Agentic MMM in Action: The World’s First End-to-End Workflow for Ecommerce and Retail

Agentic MMM in Action: The World’s First End-to-End Workflow for Ecommerce and Retail webinar is now available on demand.

In this 45-minute live session, we showed how Sellforte AI connects campaign measurement, budget simulation, and media execution into one unified workflow without switching tools.

You'll see how incrementality measurement tells you which campaigns actually drove sales, how scenario planning lets you model budget decisions before committing, and how budget recommendations execute directly in Google or Meta Ads with guardrails you control.

Agentic MMM in Action: Watch the full webinar recording

Live AMA: Full Q&A from the Webinar

Meridian is a modeling library, whereas Sellforte provides an end-to-end workflow that connects measurement, optimization, and execution into an automated SaaS solution.

Unlike in-house MMM tools that focus mainly on reporting on quarterly/monthly basis, Sellforte operationalizes automated daily insights into concrete recommendations that can be activated in ad platforms.

Compared to in-house solutions, Sellforte reduces time-to-value, leverages built-in experiment-calibration, and removes the need for large dedicated data science teams while maintaining transparency and control. 

Read more: Meridian vs. Sellforte MMM SaaS: The Complete Comparison

2. Does the model factor in marginal ROAS across all channels before making a recommendation?

Yes. Recommendations are based on marginal incremental ROAS across all included channels. The optimizer continuously reallocates budget to balance marginal returns, ensuring investment flows to the highest-performing opportunities.

The user can create “planning groups” where the system only optimizes spend within a planning group, but not across different planning groups. This way the user can have different miROAS targets for paid search, paid shopping and paid social channels. 

Read more: Marginal Incremental ROAS (miROAS): What is it? And why does it matter to marketers?

3. Is the MMM model itself built by AI, or does AI just interpret the outputs?

The MMM model and optimization tools are built in the "traditional way". The AI is an additional layer in the technology stack, using the MMM results and the optimization tools, as illustrated in the image below:

Read more about Agentic MMM: The Rise of Agentic MMM (Marketing Mix Modeling): How AI Is Transforming Media Optimization

4. What's the risk of hallucinations? What verification layers ensure query correctness and prevent hallucinated outputs?

As mentioned in the previous questions, the AI is an additional layer that sits on top of models and optimization tools built in the "traditional way". That means that all outputs from the AI are traceable back to source data, dashboards and optimization tools within Sellforte. 

The AI has specific safeguards and instructions for each prompt type and analysis. As an example, it is forced to use Sellforte Optimizer tool to answer questions about budget optimization across channels. 

5. Do granular campaign-level recommendations account for seasonality? E.g. when there's Black Friday?

Seasonality is taken into account in campaign-level recommendations in cases where the platform is allowed to scale spend based on targeted cost or efficiency levels, such as Target ROAS.

We are currently working on including seasonality for campaigns that are using fixed spend limits, such as "daily spend".

6. Does the model support offline channels like TV and OOH?

Yes, Sellforte can measure all media, including digital channels, offline channels, and own media (such as email, SMS).

Sellforte can also measure the impact of all media to sales in all channels, such as ecommerce, physical stores and marketplaces.

7. Can the AI print model validation & accuracy metrics?

Model validation metrics, such as R² and MAPE are available in the Model Validation view (see screenshot below).

We have not yet developed a skill for the AI to access model validation metrics so that they can be requested from the chat, but this is a great idea, and we'll add it to the backlog!

8. How do you maintain model stability at a granular level with limited observations?

Sellforte's Next Gen modeling approach uses vastly more data, compared to traditional Marketing Mix Models. The core method is an advanced version of Bayesian Marketing Mix Model that is calibrated with experiments and other data that can be used in forming informative priors. This method is complemented with causal attribution, pre/post analysis of bidding parameters changes, and channel/campaign -specific additional data that reveal info about level of saturation.

Read more about Sellforte methodology:

• Calibrating Marketing Mix Models with Experiments and Attribution data
  
• Pre/post-analysis of bidding changes: Change Intelligence

• Causal Attribution
  

We are also very happy to share more: book a short discussion with Sellforte.

9. Are you using simple prompt engineering or orchestration to guide the performance analysis? How customizable is the analysis?

The system uses orchestration with multiple agents rather than simple prompting. Under the hood of Sellforte AI, there are different agents answering questions. As an example, Media Planner Agent is built specifically to answer future-looking optimization and forecasting questions using Sellforte Optimizer.

Author

Daria Alén is Senior Marketing Manager at Sellforte, where she builds educational programs, webinars, and events for ecommerce and DTC growth teams. She has over 10 years of marketing experience in B2B SaaS and Tech with specialization in go-to-market strategy and marketing analytics. Follow Daria on LinkedIn for more about marketing and growth.