> For the complete documentation index, see [llms.txt](https://docs.chord.co/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.chord.co/model-your-data/predictive-models/marketing-mix-modeling-mmm.md). # Marketing mix modeling (MMM) ### Introduction **Marketing Mix Modeling (MMM)** is a statistical approach used to measure the impact of various marketing activities, like digital ads, influencer campaigns, email, TV, or even promotions, on business outcomes, such as sales or conversions. MMM is a top-down approach to estimating the impact of marketing efforts. This approach is *top-down* because it aggregates summary data to estimate correlations between the time that marketing dollars were spent and corresponding outcomes (conversions, revenue, etc). This model estimates what impact a $1 of marketing spend in channel X had on time period Y, while taking into account all other factors (including other marketing efforts in that time period). We are left with a baseline demand plus the impact of marketing. In combination with client-side event attribution analytics, we can plug in our estimates and obtain marketing spend optimizations and recommendations for future spend. **Why It Matters for E-commerce Operators & Analysts** 1\. **Channel-Level ROI, Even Without Granular Attribution** MMM doesn’t rely on user-level tracking (like pixels or cookies). That makes it especially powerful in a **privacy-first world**, where attribution data is increasingly limited. It helps answer: {% hint style="info" %} “How much incremental revenue came from Meta ads vs. organic email?” {% endhint %} 2\. **Full-Funnel Visibility** MMM accounts for both **upper-funnel (brand awareness)** and **lower-funnel (conversions)** activities. It allows you to assign value to **non-clickable** channels like influencer marketing, PR, or even SMS campaigns that can’t be tracked with standard click-based attribution. 3\. **Scenario Planning & Budget Optimization** Operators can use MMM to run **“what-if” simulations**: {% hint style="info" %} “What would happen if we moved 15% of our TikTok budget to Meta?” This is valuable for quarterly planning, budget allocation, and pitching leadership on spend strategy. {% endhint %} While we don’t yet offer scenario-based simulations, the current model lays the foundation for future “what-if” capabilities by establishing channel-level response curves and measurable marginal returns. This positions Chord well to support simulation use cases in future iterations. 4\. **De-biasing Attribution Models** Most e-commerce attribution tools overvalue last-click or post-purchase touchpoints. MMM helps **balance that bias** by taking a step back and analyzing **macro-level trends** in your marketing and sales data. **Why MMM vs. Multi-Touch Attribution (MTA)** But why not use direct marketing attribution via multi-touch attribution (MTA) instead? The short answer is that we would love to. It is directly measurable and provides a detailed order-level interpretation. However, unfortunately, the world is messy and we often cannot often 100% attribute a sale directly to its source. The classic example of the need for MMM analysis is print or TV advertising. In these cases, while the marketing events may have led potential customers to a desired sales outcome, the digital breadcrumbs of the events may not be preserved. However, in reality, this type of unobservability also plagues most of our online marketing efforts. Opt-outs, VPNs, tech stack issues, and numerous other factors contribute to the potential for unobservability across all our marketing channels. In reality, we should make both top-down and bottom-up estimates to help drive better decisions. At Chord, we empower you by bringing your client-side eventing, order management system, and ad spend data all to one trusted place, where you can combine sources of truth to obtain the most accurate view of your marketing spend analysis. ## How it works At its most basic, an MMM model is simply correlating the spend in time t with sales events in that time period. We feed the spend data and the outcome data into the model, and estimate the weights. **Outputs** After estimating the weights for each marketing input above, we can generate the contribution for each input: * **Historical Contribution**: allotted channel contribution to KPI via estimated weight and actual spend in time period t. * By baseline and marketing channels * Over time, by baseline and marketing channels * Ranked over time by baseline and marketing channels * **Return on Investment** * **ROI vs Effectiveness:** Effectiveness measures the incremental outcome generated per impression. * **ROI vs Marginal ROI**: the additional return generated for every additional dollar spent. * **ROI by Channel with forecast intervals**: credible range of channel’s return over time. * **Cost per incremental KPI** for each channel. * **Response Curves for Each Channel:** Response curves illustrate the incremental revenue generated for each dollar spent. * **Spend Optimization** * Recommended change in spend per channel * Optimized incremental revenue across all channels * Optimize spend response curves ## In conclusion Over the past few years, the industry has seen a resurgence of MMM modeling techniques emerge that leverage advances in ML and AI, offering flexibility and power that would have been hard to imagine even a few years ago. With Bayesian simulation now available on GPU chip acceleration, along with the release of powerful open-source packages from both Facebook and Google that have seen widespread adoption, it is becoming increasingly common for MMM to be a trusted modeling tool in the tool chests of most advanced marketing teams. ## Eligibility Interested in adding this tool to your marketing tool chest today? To get started: * Connect your ad spend sources in Chord's "Data Sources." * For any spend sources not covered in "Data Sources," utilize the Ad Spend Amplification in the "Amplifications" portion of the platform, especially for offline or manual spend. Then, reach out to to join the beta! --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://docs.chord.co/model-your-data/predictive-models/marketing-mix-modeling-mmm.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.