Interactive Agency Organic Predicts Chrysler Sales With Media-Mix ROI Model

Automaker's Digital Agency Correctly Estimated 2008 Numbers Before Talk of Big Three Trouble Even Began

By Published on .

NEW YORK ( -- It sounds mind-boggling: In July, before the banks fell apart, before the bottom dropped out of the credit market, before talk that two of the Big Three autos could potentially end up bankrupt, Chrysler's digital agency, Organic, predicted Chrysler's total retail sales.

Unusual accuracy: Organic was off less than 1% for Jeep sales in fiscal 2008.
Unusual accuracy: Organic was off less than 1% for Jeep sales in fiscal 2008.
It estimated that Chrysler would sell 418,000 vehicles in the U.S. during the second half of 2008. In actuality, it sold 411,000. Organic was off less than 1% for Jeep sales in the 2008 fiscal year, and its calculations for the Chrysler and Dodge brands for fiscal year 2008 were off 3% to 6%. Generally, an automaker is satisfied, even happy, if it can predict sales within 10% to 15%.

How could Organic have done that? It started when Chrysler tasked the agency with building a predictive econometric model that could help it optimize its 2008 marketing spending. The agency pulled together a large set of inputs, including social-media and search data, and then went a step further: What if it turned the media-optimization model around? Instead of using retail sales to determine media allocation, it would use media spending to predict sales.

"In July, we knew what Jeep would come in at," said Chuck Sullivan, director of interactive at Chrysler. "Naturally, as we took the model through to the top of the house, we were met with skepticism. So we proved it, and we were able to prove it amazingly within a few thousand units."

Understanding future sales is critical for automakers. For example, Chrysler was the first domestic automaker to cut production last year, a major cost-saving move for the struggling company, Mr. Sullivan said.

Feeding the beast
The model was "very data-hungry," said Steve Kerho, VP-analytics, media and optimization at Organic. It looked at five years of historical ad spending by market by week across all the channels. It used lots of online data, from web clicks to behavioral, search and social-media data. It factored in macroeconomic information such as housing starts, the value of the dollar, gasoline prices, and inflation and interest-rate data. And it included auto-industry-specific information such as Chrysler's net contribution after incentives by model by month.

In addition to its forecasting capabilities, Organic offered up other lessons. For example, in one specific campaign it found the combination of TV and interactive generated the greatest possible return.

Marketing-effectiveness and media-modeling experts say these models are meant to be forecasting tools but predicting sales in such a tumultuous -- and unusual -- economy is very difficult.

"Marketers are realizing you can't drive your marketing budgets by looking through the rearview mirror," said Rex Briggs, CEO of Marketing Evolution and co-author of "What Sticks."

Unpredictable factors
Wes Brown, VP of consultant Iceology, said any prediction of annual vehicle sales volume last year would had to have been made midyear and based on the assumption that gas prices would stay high for the rest of the year.

That didn't happen, but the credit crunch that began in late summer, when Cerberus-controlled GMAC started demanding higher credit scores, and was exacerbated by the Wall Street debacle led to a steep decline in industrywide new-vehicle sales.

The original task handed to Organic was part of a mandate to be more efficient and get better return on investment -- an emphasis that re-emerges in every recession.

"We were challenged by senior management to make powerful, quick, important changes to everything we do, measure the success and, those that are successful, repeat," Mr. Sullivan said. "We knew if we had done everything how we had been doing it in the past that would not be a recipe for success."

The company is, in fact, using the model again in 2009 -- and senior management has asked Mr. Sullivan and his team to keep up the forecasts.

Most Popular
In this article: