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

I LOVE BASEBALL. I bought my first pack of baseball cards when I was 8 and went to my first Yankee game when I was 9. Yankee stadium was so beautiful and expansive…nothing like it growing up on crowded, dirty streets in Brooklyn.  Then there was Mickey Mantle. I remember seeing him hitting two home runs mid-way up in the upper deck during a doubleheader; not even Aaron Judge hits homers like Mantle!

But this blog is about more than me as a fan. It’s about the influence that baseball has had on my career and how it affects my ideas even today for transforming marketing analytics.

I’m not sure if my math brain drew me to baseball or baseball made me love math…but here is how I came to use stats and modeling for work daily.

In high school, I discovered Strat-O-Matic baseball, a remarkable game that purported to accurately replicate the statistics of every hitter and pitcher in the major leagues.  I had to see for myself!  When I was 16, I spend the summer creating an 8-team league (some constructed to be pitching heavy, some hitting heavy, some all power hitters), and played a 154-game schedule (each game took 30 minutes to play and record outcomes)…my mom was not allowed to touch the cards and sheets of calculations strewn in front of the TV or to vacuum the rug that summer! I keep stats on every hitter and pitcher in my league.  This was no small feat as there were no calculators back then, let alone computers or smart phones! Strat-O-Matic actually replicated player performance remarkably well, but there were some variations… as there are in baseball for real players from one year to the next.

Boom, there I was at the age of 16, living in the realm of statistics, probability theory, causality, and Monte Carlo simulation.

In 1972, a stats professor at the University of Chicago (where I got my MBA with concentrations in stats and economics) introduced me to an amazing book…Earnshaw Cook’s “Percentage Baseball”. This book was the first of its kind…using Bayesian math and decision trees to reach some very unconventional results, like bunting a runner to second by non-pitchers was a dumb thing to do except situationally in late innings.

Motivated by Cook’s work, later in 1972, I wrote a term paper for a stat class at U of C, “Predicting baseball outcomes”. I build a two-stage regression model…I predicted won-loss percentage as a function of run differential (runs scored-runs allowed).  Then I predicted runs scored with various batting statistics and I predicted runs allowed with pitching and fielding statistics. The model had amazingly good fit to the data. All of this before Bill James, and Sabermetrics.

Big insights came out of the model I built in 1972.

1-      On base percentage (OBP) was clearly a superior predictor vs. batting average. That finding was not well understood until 30 years later, as documented in Moneyball.

2-      For pitchers, I created a WHIP statistic (walks plus hits per 9 innings), figuring it was analogous to OBP. WHIP was highly correlated with Earned Run Average.

  • The old adage, “a walk’s as good as a hit” is basically true!

3-      I could simulate who was worth more, Mickey Mantle or Sandy Koufax at their best.  I did this by simulating the difference in runs scored or runs allowed by substituting the average player performance at their position for their performance on predictive metrics.  Then, I predicted changes in run differential which allowed me to predict a change in won-loss percentage.  Basically, I had created WAR (wins above replacement), perhaps the most important statistic today for player evaluation, 20+ years before Sabermetrics.  (BTW, Mantle was worth a few more games.)

By now, I had committed to statistics and analytics as a career path. At the age of 28 I was running analytics for Unilever in the US, and then for 25 years was the Chief Research Officer at The NPD Group. After leaving the ARF in 2010 as CRO, I have been in my own consulting business.

While at NPD in the 1990s, I began playing fantasy baseball.  The trick is understanding it is not real baseball and you had to game the scoring system. The head of research at Kimberly-Clark brought me into his league as co-captain and we won! I remember making Bip Roberts our number one draft pick and others laughed!  What they didn’t realize was that you needed a player at every position so the idea was not to draft the best player, but the one with the biggest gap vs. others at the position.

