What the data said
So about those academic papers. In September 2002, Richard Wittink from the Yale School of Management published a study, "Analysis
of ROI for Pharmaceutical Promotion" (ARPP). An update to another study done in 1999 by Scott Neslin of Dartmouth, ARPP examined
major promotional activities by the pharma industry, from 1994-2000, including the "new" DTC. (In the early 1990s, the FDA
had loosened regulations to allow DTC advertising.) Arguably the first DTC TV commercial was a 60 second ad for Rogaine, a
treatment for male pattern baldness, which first aired in 1994, marking the beginning of "DTC" as we know it now, and the
starting point for Wittink's study. Wittink examined a total of 392 brands, each with a minimum revenue of $25 million; 63
brands had more than $500 million in revenue. In this period, DTC advertising grew at an astonishing rate of 44 percent a
Despite the house-on-fire growth and enormous spend levels (quite a few brands were spending more than $100 million annually),
the ARPP study showed barely any positive ROI for DTC. To quote Wittink's paper: "For the 192 brands with intermediate revenues
($100-$500 million) DTC...return is close to zero." For larger brands (revenue above $500 million) ROI for DTC was just barely
positive—.02." So for every dollar spent on DTC for a big brand from 1994-2000, you got back a measly two cents. If you had
a medium size brand, you were basically firing up thousand dollar bills with your Bic cigarette lighter.
Granted, this was a long time ago. Other studies were done later, some of which showed an ROI for DTC of between 1.5-2.0 to
1. It should be noted that some of these other studies counted revenue over a three-year period post-campaign, which made
the numbers look a whole lot better.
Nevertheless I decided to be generous and use the higher estimates as the benchmark. That meant that for our digital efforts
to prove themselves "superior" to traditional DTC, we had to show better than a 2:1 ROI.
The path seemed clear, but it turned out the road was strewn with obstacles. First, no one was even measuring digital as a
channel back then. We had a chicken/egg problem. In order for digital spend to show up using the standard measurement models,
there needed to be much greater spend, but in order to justify greater spend, we needed to prove it could work. You see the
Just about this time, Paul Ivans, a young Cornell-educated chemist with a big dose of computer science and business experience,
tried to solve this problem. Over the next 10 years, Paul and his company, Evolution Road, would have arguably more influence
on the evolution of digital marketing measurement in pharma than any other single entity.
"The pharma marketing and sales model was broken a decade ago," recalled Ivans. "I saw that the digital wave was coming, and
there needed to be a better way for patients and physicians to get information, make choices and create better outcomes for
patients. In order to change the model in a significant way, I concluded that we needed better ways to measure ROI."
So Ivans set to work on an approach to measuring the effect of digital initiatives that would break the industry out of its
Catch 22 predicament. Along the way, he innovated new methodologies with the help of a variety of partners. First, around
2003, he pioneered a technique called PLA (patient longitudinal analysis), also sometimes referred to as PBM Matchback. Whatever
the name, the central idea was this: in cases where there was a pharma marketing campaign without database name capture, (i.e.,
most digital campaigns) these techniques could connect digital marketing activities to actual script lift, using de-identified
patient data so that activity could be matched without violating patient privacy.
"At the time, I thought this would be the best thing in the world," admitted Ivans. But, as it turned out, for a number of
technical reasons, "the analytics groups at major pharma companies had issues with the PLA approach." As he summarizes the
problem: "The Achilles heel is getting enough people's names and addresses, and that's historically been hard to do. Sample
sizes are the rate-limiting factor in PLA." However, Ivans notes that recently some players have been developing approaches
to address this fundamental issue.