top of page

How to Build a Business Case for Marketing Mix Modeling: The Questions Only MMM Can Answer

Writer: Linda Orr
Linda Orr
5 days ago
8 min read

Every marketing mix model I've built has a moment I wait for. The response curve for a channel climbs, bends and goes flat. Somewhere on that curve is the dollar where spending more stops producing more. Almost every company I work with is spending past that point in at least one channel, usually the one its dashboards make look best.


That moment is the business case for MMM. Your CFO is not really asking whether marketing works. They are asking where the next dollar should go, where the last dollar was wasted, and what happens to revenue if the budget gets cut. Channel reports cannot answer those questions. A well-built model can.


Cover graphic: Where does each channel stop paying off? A response curve showing the point of diminishing returns

1. What question is your CFO really asking?

Most marketing budget conversations get stuck on "what's our ROI." The question that changes decisions is a different one: if we had one more dollar, or one less, which channel should it come from? That is a question about the margin, about the next dollar. It can't be answered from averages, and every report your ad platforms send you is an average.


When I frame an MMM project for a finance team, I start with the decisions on the table. Should we keep TV next year? Can we cut paid social by a third without losing sales? Is direct mail worth the printing bill? Would more streaming help, or are we already saturated? If the model won't change one of those decisions, it isn't worth building.


2. Why can't you just compare ROAS by channel?

Because channel ROAS answers a different question, and it answers it badly. There are four problems I see in nearly every account I audit.


The first is that platforms grade their own homework. Google, Meta, Amazon and every streaming platform count the sales they touched, using their own rules. Add up what each platform claims and the total is often well above what the company actually sold. The same customer saw a Meta ad, searched your brand on Google, and clicked an email, and all three take full credit.


The second is that some channels harvest demand that others create. Branded search almost always looks like your best channel. It sits at the very end of the journey, catching people who already decided to buy because of a billboard, a podcast, a referral or a TV spot. Cut the channels that created the demand and branded search shrinks with them, a month or two later, when nobody connects the two.


The third is timing. A billboard or a TV flight this month shows up in sales over the following weeks, sometimes months. Attribution windows of seven or thirty days miss most of that, so the channels with the longest effects look the weakest.


The fourth is the most important. ROAS is an average across all the money you spent, and the average hides the margin. I worked with a connected TV campaign where the platform reported a cost per action of $1.61. A holdout test told a different story: the households that never saw an ad acted at a higher rate than the ones that did. The average looked wonderful, and the media was not causing the sales.


3. Where does a channel stop paying off?

Every channel has a point of diminishing returns, and I have found it in almost every model I've built. The first dollars in a channel reach the most responsive people. As you spend more, you reach people who are less likely to buy, you show the same people the ad again, and you bid against yourself in the auction. The curve bends.


Here is a simple illustration of why that matters. The numbers are made up to show the arithmetic. Picture a channel where the first $20,000 a month returns $4 in sales for every dollar, the next $20,000 returns $2, and the next $20,000 returns 80 cents. Across all $60,000, the channel's ROAS is a respectable 2.3. Every report will call it a winner. The last $20,000 is losing money, and you would never see it in an average.


Spend also carries over. A TV spot, a billboard or a streaming campaign keeps working for days or weeks after it runs, fading as people forget it. Statisticians call that carryover adstock, and a good model estimates how quickly each channel's effect decays. I have built it into models for a health plan and a sports camp company, among others. More recently I looked at the lag for a personal injury law firm advertising on Spanish-language TV, measuring how long after the spots aired the calls started coming in. Ignore carryover and the channels with the longest effects look like the weakest ones, which is exactly backward.


A marketing mix model estimates those curves for each channel from your own history, including the lag, the carryover from one week to the next, and the point where the channel saturates. That is how you find out where to stop.


4. What is an efficiency index, and how do you use it?

Once you have the curves, the most useful thing I give a leadership team is an efficiency index. For each channel, I estimate how much incremental revenue the next dollar would produce at your current spend, then index it so the portfolio average equals 100. A channel at 140 deserves the next dollar. A channel at 60 is past its point, and money there would work harder somewhere else.


The plan follows from the index. You move budget from the channels below 100 to the ones above it until the marginal returns start to even out. That is a very different conversation from "this channel has the best ROAS." When I built the first marketing mix model at a large tech marketplace, covering an $88 million media portfolio, that kind of reallocation supported a 25 percent reduction in media spending while sales continued to grow.


