Marketing-mix modeling has been around for two decades, but it's suddenly making its mark on marketing budgets as never before. The reasons, according to practitioners, include better analytical tools and growing familiarity among an increasing pool of marketers. It's also finding its way into CEOs' discussions with Wall Street and into industries beyond its traditional package-goods stronghold.
"Companies are both using [marketing-mix analytics] more and talking about it more," said Andrew Shore, analyst with Deutsche Bank Securities. Besides shifting money from ads to promotion for Clorox bleach and Kingsford charcoal, marketing-mix analysis also has helped Clorox save an estimated $65 million on wasteful trade promotions, Mr. Shore said.
Marketing-mix models analyze data from a variety of sources, including retail scanner data, company shipment data, pricing, media and promotion spending data and even public relations media-impression data, using correlation and regression analysis to sort how each element of the marketing mix affects sales for a particular brand.
The broad impact on spending isn't clear-cut. While marketing mix led Clorox to shift money from ads to promotions, P&G's embrace of marketing mix coincided with a 25% increase in U.S. media spending last year, according to TNS Media Intelligence/CMR.
no broad pattern
There's been no broad pattern to what the analysis recommends, said P&G Global Marketing Officer Jim Stengel. In some cases, the analysis indicated shifts between advertising and promotion; in others, shifts from one medium or even from one daypart to another.
P&G was never exactly a slouch in the number-crunching department. But its recent stepped-up use of marketing-mix analytics has been driven in large part by a marketing-mix expert acquired as part of the 2001 Clairol acquisition-Joe Auriemma, now P&G's director-consumer market knowledge.
"I see e-mails every week on better decisions we're making because of [marketing-mix modeling]," Mr. Stengel said. But while marketing mix does "a great job of refining what you know," he notes that the analysis still primarily looks at how each part of the mix works independently rather than at optimizing how all parts work best together.
Movement of package-goods executives into other industries has spread use of marketing-mix modeling, too, said Randolph Stone, president of Aegis Group's Marketing Management Analytics, the Wilton, Conn., firm that largely founded the approach and is still acknowledged by competitors as the biggest player in the field. Mr. Stone sees growing use of the approach among automotive, telecom, retail, entertainment and pharmaceutical marketers.
The models themselves have "gotten better by an order of magnitude over the past three or four years," said Leonard Lodish, professor of marketing at the University of Pennsylvania's Wharton School and a consultant to Symphony Technology's Information Resources Inc., another of the leading marketing-mix players.
The improvement has come from an estimation process that helps identify and minimize impact of bad data points, provide more detailed forecasts and sort the impact of good ad copy from that of bad ad copy.
One problem with marketing-mix models is that they traditionally have been done only once every year or two. And amalgamating the huge stockpiles of data from disparate sources can itself take up to 10 weeks, meaning the models are often based on data a quarter or two old by the time they're complete.
ImmediateFX launched last year with the idea of helping companies compile and analyze their own marketing data continuously so they can produce their own models more often and faster, said Gregg Ambach, VP-analytic services. The problem, he found, was that many companies still want a third party to referee an analysis that can become a political football.
Marketing executives want the analysis to prove their brands should spend more on advertising and less on trade promotion, Mr. Ambach said. But sales executives often warn cutting the trade spending will hurt sales. "The problem is the sales organization has the power to create a self-fulfilling prophecy," he said.
A bigger problem still is that while marketing-mix modeling has never been more popular, the data supporting it arguably have never been worse. That's true at least for marketers that sell products through Wal-Mart Stores, which stopped providing scanner data to syndicated research firms in 2001.
Marketers have Wal-Mart's own Retail Link data for their brands and categories, but generally lack data on competitive brands or what promotions occurred at what stores. And marketers generally can't share the data they do have with outside analysts, Mr. Stone said.
"It's made the marketing-mix analysis process much harder," Mr. Ambach said. While he believes Wal-Mart consumers respond to advertising largely the same way as consumers generally, the chain handles pricing and promotion much differently than other stores, making it hard to extrapolate promotion impact based on data from other stores.
"I've got projects going on right now with clients where I've built decent enough Wal-Mart historical consumption models, but they're just not holding up as a forecasting tool," Mr. Ambach said.
Now nearly 3 years old, historical Wal-Mart scanner data is getting stale. And Wal-Mart, already accounting for a third or more of many categories, keeps becoming a bigger factor in the sales impact that models must measure.
But Mr. Lodish said: "If you can manage 70% of your business well, the fact that you're not going to manage the other part as well shouldn't stop you."