Marketing and Research Consulting for a Brave New World
Subscribe via RSS

I have always tried to impress upon clients that brand equity and brand health are different things. Brand equity is about the size of the brand while brand health is about growth.

If you think brand equity and brand health are the same thing, ask Meta/Facebook what happened to their cap value when they announced their daily user counts had dropped.

Measuring brand health. So, you really need to measure brand health…but what is it? Let’s start with what it’s not…it’s not Market share, penetration and measures of awareness. These are lagging indicators that measure size not direction.

What do lagging indicators of brand size have in common? They occur in calendar time…i.e. they align to weeks, months, quarters. That means any movement is likely to be reversible as it might just reflect the advertising and promotional give and take between you and your competitors that month. 

So we need something else…something more to measure brand health. There is a set of measures that are not in “calendar time” but in “event time” … which tend to be about repeat buying. These require longitudinal data (same consumers over time). These give us a better window into brand health and brand trajectory.  

To understand the power of measuring repeat rates, let’s start with something most of us are familiar with…how to predict if a new brand is likely to be successful in the long run. What do you focus on? The brand’s first repeat rate (i.e., what percent of triers ever buy it again) and the depth of repeat (i.e., how many additional purchases these repeaters make). Repeat purchases occur in event time…the first repeat purchase (if it occurs) might occur the same month, the next month, or in 6 months…it does not follow the calendar, it is a calculation of a consumer action whenever it might occur. First repeat and depth of repeat are not in calendar time but in event time. They require longitudinal data (same consumers over time) and they tell us about a new brand’s future fortunes.

How about existing brands? If you could only have one measure of brand health, the best measure would be the Markov repeat rate. It is readily available via receipt scanning, household panel, and frequent shopper data and should be used by more marketers. If your Markov repeat rate is increasing your share must grow.

How is the Markov repeat rate calculated? It is the percent of those buying your brand who buy it on their next purchase occasion. It is the conditional probability that a consumer will choose your brand given they bought it on the last occasion. If that makes it seem like it has mathematical properties that’s right, it does.

  1. It is mathematically predictive of future steady-state market share (not just something we observed that might be a coincidence), based on linear algebra calculations.
  2. If your repeat rate has changed, the whole distribution of how consumers prefer your brand must have changed! Literally, the second moment of the Beta distribution is the numerator of the Markov repeat rate which is calculated from the two parameters of the Beta distribution. (For any of you who want to get even more into the math, please message me at
  3. Markov repeat rates are one side of the brand switching coin. The other is the consumers who switch. In steady state, the number of consumers you lose to a given competitor must be the same as the number you win. Subscription businesses like telecom might think of this as “churn” and the balance between acquisition and customer loss is where growth lies. If you are gaining more than you are losing, your Markov repeat rate will increase and that is your predictive marker for sustainable gains in market share.
  4. When your repeat rate drops, everything starts going to hell. The purchase cycle for your brand gets longer. The value of a customer drops. Your penetration starts declining. Basically the brand will hemorrhage.
  5. To magnify the predictive power of the Markov repeat rate you even have the flexibility to calculate it off of non-consecutive purchases, for example, plan to look at the repeat rate from the first purchase of 2022 to the first purchase after June 1, 2022. If the non-consecutive repeat rate is higher than the consecutive purchase one, that is another powerful signal that the brand share is trending up rather than just bouncing.

Final point, if consumers are shifting their loyalties, you want to know WHY. Brand tracker data is the key. Create a driver model of attributes to predict purchase probabilities. Those repeat buyers who do not have commensurate perceptions are more likely to defect. Conversely, those repeat buyers of your competitors who have favorable views of your brand are Prime Prospects who are five times more likely to be converted over time (research here.)

Your brand guidance dashboard should go beyond share and penetration to contain measures of BRAND HEALTH. And your best measure of brand health is one you probably are not yet using…your brand’s Markov Repeat Rate.

Acknowledgement: I want to thank @JaredSchrieber founder of Numerator, for a stimulating one on one discussion that led to this ah-ha about calendar time vs. event time. I highly recommend his forthcoming book, “Breakout brands” whose principles are based on millions of shopper data points from Numerator data. A preview of his book’s ideas can be found here.

Tags: , , , , ,


2 Responses to “Which KPI is your best leading indicator of brand health?”

  1. Do you feel like this is somewhat limited to consumer packaged goods that naturally are bought with some frequency? What about products like automobiles, heavy appliances, or (dare I say) market research services?

    • joel

      Actually, the concept should be universal and reflects fundamental human characteristics of attachment and simplification. Probability of purchase is a latent characteristic in all cases but the manifestation of it in data is more obvious for CPG or QSR or gas pumps. however, I bet these probability ideas apply to autos and while I do not work in that category, I’d be shocked if the repeat rate for a brand of car is not considerably higher than its share and, if so, that proves the applicability of the model.