EpiPen Pricing Puts the Predictive Interviewing Model to the Test
Communication models work to simplify complex phenomena by focusing on the most relevant dimensions and useful particulars. Models help us see and understand relationships and all us to see patterns. They provide structure in an environment of uncertainty, and while they never perfectly reflect reality – because they’re not supposed to – they give us ways to logically process what may initially appear to be completely irrational experiences.
Models help us see and understand relationships and all us to see patterns.
We’ve analyzed hundreds of interviews using our own Predictive Interviewing Model (PIM) over the past five years, and found it to be an excellent guide to the typical behavioral pattern of a news interview. To build the PIM, we captured and analyzed 505 interviews aired on National Public Radio then focused on the types and sequence of questions used by reporters. The median length of the interviews we studied was four questions.
What happens to the model, though, when an interview blows past that median length?
Heather Bresch, CEO of Mylan Labs, the maker of the EpiPen, just provided a rare opportunity to test the question. She was interviewed on August 25 by Brian Sullivan of CNBC about the skyrocketing price of the EpiPen, a life-saving product used by millions of people.
Astonishingly, the interview ran for 17:48 without commercial interruption, during which Sullivan fired 34 separate questions at Bresch – a very long interview by any standard. On the whole, Bresch was pretty darn good. She wasn’t perfect in her responses, but there’s no question she was well-prepared. The telltale sign was her response to the first question, when she quickly threw the flag (a technique we teach in our Predictive Media training), “First and foremost…” then followed it with a strong headline message.
We used the PIM to look at Sullivan’s questions and the sequence in which they were asked. The 34 questions and statements Sullivan leveled at Bresch are transcribed in Table 1 and summarized in Table 2.
Table 1: CNBC’s Brian Sullivan Interview with Mylan Labs CEO Heather Bresch — Question Transcript
Table 2: CNBC’s Brian Sullivan Interview with Mylan Labs CEO Heather Bresch — Question-type Counts
By analyzing the content, we see questions asked just as the PIM predicts, but also two very intriguing and hard-to-predict patterns in Sullivan’s interview style:
- Just like the PIM predicts, an open-ended question-type started the interview.
- Just like the PIM predicts, probe question-types were the most common question-types used.
- Just like the PIM predicts, hypothetical question-types are used less often than the others, but appear after the fourth question has been asked.
- And finally, just like the PIM predicts, an offer served as the wrap-up moment for the interview. Bresch, by the way, did a great job of using that particular moment to finish strong by repeating her headline message.
What made this interview skew from the PIM’s normal sequence progression was Sullivan’s habit of loading up questions; he jams unjustified assumptions or controversy into setups before delivering the actual question. On five occasions, Sullivan used this technique, including his long preamble to the very first question of the interview. His favorite loading device? The proxy.
Proxy question-types are source generalization questions used by reporters through statements like, “I’ve spoken to many people who…” or “Many are upset over this… .” The technique has been described by academic researchers as a reliance on faceless sources to create an image of investigative reporting, even though none has actually taken place. This is important for interviewees to notice because it’s not an unusual trick. The question, however, is how is it best addressed?
Our advice is to focus on the premise of the final question and ignore all the rest. This is easy advice to give, but staying on message under the weight of loaded questions is much less easy to do — especially when, as Sullivan does, an interview isn’t even trying to disguise the attempt. Consider Sullivan’s third question, where he states “As somebody I talked to last night said…” to see about as pure a proxy as you’ll ever find.
Also unique in this interview was Sullivan’s pattern of statement-making. On six occasions, he made statements intended to cause Bresch to react. (Questions 3 and 25 are the best examples.) In question 25, by the way, it’s not difficult to know why Sullivan laid this line down. Bresch is the daughter of a sitting U.S. Senator, Joe Manchin of West Virginia. It’s a good bet that he was trying to bait Bresch into defending her father’s profession. In case you’re wondering, she didn’t take that bait.
This kind of statement-making isn’t common among reporters who are aiming to present themselves as objective. So it’s worth speculating as to why they may have appeared in this interview:
- In a long interview (nearly 18 minutes in this case), a reporter can throw a lot of questions, and include about as much commentary as he wants without facing time constraints.
- When watching the video, one can notice that Sullivan was wearing an earpiece but had no notes in hand. It’s a good bet that production staff was feeding him question — and background information he could then present as statements.
- Sullivan cited the New York Times on two occasions during the interview. Since he had a named source, he may have felt he could bottom-line those exchanges with statements rather than ask questions to seek the same information.
Whenever prepping for an interview, ask yourself, “What’s the worst question I could possibly get?”
One last observation: Whenever prepping for an interview, ask yourself, “What’s the worst question I could possibly get?” See questions 10 and 22 as my picks for Bresch. Like most all the others, she handled these really well.
The good news is that a number of the PIM’s attributes were affirmed, even by this very unusual interview. However, models are intended to simplify the world and this interview, because of its significant length, didn’t conform perfectly to the sequence the PIM predicts. As with any good science, we’ll use this to ask whether we should modify the model and look for more opportunities to test its boundaries.