5. Should you spend more on streaming, billboards, direct mail or catalogs?

You can't answer that from attribution, because attribution barely sees those channels. Nobody clicks a billboard. A catalog lands on a kitchen counter and the order comes in by phone or through a branded search a week later. Streaming TV gets watched on one screen and acted on from another. These are often the channels doing the most work, and they're the ones companies cut first because nothing in the dashboard defends them.


MMM measures them on the same terms as everything else, because it works from spend and sales over time rather than from clicks. It also catches interactions. Channels often work better together than apart, and the model can show when radio makes TV more effective, or when direct mail lifts search. When I validated a health plan's marketing mix model, adding the economic conditions that drive enrollment moved out-of-home from the seventh most effective channel to the first. The original model's math was fine. It was missing what actually drives the business.


6. What has to go into the model before any math?

A hypothesis. Real statistics starts with thinking about what should be driving your sales, then building the variables to test it. Before I model anything, I want to know what else moves demand in your category: seasonality, weather, pricing and promotions, distribution and shelf position, competitor launches and spending, PR, the economy, and one-time shocks like COVID. Then I test whether the data is clean and representative enough to support those questions.


I've watched what happens when that step is skipped. At one company where I led measurement, the business later paid an outside firm about a million dollars for an MMM. Where the firm used the same variables I had, its results matched my analysis, which cost a fraction of that. Where the results differed, the outside model had left out weather, COVID, competitors, PR and TV. A model can be made to say almost anything through the variables you include, how you cut the data, and whether you assume straight-line or curved responses. Garbage in really does mean garbage out.


7. Why won't MMM software do this for you?

When I was in my PhD program, a young statistics professor got teased by older faculty for what they called his cookbook approach: put the data into SPSS, run the procedure, report what comes out. I liked him, and I think about that critique every time I see MMM software pitched as a solution. The software models whatever data it is fed. It doesn't ask whether a competitor's promotion, a rainy month or a lost retail account explains the dip. It can give quick answers and help shift budget a little within the walled gardens of the ad platforms, but it can't build a plan.


Open-source tools like Google's Meridian and Meta's Robyn are powerful in the right hands, and I compare them in Meridian vs. Robyn vs. an MMM expert. The right hands matter. The person building your model should have planned and forecast marketing budgets at the CMO level, or worked closely with people who have, so the strategy is built into the model. A pure statistician can produce a technically sound model that answers the wrong question. A Harvard Business Review Analytic Services survey sponsored by Google found the top barriers to getting value from MMM were slow processes (46 percent), lack of expertise (45 percent) and siloed teams (41 percent). Software doesn't fix any of the three.


8. What does MMM cost, and how do you show payback?

My MMM builds are custom quoted. The smallest run about $8,000, multi-channel models for companies spending millions typically run $30,000 to $40,000, and a very complex project with regional and national models and messy data can reach about $75,000. Compare that with what's at stake. Gartner's 2026 CMO Spend Survey found paid media now takes 31.4 percent of marketing budgets, a five-year high.


For payback, use a conservative example and label it as one. A company spending $2 million a year on media that finds just 5 percent of that spend sitting past its point of diminishing returns can redeploy $100,000 a year. A $30,000 model pays for itself in the first reallocation. In my experience the finding is usually larger, but the small number is the one to put in front of a CFO.


9. What should the one-page business case say?

State the budget decisions the model will inform and what each one is worth. Give the cost and the conservative payback example. Describe the data you have (roughly two years of weekly spend and sales is the place to start) and any gaps you'll close first. Name the outside factors the model will account for. Name who will build it and who will interpret it. Explain how you'll validate the results, ideally with a holdout or geographic test on the first recommendation. And plan to rerun the model as the business changes, because the curves move when your market does. Timing depends on scope and on how clean the data is, so say that plainly instead of promising a date.


If you want to know where each of your channels stops paying off, see how I approach marketing mix modeling. If your data isn't ready, start with a GA4 and measurement audit. Or book a call and bring your channel list.


Sources: Gartner 2026 CMO Spend Survey, May 2026; Harvard Business Review Analytic Services survey sponsored by Google, as reported by EMARKETER; Orr Consulting client work (anonymized), including a connected TV incrementality test and a health plan MMM validation.



