Marketing mix modeling (MMM) is a powerful methodology for evaluating media effectiveness, particularly for brand media. Below, I share my personal experience collaborating with agency partners to build MMMs and deliver results. Rather than detailing the end-to-end process, I’ll focus on key learnings across three critical areas: data collection, communication, and reporting.

Data Collection

Data collection is often underestimated. It’s not just about exporting everything from the data warehouse. First, you must help the agency deeply understand your business. Every business is unique, and an MMM that doesn’t reflect your specific context is ineffective. Second, not all data points, even those relevant to the business, should be fed into the model. MMM focuses on marketing effectiveness, not every factor influencing the business. Finally, data collection requires both on-site data (e.g., sessions, revenue) and off-site data (e.g., spend, impressions), which often don’t align perfectly. For example, marketers may use different UTM campaign names than those tracked in paid traffic. Joining on-site and off-site data may also require deduplication efforts to ensure accuracy.

Communication

Unlike in-house modeling teams, which are naturally aligned with business and engineering updates, agencies rely on you to proactively share critical information. This includes business changes, engineering updates, timelines, challenges, and limitations. Transparency is essential. For instance, early discussions should clarify whether the MMM should be session-based, which aligns with media strategies, or revenue-based, which ties directly to business outcomes. Clear communication ensures both parties are aligned on objectives and expectations.

Reporting

Receiving the MMM results feels like a milestone, but it’s just the beginning. Validation is crucial, but the real challenge lies in presenting findings to stakeholders. How do you convince the performance marketing team that SEM’s ROAS is 80% lower than last-touch attribution reports without getting your face rearranged? How do you explain that some performance channels are saturated and budget should shift to brand media? Similarly, when presenting to C-level executives, who may lack marketing expertise, how do you demonstrate that a session-based model reliably informs revenue outcomes? How do you instill confidence in budget allocation decisions when the MMM itself is a complex, often opaque model? These challenges determine whether MMM results are actionable. If not addressed properly, the model risks becoming just another set of PowerPoint slides.