THE TOP 10 INSIGHTS

Chicago, June 26-27, 2003

1. ROMI is the language that will catch management’s ear. Guy Powell, expanding on his new book "Return on Marketing Investment" (i.e., ROMI) said the first thought that usually comes to the CEO in BtoB is, rather than increase advertising, wouldn’t it be better to hire more salesmen? His point was, when marketing can prove it provides a higher return, it can get the money. He and every other speaker at the conference addressed the issue of how that can best be done.

2. Modeling is the key. KK Davey and Russ Booth of Grey’s modeling group, Insight Partners, described modeling as the key to optimizing media mix and monitoring its performance. It’s important because the proliferation of channels and publications makes large mass audiences harder to come by. But Reebok provided a case history showing it can still be done. Their sponsorship of the original Survivor, and their use of every conceivable tie-in, provided a lasting turnaround in the price of Reebok stock. It shows rigorous identification of everything driving sales, and its impact, is getting to be the price of entry into today’s communication arena. They have been particularly successful using the models they produce to provide a single page report card on the efficiency of a given mix of media: Insight Partner’s Performance Tracker.

3. Identifying the exact costs of marketing mix elements is critical at Coke. This was emphasized by Craig Stacey, their new Director of Marketing Science, as the basis of what most call modeling, but is referred to as Marketing Variance Analysis (MVA) at Coke. It depends on three factors: data (he mentioned vending machines are a problem), capabilities (size and abilities of analytical staff), and business issues (advertising only gets simulations of what will happen by markets, not by stores, because advertising is the same for all stores in a market). They use A.C.Nielsen’s SPECTRA (See # 10.5). The end result is always the simulation model. They already have many aggregate level regression models. They are working on attraction-based (MCI) models that will cover all competitors and split categories on a quarterly basis.

4. Who says you can’t measure PR’s impact? Miller is doing it. They have had data showing how many were reached by their PR for some time. They feel it’s more credible and more serious than advertising, and it leaves a longer impression. So they started modeling and found they could use common metrics: number of impressions, % favorable, etc. They even found they could handle different subjects being covered at the same time. This was a subject that came up repeatedly during the conference: The arithmetic of modeling can show how important things are when they happen one at a time. But when they all happen at the same time you need something more. (A recognition-based ad tracking survey to measure the impact of all commercials that start running at the same time, is an example that surprisingly comes to mind for this author.) For Miller, the key comparison is reach per dollar and they find ad stock models as useful for PR as for advertising, except the half-life for PR turns out to be significantly longer.

5. You should manage multiple brands with a portfolio approach. Todd Kirk, VP at MMA, one of the leading model building firms, cited Aaker & Joachimsthaler (2000) in challenging the classic P&G approach that encouraged competition between its own brands. Kirk pointed to the downward spiral that can result. The winning brand causes the losing brand to cut its price, which the winning brand finds it has to match, so the losing brand has to cut more, and the cycle has begun. A portfolio approach would suggest dropping one of the company’s brands. A portfolio approach can also use brands as a firewall to protect the major brand. When clear Crystal Pepsi was introduced a few years back Coke did not want to endanger its main brand with a questionable extension, so it came out with Clear Tab. Now, both Crystal Pepsi and Clear Tab are gone, but Coke was spared being associated with a failed brand extension. To be successful in portfolio management he advocated a more holistic approach to modeling, as reflected in his firm’s Competitive Interaction Analysis (CIA). One set of attraction-based models is used to estimate market shares for all competitors. Then category size is estimated separately. It is basically the approach originally detailed in Cooper & Nakanishi (1988). Estimating that many parameters can reduce degrees of freedom, and a model’s ability to produce reasonable estimates. Kirk advocated the Fader & Hardie (1996) approach, reducing all the detail on brand sales that comes out of a scanner system to a more manageable set of "SKU attributes" (brand, size, color, flavor, etc.) Many similar citations and documentations make this a paper that should be sought by anyone interested in learning more, or in checking out the validity of current modeling techniques. (Try: todd.kirk@mma.com)

6. Ad spending can effect the firm’s stock price. George Zinkhan, Professor, University of Georgia, reviewed 88 models and 16 studies that looked at this issue over the past 20 years. He found that announcements of new campaigns in the Wall Street Journal had a positive effect in 67 of the cases and a negative effect in only 21. Other speakers tended to support these findings by also citing higher stock prices as evidence that the investor community believes advertising works.

