Beyond the Goldfish Bowl: Why DIY MMM Rarely Makes a Splash

Building your own marketing mix model (MMM) can be an enlightening journey—consider it as a crash course in DIY analytics. There's no shortage of free tools and online curricula available to lower the barrier to entry. Curious? Dive right in! But buyer beware: you may have lottery-sized aspirations when your in-house marketing mix modeling project gets the green light, but in reality, you’ve just snagged a free goldfish at the county fair.

Sure, a goldfish seems easy —small, contained, and unassuming. How hard could it be? But soon enough, you find yourself shelling out for the bowl, the aquarium, or whatever "ecosystem" your new pet needs to survive. Before long, you’re deep into research about water dechlorination, filtration systems, and balancing the nitrogen cycle. It's not impossible, of course, but did you really anticipate siphoning fish poop on your weekends?

And after buying new equipment and all that maintenance, what’s the payoff? Not much. It's a low-ROI pet—there’s a fair amount of work for minimal impact. Sure, the goldfish started out free, but it required investments of time and money to maintain it in-house. Free goldfish rarely equate to high value.

The truth is that "free" goldfish – much like a DIY MMM solution – demand resources and upkeep while providing little benefit. Sure, it’s feasible to keep it alive, but is your MMM adding value? Is it influencing company forecasts? Is leadership relying on it?

From Fish Tanks to Financial Stakes: MMM Isn’t as Simple as It Seems

Marketing Mix Modeling (MMM) is a deceptively intricate discipline—far from the plug-and-play solutions some would have you believe. While the explosion of AI and open-source tools makes DIY modeling more tempting than ever, without seasoned expertise, your model can quickly go belly-up.

At Analytic Partners, we sometimes supplement or replace in-house MMM models. Across these, we've identified common, goldfish-like deficiencies:

• Oversimplified models restrict accuracy: Commercial factors interacting with marketing, such as pricing and macroeconomic trends, are incomplete or absent. Critical marketing channels and halo effects across channels are regularly overlooked, and modeling multiple performance indicators is rare. Want more depth on why a holistic and commercial approach? Analytic Partners has a report on that.
• No simulator and no dashboard reduce value: Without core business interfaces, adoption falters, self-service is limited, and the agility needed for same-day support evaporates. A common bottleneck? The single point of failure—the analyst who built the model. Simulators are a huge value realization engine. This is one reason why “business scenario planning” is the highest-scoring critical capability in Gartner’s latest Marketing Mix Modeling Solutions research.
• Low model update frequency limits timeliness: Refreshing models and results every six months is uncommon, with updates typically occurring only annually or even biennially. MMMs should fuel organizational learning, but developing the necessary behavior change with annual iterations is tough.
• Lack of technical customization prevents bespoke solutions: Hierarchical data structures and industry benchmarks are generally absent. Econometric details, such as ridge regression or time-varying coefficients, depend on the free tool used rather than the method most appropriate for the key business question.

The risks aren't trivial. MMM, done right, informs investments worth tens or hundreds of millions. Done wrong, and you’re basing mission-critical decisions on insights about as reliable as a carnival goldfish’s lifespan. (Hint: The median lifespan is short.)

DIY MMM: Your Goldfish Might Need a Full-Time Team

While DIY MMM might seem budget-friendly initially, labor costs quickly add up. Using data from Comptryx and the Total Remuneration Survey, Analytic Partners estimates that hiring and salary costs for an in-house MMM team in year one can easily soar to a staggering $629K! Even with high employee retention, these costs remain substantial, averaging $419K per year in subsequent years. These calculations are based on a small MMM program delivering roughly two model updates per year and addressing two KPIs for just one brand, that's a hefty investment.  If this were your goldfish, that's an awful lot of fish poop siphoning.

And despite having the equivalent of 3.5 full-time employees on your goldfish care team for year one (in this example), validated results typically don’t appear until nine months in. That's nine months of resources, investment, and effort before you know if your goldfish—sorry, your MMM—is even going to swim.

We estimate the 3-year salary costs for building and maintaining a simple pair of marketing mix models for a single brand to be $1.3 million USD.

The Real Question: Is DIY MMM Ever Worth It?

There's tremendous value in understanding how analytics enhance forward-looking decisions. MMM, specifically, relies heavily on causal and prescriptive machine learning—areas where many analytics teams lack deep expertise. But realistically, how much time and effort can you afford to invest in training a team before they deliver tangible business value? Many brands today face increasing volatility and uncertainty. In this environment, wouldn’t it make more sense to allocate resources to proven use cases and ensure immediate strategic impact? Want to dive deeper into the critical considerations around building versus buying MMM? We have a white paper for that too.

When it comes to MMM, you don’t just want your model to survive—you want it to deliver robust, reliable and non-obvious insights that fuel growth year after year. Ambitious? Absolutely. But remember, there’s no such thing as a truly free goldfish.

Ready to step beyond the goldfish bowl? Explore Commercial Analytics—where MMM isn’t just alive, it’s thriving.