Select Page

A Conversation with Harry Bright … an Absolute Middlegame Champion





It’s been almost 20 years since I met one of Middlegame’s long-standing champions. We mentioned Harry Bright in an earlier blog and he has been an absolute inspiration from the time that Middlegame was just an idea. Although I met a number of interesting people at The Coca-Cola Company (TCCC) at my interview in 1996, I was particularly fascinated by my discussion with one individual. Four days later, Harry called me from Geneva, Switzerland asking me to work for him. His goal was to bring the same kind of marketing analytics I had been working on with scanner data to the emerging markets. I think I accepted the job before he even fully explained the offer and loved every day of the next five years I spent there.

A few weeks back, I had the chance to reconnect with Harry. Eventually, the discussion evolved back to those same topics that were part of that immediate connection when we first met in the Coke cafeteria. After Harry lit up the room with a couple of real gems, I asked if I could record the rest of our conversation since he was reiterating the prehistory of Middlegame. He graciously agreed that I could share parts of that here on the Middlegame blog. Here is an excerpt I thought was too good not to share:

(Todd) I start many client discussions with an immediate divide between scanner markets and audit markets. You were the one that taught me the true meaning of the audit results, so I get it. [But] a lot of the reasons that people claim these are so different really isn’t what they should focus on. I am admittedly coming at this from a modeling perspective. Was it the same when scanning revolutionized the retail measurement business?

(Harry) One of the perceived advantages of scanning data over audit data has always been the ability to identify marketing response impacts based on time-series models. Weekly scanner sales quickly demonstrated reliable correlation with distribution, pricing, and in-store activity. Further analysis was able to leverage the time-aligned variation to explain the impact of advertising and mass media. Monthly audit sales were subsequently determined to have little relationship with the in-store conditions on the day of the audit. The monthly or even bimonthly audits also provided far too few observations to meaningfully derive advertising impact with the time-series approaches.

(Todd) Clearly, this is the core reason why Middlegame exists. Our focus on emerging markets stems from this very problem. But we would have never understood this issue without what I learned from you. How did you get so engaged in solving this problem?

(Harry) In the nineties, Nielsen asked me to lead the development team, building an international database that integrated scanner and audit sources. This gave me the time to think about the root cause of the analytical differences between the two forms of collection. I came to the conclusion that we were thinking about the problem in reverse. The market research community was approaching the problem starting with the data and moving forward. The answer was addressing the goals of the analysis and working backwards to the data just like Art Nielsen Senior had done. Taking this different perspective led to the foundation of a novel idea. Eventually we would develop that into the ISO algorithm.

(Todd) The team at Coke became the real champion for this which is obviously where I met you. When I got there you were out of concept development and already implementing this for TCCC (The Coca-Cola Company) in East Africa, India, and Latin America.

(Harry) You know as well as I do how impatient TCCC can be getting good ideas into the market. I was presenting the first phases of the scanner and audit integration parts of the international database to our friends there since more than 80% of their global sales were generated in outlets without scanners. As the conversation turned toward the other ideas I had for the individual sample store in future phases, they wanted me to come to TCCC full-time and directly develop. Knowing that was where I really wanted to focus my efforts, I agreed and joined SMPRT (Strategic Marketing Presearch Research & Trends).

(Todd) I know the opportunity that Coke gave you to fully develop the preliminary ISO algorithm was really important to everything that followed.

(Harry) Since I needed the auditors to perform their job [in a] slightly different [way], we had to look at alternatives for data collection. One of those eventually became TRAC, the retail auditing company he co-founded in the UK. One of the initial keys was that Coke, Fanta, Sprite and the other products TCCC marketed had very high turns. Once auditors recorded the store purchases by date, we could estimate the sales during the week the auditor was in the store based on the purchase and repurchase patterns in that store by item (SKU). I called this the ISO algorithm and we suddenly started to have the same correlation between sales and distribution, and pricing, as well as in-store activity.

(Todd) I immediately saw the power of version one of ISO firsthand. I remember the feedback we got from teams in the field after the audit sample settled in the fourth or fifth month.

(Harry) We were immediately delivering all kinds of new insights. The only problem was that there was still no opportunity to leverage a time-series approach for sophisticated marketing response models. The estimated sales of the ISO algorithm and the in-store activity collected by the auditor were really only available for a single week during the month. The other three or four weeks had missing values. Somehow, we found Todd Kirk and he solved this problem.

(Todd) I remember calling my father after my interview and telling him that your description of this was the only thing that really interested me in going to The Coca-Cola Company.

(Harry) Your use of a cross-sectional approach that was able to deliver marketing response models without the time-series variation crossed that off my list. It also gave us a direct approach to solving portfolio questions. What happens to other packs if we only support the two litre? What happens to Coke and Fanta if we only support Sprite? As you further built-out your modeling approach, I developed the classification of in-store activity based on whether the stimulus impacted just an item, a brand of multiple items, or all items for a manufacturer. This greatly improved audit collection and I think it made the modeling better.

(Todd) It does make the modeling better and a lot of very positive lessons learned in those days that really ensure the face validity of the analytics we provide today.

(Harry) The approach was successfully demonstrated in a number of countries and TCCC thought so highly of the work, that they patented the algorithm that we developed before I moved on. Admittedly, the method I invented that could be directly incorporated into the TRAC production database is far superior.

(Todd) I tend to be very discouraged by all of the modeling projects I see that start with the models and immediately jump into findings. I really think that the models should augment the exploratory nature of simply investigating the data.

(Harry) I continue to believe that modeling, as well as any shopper response analysis, is best served by investigating the disaggregate data at the intersection between store, item, and week. The interactions are [best] observed the closer you get to this level in the data. You can always aggregate the results of your model to where a particular audience needs to make their decisions. The power of the Middlegame CIA® approach was originally thought to be overcoming and that the Multiplicative Competitive Interaction (MCI) model has significant advantages over the time-series approach. But the real power was allowing one model to be used for many simulations based on seeing opportunities in the data.

It was incredible to capture all of that history and even relive it once again by transcribing the conversation to this blog. We want to thank Harry for helping us support Middlegame clients the way that we do. As we mentioned before, Harry is one of the owner of TRAC. Here, he developed an entirely new and improved ISO algorithm which is directly imbedded into the each production database. The “on the fly” analytics that TRAC can deliver based on this is amazing. I encourage you to learn more about this at

Middlegame is the only ROMI consultancy of its kind that offers a holistic view of the implications of resource allocation and investment in the marketplace. Our approach to scenario-planning differs from other marketing analytics providers by addressing the anticipated outcome for every SKU (your portfolio and your competitors’) in every channel. Similar to the pieces in chess, each stakeholder can now evaluate the trade-offs of potential choices and collectively apply them to create win-win results.