Health knowledge made personal
Join this community!
› Share page:
Search posts:

DTC Pharma-Twitter Experiment: Achieving a 14% Response Rate

Posted Jul 30 2009 12:19am

The first pharmaceutical branded tweet was sent out on June 18, 2009 on behalf of Novo Nordisk. It was sent by racecar driver and Type I diabetic, Charlie Kimbell, using his @RaceWithInsulin Twitter profile.


Labeled as “sleazy Twitter spam” by some and “a great start” by others, the resulting coverage revealed just how little is known about how Twitter can be used as part of a brand manager’s direct-to-consumer plan.

Recently, Krū Research conducted a marketing experiment to determine the optimal way a pharma brand can connect with health consumers on Twitter. We will leave the more philosophical questions to others. This article is not designed to address the ethical issues of using social media for sales, or even the often sited claim that a sales goal in itself is counter to building authentic relationships.

Our experiment is simply designed to understand the behavior of Twitterers as it relates to healthcare marketing. Similar to how pharma uses CRM opt-in email campaigns, the goal was to see if Twitter could be used to build a list of target prospects who could then be contacted and converted  to new patient starts.

Questions We Sought to Answer

In Phase One of our Pharma-Twitter experiment we set out to answer the following questions:

1)      If you use Twitter search to identify and then contact people with a specific condition, will it be rejected as Twitter-spam or be welcomed for its contextual relevance and timeliness?

2)      Do people prefer to follow a person or a brand, and by what margin of difference?

3)      What type of person-profile will get the most follow-backs? Specifically

a.  A person who shares your condition

b.  A person who shares your condition, but is representing an unbranded website

c.  A brand manager for an insomnia drug

In Phase Two we sought to answer a fourth question:

4)      What impact does the profile picture have on follow-back response rates?

Choosing a Condition - Insomnia

For the experiment we chose to focus on the condition of chronic insomnia. It was picked for a variety of reasons including the fact that insomnia is a common condition, thus likely to have a lot of observable mentions on Twitter. It’s also a condition where patients have a lot of influence over whether or not they choose to treat the problem, and over which type of insomnia medication to choose.

Creating a Fictitious Drug and Profiles

We created four different fictitious profiles on Twitter. These were:

1.       A regular person with no association to insomnia or a drug; this was our control

2.       A person who mentions they have insomnia in their profile; this was considered a patient-peer unaffiliated with pharma

3.       A person with insomnia who is representing an unbranded insomnia website; this was our paid patient opinion leader profile

4.       An insomnia brand

The fictitious brand was called Restira, and the simple profile is shown below:


All three people profiles looked something like this:


All profiles used similar photographs of the same person and similar, generic information for the name and bio. All profiles posted one tweet on the first day of the test and then were silent.

Searching Twitter for Insomniacs - 400 People a Day

We searched twitter for the term “insomnia”, then scrubbed the list to eliminate all those who were talking about the song insomnia, the band insomnia123, the book, the movie, nightclubs, coffee shops, and people who wished they had insomnia in order to get more work done.  We kept only those Twitterers who were talking about having insomnia, and after looking back about 8 hours in the course of a single day, we found more than 400 people who fit our criteria.

We followed all the people on Twitter within a day of their tweet about “insomnia.” We rotated the fictitious profiles so each profile followed 100 Twitter insomnia sufferers. We then watched what happened over the next two weeks.

Which Profile Got the Most Follow-backs?

Our control profile, a mom with no relevance to insomnia, was followed back by 7 percent of the people. Our two self-proclaimed insomniacs came in virtually tied with a 14% response for the person associated with the unbranded website, and 13% for the person without that affiliation. The non-person profile, the brand Restira, received a 5% response.


Discussion on Response Results

Our control profile, the generic mother of two, received a 7% follow-back response rate, which is actually lower than we would have guessed since we falsely assumed that most people have setup an auto-follow feature on Twitter.

While it’s true that the Restira brand performed poorest of our four profiles, a 5% response rate would be considered an amazing success by most direct response marketers.

What’s most interesting is that in this experiment people were twice as likely to follow a fellow insomniac as a generic person, and almost three times more likely to follow a health peer than a branded drug. Rather than being turned off by the obvious search and follow approach, it actually doubled the chance that someone who be interested in following someone back.

Phase Two - What Difference Does An Image Make?

Still wondering whether existing in Twitter as a brand was possible if the goal is to gain followers, we set up two new profiles. Specifically we wanted to test the idea of personalizing the brand with the name and picture of the product manager, and also test the idea of using more of a conceptual, interesting profile photo.

For both profiles we used a similar profile name for our fictitious drug Restira, the same bio, and custom backgrounds. Then we chose two different pictures. One was a professional head shot of what could be the Restira brand manager, and the other was a conceptual image of a sleeping woman. These profiles are shown below.


Over a one week period each profile followed 200 people who matched our previous insomnia criteria.

Phase II Results

Using a person’s name and image instead of the brand logo indeed had a dramatic impact. The follow-back response for this profile was 8%, significantly better than the 5% rate the logo alone got. Very surprising though was the fact that the “sleeping girl” picture received a 10% response rate. This doubles the result of the logo alone, and is 25% better result than a picture of an actual person.

Additionally, after a one week period, all of our profiles had a similar block rate of about 14%.


Results Will Vary, and Next Steps

The results of the first two phases of our research suggest that brand managers would be wise to work with a patient opinion leader to lead an unbranded website to maximize their Twitter followers. If the preference is to tweet from the brand itself, the use of an interesting photo that is not the brand logo will get the best results.

As with any study, the results presented here should be viewed as general indicators and should be used to shape the design of further research. One cannot assume that response rates in different disease states would match those we found in insomnia. Critics of our study might point to the limited number of people followed, the fact that we did not have a long chain of prior tweets in each profile, or even be critical of the profile pictures themselves. Our next report will show the results of our experiment to drive followers to an unbranded website, and ultimately to convert to coupons for prescription starts.

To get future research, click here to subscribe to the Kru Report.


Post a comment
Write a comment: