MMMs Can Tell You If CTV Worked. They Can’t Always Tell You Why
| |
By Dan Larkman, CEO, Keynes
|
Marketing mix models (MMMs) deserve the attention they are getting. Used well, they can pull brands out of a measurement culture that still gives too much credit to the click, especially in channels where clicking was never the point.
That is a healthy shift. A click can be accidental, low intent or simply easy to capture. Anyone who has opened Instagram has probably inadvertently tapped a Reel or ad. The platform still records the action. Leadership sees a number, and a weak signal eventually becomes a budget argument.
When properly considered, MMMs give marketers a broader view of how channels contribute to growth, where spend creates incremental impact and how different parts of the media mix interact. This is especially useful when so much marketing activity happens on mobile devices, while many conversions still happen elsewhere. A consumer may see the ad on a phone, search later from a laptop and convert on a desktop. Looking only at the last visible action misses the way demand is actually created.
The best MMM work also helps marketers understand marginal impact. Not just whether a channel worked in the aggregate but what happens when the next dollar is added. For CTV, that discipline is badly needed.
The problem is that many marketers expect MMMs to answer questions they were not built for.
One CTV number can hide a lot
CTV is often treated as a single channel in a model, but it does not behave like one in practice. A campaign might run across hundreds of networks, dozens or hundreds of audience segments, multiple devices, different inventory paths, varied frequency levels and different creative rotations. Two campaigns can both be labeled “CTV” and have very little in common operationally.
If an MMM says CTV was incremental, the result is encouraging but incomplete. The reverse is even more dangerous. If an MMM says CTV underperformed, the conclusion can sound cleaner than it is. A brand might read that as a channel failure when the issue was actually a planning failure, a supply issue, a poor audience match, an overexposed segment or a bad test design.
Imagine a brand running one campaign on Hulu and another on Peacock. One produces incremental reach and downstream lift. The other does not. A channel-level readout may collapse both into the same CTV bucket. The result might be technically useful, but it does not tell the marketer what to do next with their budget.
That is where MMMs should be treated as calibration tools rather than arbiters of judgment. A good calibration tool can tell you something is off or where the mix needs attention. It can help a brand understand which channels are contributing. But it can’t isolate every ingredient inside a complicated campaign.
Geo-testing adds another layer of complexity. If a brand excludes one market from a campaign, but a small amount of media still reaches people in that market because of imperfect location signals or cross-device matching, the test can be contaminated. When the expected lift is small, even minor leakage can change the read. A 2% lift that appears as 1.8% may look like a rounding difference on paper, but it can influence millions of dollars in planning decisions.
This is not an argument against MMMs. CTV marketers should be leaning into them. Independent incrementality work is a necessary counterweight to platform-reported attribution, especially in a non-clickable channel. It helps marketers escape the false comfort of last-click reporting and gives finance teams a more credible way to evaluate media investment.
The better argument is for using MMMs in the right sequence. The model should always be to sharpen the next question rather than end the conversation.
- Start with the model to understand the broader contribution and marginal impact.
- Use more granular diagnostics to figure out why the result happened.
- Look at network mix, audience composition, frequency, creative exposure, device distribution, geo delivery and the relationship between CTV and other demand signals.
Accountability without oversimplification
CTV does not need to be judged by the same metrics that made search and social easy to buy. It also should not be excused from accountability because it is harder to measure. The channel can create demand, influence other channels and drive business outcomes. Proving that requires a measurement approach that respects the complexity of how media is bought and consumed.
MMMs are useful because they reduce the industry’s dependence on the click. They become risky when they replace one oversimplification with another.
For CTV, the right use of MMM is: “What did we learn, what needs to be adjusted and where should the next dollar go?”
That is the kind of answer marketers can actually use.
Follow Keynes and AdExchanger on LinkedIn.
|