Since 2015, baseball geeks like me have a new candy store called StatCast.  Now, they can measure things like spin rates and movement on breaking pitches, batting average on different pitches (guides pitch selection), exit velocity (how hard hitters hit), spray charts that lead to fielding shifts and hundreds of other metrics. It was metrics like exit velo, barrel rate, and BABIP that led the Yankees to discover Luke Voit their star first baseman who was languishing in the Cardinals organization.

I regularly think about how to Moneyball marketing… it’s a great metaphor.  I have actually created new product forecasting systems that predict new product sales performance the way Nate Silver’s Pecota system predicts player performance.

Consider a Synectics excursion from the statistics of baseball to marketing.

Data driven Baseball Marketing and media analogy Comments
No longer solely relying on scouts’ judgment, but using Sabermetrics and Statcast statistics Good creative is not based on judgment, it is the creative that WORKS to produce high ROAS and brand lift. Kevin Youkilis (called the ‘Greek god of walks’ in Moneyball… but actually Jewish) wasn’t fast or classically built as a ball player, but his ability to get on base was tremendous. Creative is no longer about artistry or production value, it is about driving performance which is now highly measurable.
Batting average by pitch type. Understand what media tactics and publishers work best within the overall campaign. Starters have at least 3 pitches and they adjust pitch selection based on statistical analysis of batting averages…can vary by 100 points or more by pitch.  Why not media tactics?  Just like batting average by pitch is more granular than ERA, why not get more granular in media evaluation of publishers, creative, and tactics? That is what MTA does best.
Pitching splits (e.g. vs. left handed, right handed batters) Match creative to target and to marketing goal Especially as it gets into late innings, managers often mix and match the reliever to the hitter. The creative that works best for one consumer target is not necessarily best for another. The marketing content that works best for one stage in the customer journey does not work best for another. This has led to creative dynamic platforms that can improve effectiveness by matching in real time.
On base percentage (OBP) vs. Batting average (BA). Exit velocity on batted balls. Spin rates on breaking pitches. Choose a conversion metric that is broader (but highly predictive) vs. sales. These are like up-funnel metrics.  You don’t score runs with OBP or BA but you build runs. High exit velo is not a guarantee of lots of extra base hits, but it is highly predictive. What upstream (digital) behavior is most predictive for you of sales?  Search? Product page views? Find the right conversion metric!
Movement and speed metrics to detect player fatigue Carefully monitor ad frequency and pivot when saturation occurs Monitors now detect when a player is getting fatigued which means loss of performance and also increased possibility of injury. Marketing needs to measure individual users’ ad frequency exposure by channel in real time vs. incremental conversions.


Baseball lineups used to be the same every day. (Little known fact…when the Yankees first put numbers on players’ jerseys in 1929, it was their position in the lineup!) Marketing plans used to be the same from year to year.  Now baseball teams use predictive simulations and the Yankees, for example, have had a different lineup every day.  It is not random or just about injuries…they do massive simulations that predict performance for different lineups (based on who’s hot, matchups to the pitcher, etc.).  Very Bayesian, and very future focused.  When I think about brand trackers and marketing mix models, they feel much more like using the same lineup every day…backward looking, glacial, static…not predictive.

Fielding shifts and pitch selection can reduce batting averages by 50 points or more…very significant. Media effectiveness and ROI can be improved by 30% using constant testing, MTA and dynamic optimization.

Isn’t it time to transform the way that marketing works to be agile, data driven, and predictive…just like baseball did 15 years ago?

Sensitive measurement system for ad effects…tune to behavior not surveys, large scale vs. small scale, user level vs. aggregated modeling.

Read more ...

MTA adoption is growing, but restrictions on data sharing are creating barriers. Yet workarounds and a winning solution are possible.

Read more ...

The same ad unit cannot be optimal for both performance and brand marketing. How to decide what your brand needs now via an ad guidance system.

Read more ...

Stat testing done right is essential for fact-based marketing. Without it, we would have endless polarized debate…just like politics!

Read more ...