Featured In Leading Media & Journals

Quoted in national media, published in peer-reviewed journals, and the author of five business books.

Forbes logo, linking to the Forbes article quoting Dr. Linda Orr on solo marketers
Entrepreneur logo, linking to Dr. Linda Orr’s articles on Entrepreneur.com
Yahoo logo
The Daily Upside logo, linking to the Daily Upside article quoting Dr. Linda Orr on AI marketing and brand risk
business.com logo, linking to the business.com article featuring Dr. Linda Orr on offline marketing tactics
Glossy logo, linking to the Glossy Fashion Briefing article quoting Dr. Linda Orr
Journal of the Academy of Marketing Science logo, linking to Dr. Linda Orr’s article in the journal
Journal of Business Research logo, linking to Dr. Linda Orr’s article in the journal
Psychology & Marketing logo, linking to Dr. Linda Orr’s article in the journal
Journal of Business Ethics logo, linking to Dr. Linda Orr’s article in the journal
Marketing Management Journal logo, linking to Dr. Linda Orr’s article A Re-Inquiry of Hofstede’s Cultural Dimensions

Why Orr Consulting

You work directly with a PhD-level strategist.

When you work with Orr Consulting, you work directly with me. I set the strategy, decide where the budget goes, and stay accountable for the results.

 

I also lead the people you already have: your internal team, agencies, developers, analysts, creatives, and specialists. You get senior direction without replacing the team you have built.

 

When you need more hands, my team does the implementation: tracking and CRM setup, web pages, email marketing, Meta and paid social, SEO and AEO, and video and image editing. I manage Google Ads myself and review everything the team delivers, so you have one point of accountability.

 

My background combines 25+ years of marketing leadership, two decades as a marketing professor, and a PhD in Marketing with depth in statistics and psychology. That work includes building Upwork's first marketing mix model across an $88M annual media portfolio.

 

Dr. Linda Orr, PhD, Founder

Dr. Linda Orr, founder of Orr Consulting

SELECTED CLIENTS AND COMPANIES

Selected Orr Consulting clients: PepsiCo, Upwork, McKesson Foster and Thrive, Midmark, FedEx Custom Critical, Arm and Hammer, TheraBreath, Mavrik Dental Systems, medmix, Elos, GFOA, Indeed, Magenta Inc, PreVu Dental, SMRTFT, Sparks, The Critical Thinking Institute, Tahoe Modern, Mayberry Law, Ollie Pediatrics, Therapevo, Mind On Photography, Artemis, Disc It
Selected Orr Consulting clients: PepsiCo, Upwork, McKesson Foster and Thrive, Midmark, FedEx Custom Critical, Arm and Hammer, TheraBreath, Mavrik Dental Systems, medmix, Elos, GFOA, Indeed, Magenta Inc, PreVu Dental, SMRTFT, Sparks, The Critical Thinking Institute, Tahoe Modern, Mayberry Law, Ollie Pediatrics, Therapevo, Mind On Photography, Artemis, Disc It (continued)

What Clients Say

Most engagements are under NDA. Here is what clients say, plus our reviews on Google.

★★★★★

“Dr. Orr's exceptional leadership and advanced knowledge in sales and marketing were invaluable. She provided practical tools our team could use immediately to perform with confidence.”

B. MacDonald

B2B Sales Enablement · Sales Training

★★★★★

“Linda was a pleasure to work with and extremely thorough. She led a market research project for our 3D dental imaging launch, and the insights directly shaped our positioning and go-to-market strategy.”

Midmark Dental

Healthcare · Market Research · Launch Strategy

★★★★★

“Working with Linda was an absolute pleasure, and the results she delivered were beyond our expectations! ... She interviewed over 20 key players in the yachting industry, ranging from yacht charter company owners to marketing managers, extracting valuable insights that we wouldn't have been able to gather on our own... The final report she presented was top-notch. It was thorough, well-structured, and packed with actionable recommendations that demonstrated her deep understanding of the industry.”

Founder, Yacht Charter Marketing Agency

Market Research · Market Entry Strategy

Prefer to Schedule Directly?

Pick a time that works for you, or start with our free guide to getting found in AI search.

Contact

Thanks for submitting!

bottom of page