Michael Kaminsky warns that MMM vendors overstate their ability to measure brand-awareness campaigns precisely. Brand-awareness data come in weekly surveys of roughly 1,000 people and contain heavy measurement error, making the spend-to-awareness link unreliable. He suggests treating awareness as a multiplier that amplifies all marketing effectiveness instead.
Chapters
Imagine you spent millions of dollars on a brand awareness campaign last quarter, and your MMM vendor is telling you that they can measure its precise impact. They are probably not telling you the whole truth. Measuring how your TV campaign or your big out-of-home campaign drove your brand awareness is typically not possible with an MMM. Everybody wants MMMs to be able to do the following: take your spend, measure how it affects brand awareness, and then measure how brand awareness later impacts sales. They want both links in the chain, but here's the problem. Brand awareness data is not like your sales data. It's not like the same predictors that you have in an MMM. Sales data you get every day with a lot of precision. Awareness data you're getting at best once a week with a sample of about 1,000 people in the US. So it's really noisy. It comes with measurement error and you're getting it with 1/7 at best the frequency that you're getting your sales data. So in a typical MMM, you're already struggling with the number of parameters that you have, but if you're using awareness data or any brand metric as a dependent variable, you just have way too little data in order to identify the effects of SPET. Here's what is possible using brand awareness as an input into the MMM itself. So there's two links in the chain, That first link, not possible. That second link, brand awareness's effect on sales, definitely is possible. And it's something we do with our clients today. We think of it like a multiplier. When your awareness goes up, everything gets more effective. You get more organic growth, your ads land better, and they're more effective. And that multiplicative effect is what we estimate in our MMM.
