quarter's budget against a causal model, and pushing the resulting changes back into Google Ads.
Juha runs it live against a realistic ecommerce scenario.
The character is Alex, a Performance Marketing Director at an $80M-spend ecommerce brand. Her team made a dozen tROAS and budget changes last week. It's Monday. She wants to know what worked, where Q2 budget should go, and she'd like to spend less than two hours getting an answer.
What you'll see:
- Change Intelligence. Pre/post analysis of a real tROAS change, with miROAS attached, and a recommendation to scale, hold, or roll back.
- Dynamic Scenarios. A 15% Q2 budget increase, allocated channel by channel against a causal model, in 60 seconds. Then the same exercise for a 20% cut, side by side.
- Sellforte Activate. The recommended change pushed to Google Ads in one click, logged automatically as the next measurement event.
Built for Heads of Performance Marketing, Marketing Analytics, and Growth at ecommerce and retail brands spending $50M+ on paid media.
Watch also:
[00:00:04] Daria: Hello everyone, welcome to the Sellforte session on Agentic MMM in action. Let's give it a couple of more minutes for people to join.
[00:00:24] This is our second session of the series. Today, we're talking about workflows for performance marketers and analysts, showing how to close the gap between finding an insight and updating a campaign by unifying measurement and execution.
[00:01:13] Could you please drop your location in the chat? We have people from the US, Brazil, India, London, Spain, and more.
[00:02:11] Quick housekeeping: the session is recorded and slides will be shared. Joining me today is Juha Nuutinen, CEO and co-founder of Sellforte.
[00:02:38] Juha: Hello, my name is Juha. My role is in product development and sales. I’m a bit nervous because presenting new features live is always exciting!
[00:03:44] Daria: I'm Daria from the marketing team. We'll keep this to 45 minutes and end with a Q&A. Sellforte is an enterprise-grade continuous MMM platform turning measurement into daily operational recommendations.
[00:05:17] Today we'll cover what Agentic MMM is and do a live demo following "Alex," a performance marketing director.
[00:06:07] In 2026, the marketer's role has shifted from pulling reports to making fast, high-quality decisions. CFOs now ask what we are doing tomorrow to drive incremental sales.
[00:07:37] Agentic MMM combines four pillars: causal MMM for ROI, incrementality experiments, optimization engines, and AI agents for daily actions.
[00:08:25] Roleplay Demo: Imagine Alex, managing $1.3M in monthly spend, preparing for a Monday 10:00 AM stand-up.
[00:09:14] Juha: I'll ask the AI: "How did last week go?" The agent interprets the question and pulls data from the historical MMM analysis agent, looking at spend and incremental sales per channel.
[00:10:36] Daria: Can you add incremental ROAS (iROAS) to the table?
[00:10:49] Juha: It adds that KPI. Now let's compare week 12 to week 11. The AI pulls the data and shows spend was 39% lower, partly because Direct Mail went to zero.
[00:13:36] Let's compare channel performance specifically. PMax declined slightly, Meta performance declined by 30%, but Meta Advantage+ increased.
[00:15:29] Instead of manual analysis, I'll ask the AI to summarize the main insights and highest impact changes. This builds the narrative for the leadership meeting.
[00:17:14] Now, let's look at recommendations. Currently, campaign-level suggestions are in the Performance tools. We have "high priority cards" recommending budget changes based on marginal incremental ROAS (miROAS).
[00:18:40] For example, a Meta sales adset has an miROAS of 13. The tool finds the optimal budget to bring that down to your target average (e.g., 3.0), maximizing total performance.
[00:20:47] Daria: What if I don't have time to update these manually?
[00:20:53] Juha: We’ve built two-way integrations for Meta and Google. You can implement the change directly in the ad platforms by applying the recommendation in Sellforte. We call this "Sellforte Activate."
[00:23:26] We also have "Change Intelligence." Every change made is tracked as an experiment. For instance, if you changed a PMax target ROAS and the campaign stopped spending, the tool calculates the miROAS of that change to see if it was beneficial.
[00:27:38] We recommend this workflow: test campaign-level changes and use the results to improve the overall model.
[00:30:04] Daria: How should I spend my budget for the next four weeks?
[00:30:16] Juha: We have pre-calculated template scenarios. The tool immediately suggests channel reallocations. For this fictional brand, it found a $200,000 incremental sales opportunity within the current budget.
[00:32:18] Daria: What if the CFO asks for a 20% spend increase?
[00:32:38] Juha: The AI calculates it. It suggests putting that extra money into Meta Ads, Attraction, and Pinterest rather than Google Ads.
[00:34:07] Daria: And if we need to reduce spend by 20% due to supply chain issues?
[00:34:56] Juha: The AI creates that scenario too. You can then compare all three scenarios (Current, +20%, -20%) in a table to show the CFO the marginal payback for each investment level.
[00:40:49] Daria: We've unified execution silos! If you want to see this with your own data, book a 30-minute deep dive via the QR code.
[00:42:31] Q&A Section: Is there a risk of hallucination? Juha: We focus on deterministic tools. The AI pulls real data from the agents; it doesn't "make up" numbers.
[00:43:31] Does the model consider marginal ROAS across all channels? Juha: Yes, and you can customize "Planning Groups" (e.g., only optimize within Paid Search or within a specific country).
[00:46:16] Is the model built by AI? Juha: No, it's a robust Bayesian model. You can't just give raw data to AI; it needs the underlying math to understand incrementality and saturation curves.
[00:47:45] Does it account for seasonality? Juha: Yes, we forecast demand, available search inventory, and even your specific promotions to adjust the recommendations.
[00:50:15] Thank you everyone for joining. See you at the next event!