7. Gateway models combine forecasting and market research. Gateway’s Tim Stoughton said they found the two disciplines had to be pulled together through modeling to provide the information needed for making marketing decisions fast enough, and with a broad enough perspective on the possible outcomes. Jeff Brazell, CEO of The Modelers, described the recently completed models and simulation software supplied to Gateway management. It is updated quarterly to reflect current conditions. Users can change product characteristics and customer satisfaction, as well as marketing efforts to see the expected outcome.

8. Advanced analytics still hold unrealized promise. Rick Watrall from the Hudson River Group, another of the leading modeling firms, keynoted the conference with a review of the current situation. He prefers the phrase "Advanced Analytics," even though modeling is the cornerstone, because many other procedures are also proving valuable. The field has evolved from the days when a simple marketing model could take a whole weekend to run on a mainframe. Now, almost anything can be run on your desktop. Although this is the age of accountability, he cited management surveys showing 68% have difficulty measuring ROMI. At the same time, two thirds say marketing mix modeling is the most important development in marketing – yet only 19% are currently using any advanced modeling to measure marketing effectiveness. He quoted the CEOs of Kraft and Starbucks as recognizers of the importance of advanced analytics. Among the challenges he saw were talent shortages, getting the needed data, and how to peel apart the effects of events that all happen at the same time. He cited preliminary conversations his firm and ours have had to see if tracking surveys can’t provide some of the missing information. Big brands and big retailers are the ones he saw getting the biggest payback. I asked if the enterprise software offered by firms currently in the news because of proposed takeovers, like Oracle, PeopleSoft and J.D. Edwards, weren’t basically systems to make all kinds of company information more accessible? Shouldn’t their adoption make information for marketing models more accessible? He agreed it should, but didn’t know of any cases where it had, and felt the firms named were showing little interest in the subject.

9. Kraft moves to "real time" metrics for "mid course" corrections. Chris Ciccarello, reviewing the long history of measuring marketing ROI at Kraft, characterized it in the title of his talk as "Great, but usually late" – a phrase other speakers picked up on as descriptive of most modeling efforts. Kraft’s next generation models are focusing on intermediate measures designed to insure success, not just explain problems. They use leading indicators like how many stores are adopting a new program, and what the results are from the first couple of days after a change. Another shift is from looking at market averages to a broader focus on consumers and their many variations, because: "Markets don’t buy products, people do."

10. You only get a high return from ROMI the first time around? That was one of the reasons ROMI hasn’t caught on more widely, cited by Lorraine Scarpa of ImmediateFX. The first time you change things to improve ROMI you usually get impressive results. But in subsequent years you are hard pressed to get further improvements. She also cited the expense, one that small brands can not afford, in describing the work her firm did setting up automated ongoing marketing analyses for Bob Woodard, VP at Campbell Soup. The object was to conquer an "ad hoc Monster" that had developed when work on optimization and ROMI had to be done on a project by project basis.

11. Who shows the best response to each marketing effort? Nielsen showed an example where responsiveness, combined with product usage, identified one consumer group where marketing expenditures would be three times as effective as on another group. They advocated running consumer mix models to capitalize on such differences. They have Spectra Lifestyle and Life Stage demographics for stores, TV and cable channels that can be used to go the extra step and optimize the marketing mix for the optimal customer types.

These are only highlights of things that caught my interest. For more complete details check the IIR website at www.iirusa.com for the availability of papers.

Bruzzone Research Company · 2515 Santa Clara Avenue · Alameda, CA 94501- 4692 · (510) 523-5505 www.Bruzzone-Research.com

The firm to call when you’re serious about measuring the effects of advertising