At MMM Pilot, we believe you should know exactly how artificial intelligence is used in your analysis. This page explains where AI is involved in the platform, what information it accesses, and how we ensure the outputs you receive are reliable and trustworthy.
Our AI Provider
MMM Pilot uses Google Gemini, a large language model developed by Google, as its AI engine. Google Gemini is used in specific, well-defined steps of the analysis and data preparation workflow. No other AI or large language model providers are used anywhere in the platform.
For details on how Google handles data processed through its AI services, please refer to Google’s AI Services Data Processing Terms.
Where Google Gemini Is Used
Google Gemini plays a supporting role in several steps of the MMM Pilot workflow. It does not run the statistical model — that is handled by a deterministic mathematical engine. Instead, Google Gemini helps you work faster, understand results more clearly, and set up your analysis with less manual effort.
1. Model Selection
After the modeling engine produces a set of candidate models, Google Gemini evaluates their quality using objective statistical scores and identifies the single best-performing model. This removes the need for you to manually compare dozens of technical metrics.
Google Gemini decides: Which model candidate best balances statistical accuracy across multiple quality dimensions.
2. Executive Appraisal
Once the best model has been identified, Google Gemini reads its outputs and generates a structured executive report. This report covers model quality assessment, a breakdown of each channel’s contribution to your results, budget recommendations, and a seasonal outlook.
When a model’s statistical reliability falls below a defined threshold, Google Gemini automatically switches into a diagnostic mode — in this mode, it deliberately withholds specific budget recommendations and instead focuses on identifying data quality issues. This is a built-in safeguard to prevent overconfident guidance when the underlying data is insufficient.
Google Gemini decides: How to interpret and communicate your results in clear, accessible language.
3. Chart Descriptions
Google Gemini reads the underlying data behind each chart in your final report and generates a concise, plain-language explanation of what the chart is showing. These descriptions are embedded in your presentation-ready report slides.
Google Gemini decides: How to translate analytical visualizations into actionable narrative insights.
4. Data Field Classification
When you upload a marketing dataset, Google Gemini reads your column headers and suggests how each field should be classified in the MMM framework — for example, identifying which columns represent paid media spend, which represent your dependent revenue variable, and which represent date information.
These are suggestions only. You review and confirm each classification before any analysis begins.
5. Field Name Formatting
Google Gemini converts your original CSV column headers into short, standardized names compatible with the modeling engine. This is a utility step that runs automatically before a model run begins.
6. Economic Indicator Suggestions
Google Gemini suggests relevant economic indicators from the FRED (Federal Reserve Economic Data) database that may improve your model’s sensitivity to external market conditions — for example, consumer confidence indices or inflation metrics relevant to your industry and country.
These are suggestions only. You select which indicators to include in your model.
7. Google Trends Term Suggestions
Google Gemini suggests Google Trends search terms that can be used as external signals in your MMM model — covering category-level trends, competitor search volume, and your own brand’s search activity.
These are suggestions only. You review and decide which terms to include.
What Data Is Sent to Google Gemini
We are specific about what information is shared with Google Gemini and what is not.
What IS sent to Google Gemini
| Information | Why it is sent |
|---|---|
| Your project’s industry, country, and product category | To make Google Gemini suggestions contextually relevant to your market |
| Competitor brand names you enter | To generate relevant Google Trends term suggestions |
| Marketing dataset column headers (field names) | To classify and format fields for the modeling engine |
| Statistical model quality scores | To identify the best-performing model candidate |
| Aggregated channel spend figures and budget figures | To generate contextual budget recommendations in the executive appraisal |
| A free-text project context note (if you provide one) | To add relevant context to the executive appraisal |
| Chart visualization data | To generate plain-language chart descriptions |
What is NEVER sent to Google Gemini
| Information | Status |
|---|---|
| Personal information (names, email addresses, phone numbers) | Never sent |
| Your customers’ data | Never sent |
| Your login credentials or payment details | Never sent |
| Raw advertising platform data or platform login credentials | Never sent |
| Health or medical data | Never sent |
No personally identifiable information (PII) or health data is ever transmitted to Google Gemini.
How We Control Google Gemini Outputs
All Google Gemini outputs in MMM Pilot are subject to strict controls before they reach you.
- Structured output enforcement. All Google Gemini responses are required to conform to a pre-defined structure. Responses that do not match the expected format are automatically flagged or discarded.
- Validation against known values. When Google Gemini selects a model or suggests data fields, its choices are validated against a known set of valid options. If Google Gemini suggests a value that does not exist in the allowable set, the suggestion is automatically rejected.
- Data quality gating. The executive appraisal step is configured to recognize when statistical quality is insufficient. When this threshold is not met, Google Gemini enters diagnostic mode and explicitly avoids making specific investment recommendations.
- Advisory-only suggestions. Google Gemini’s field classification, economic indicator suggestions, and Google Trends term suggestions are all presented as proposals for your review. You retain full control over what is used in your analysis.
- Deterministic output settings. For utility tasks like field name generation, Google Gemini is configured to minimize variation and produce predictable, consistent results.
What Google Gemini Does Not Do
- Google Gemini does not run the statistical model. The Marketing Mix Model is computed by a deterministic mathematical engine. Google Gemini only interprets the model’s outputs.
- Google Gemini does not access your Google Analytics, Google Ads, or any advertising platform. Platform data connections use OAuth authentication directly between you and the platform — Google Gemini is not involved in that flow.
- Google Gemini does not make autonomous decisions. All budget recommendations and channel insights are presented for your review. You decide how to act on them.
- Google Gemini does not learn from your data. Your project data is not used to train or fine-tune any AI model.
- Google Gemini does not generate legally binding financial advice. Results are analytical guidance for marketing decision-making and should be validated with your own business judgment.
Google API Data Usage
MMM Pilot connects to Google Analytics 4 and Google Ads to import your marketing performance data. This connection is made directly between your Google account and MMM Pilot using OAuth authorisation — you explicitly grant access, and you can revoke it at any time from your Google Account Permissions page.
Data imported from Google Analytics 4 and Google Ads is used exclusively for the purpose of running your Marketing Mix Model. It is not shared with third parties, sold, or used for advertising purposes. Specifically, Google user data is never transferred to Google Gemini or any other AI model.
For more information, see our Privacy Policy.
Questions
If you have questions about how Google Gemini or other AI technologies are used in MMM Pilot, contact us at privacy@mmmpilot.com.
Last updated: March 2026