In this blog, I will share a few suggestions to develop physician segmentations in a manner that will generate greater business value for pharmaceutical companies. The focus of this article is on segmentation based on attitudes or a combination of attitudes and behaviour. Such an approach has endured in providing strategic direction even as behavioural databases are becoming more accessible and easier to analyse.
A segmentation of physicians accrues many advantages. One can select and prioritise the target groups, tailor the communication, and create relevant engagement programmes using the most effective channels. Yet, having observed many segmentations, we are often struck by how, repeatedly, the same groups emerge across different therapeutic areas. Roughly, there are those who adopt new drugs fast, those who prioritise efficacy over everything else, those who focus on patients’ quality of life, and different variants of traditionalists – driven by price, regulation, and caution. The nomenclature and sizes vary, but with surprising consistency, these groups recur. In itself, that may not be a problem – after all, these segments may be manifestations of fundamental variations in human nature which get reflected in the context of medical practice. However, there are two problems that are significant. First, should one spend large sums of money and effort to get an answer that can be reasonably guessed? Assuming that these were indeed the correct segments, one could size them with market research at a fraction of what typical segmentation studies cost. Second, if everyone in the industry worked with a similar segmentation, there would be no differential advantage relative to competitors. There are two underlying reasons for this. First there is an implicit assumption that one true optimal segmentation exists that is arrived at with a scientific and robust, yet standardised methodology. Methodologies used for segmentation are indeed quite rigorous relative to most market research studies. You form initial hypotheses about the segmentation, conduct qualitative research to generate dimensions and refine the language, craft a structured questionnaire, field it with a large sample, conduct multivariate analysis including some form of factor and cluster analysis, choose and profile the segments, and develop recommendations. Despite this being true, it is our belief that at every step of the process there are opportunities to make choices that can cumulatively lead to radically different segmentation solutions with profoundly diverse business implications. Second, confirmation bias often plays a role in the process. A prior experience of segmentation in a different context may lead to a desire to replicate the solution which can bias decisions taken at every stage of a project.
At Insight Dojo, we have always made a conscious effort to create solutions for clients that generate business value and provide a differential advantage relative to competitors. To do that, we have drawn upon our experience of developing segmentation not just in healthcare, but in many different sectors, e.g., packaged goods, consumer durables, high Tech, B2B, telecomm, retail, media, and financial services. I am sharing five suggestions below to make segmentation deliver greater business impact:
1. Build the segmentation in the context of your biggest strategic obstacle
Companies often do a segmentation just as a routine part of the marketing process, almost independent of the business context. However, a great segmentation can be a powerful aid in resolving your business obstacles and unlocking growth. For instance, we were supporting the launch of a novel hospital antibiotic. The big challenge was that it would have been niched to a “save for last” position, and certain patients whose lives could be saved with this antibiotic would be denied this treatment because of its default positioning. It was critical for us to identify the most influential physicians who not only determined the protocol for hospitals, but also actively felt the unmet need for early use of this treatment for certain patients. We found two distinct segments that comprised a small fraction of the physician population yet were vital for realising the early use opportunity. These two segments were prioritised by our client for a highly successful launch.
Building the segmentation around the business obstacles allows one to create the right segmentation frame, elicit the appropriate information from physicians, choose the best analytical techniques, prioritise the suitable segments, and ultimately design a programme to change behaviour that will help overcome the obstacles.
2. Do qualitative research as a stand-alone module with a view to discover
One of the best ways to overcome confirmation bias is to do deep and exploratory qualitative research and imagine that there wouldn’t be any quantitative research. Such a mentality pushes us to take a fresh look at the market and discover crucial aspects about behaviour that we wouldn’t anticipate.
There are three things one must do to conduct such qualitative research. First, one must cultivate a fresh mindset with a willingness to be surprised. Second, one must use qualitative techniques such as ethnography that are more suitable for discovery. Third, the sample design needs to be robust. To pick up different segments in qualitative research, one needs high sample sizes. Our sample sizes for such research have ranged from 30 to 100 interactions with physicians.
Consider the example of a drug related to women’s sexual health. In our effort to increase sales, we conducted a robust ethnographic study and discovered a small group of female gynaecologists whom we called the “voice of the women” segment. They were passionate about reducing gender inequality in healthcare and believed that the client’s product was more respectful of women’s lifestyle and needs. They were strong advocates and prescribed this drug to a much larger extent than any other group. The significance of this group would have been impossible to assess if we had relied mainly on quantitative research. It would have been difficult to spot such a niche segment without meeting these physicians face-to-face and interacting with them ethnographically in their own environments. Additionally, the relatively robust qualitative sample size allowed us to profile this segment meaningfully, before sensitively capturing and sizing it in the quantitative research.
It is also true that certain situations warrant a segmentation based only on qualitative research, e.g., if the universe of doctors is really small, as happens in the case of certain rare diseases, or when a market is rapidly changing. However, even if there is a follow-up quantitative research planned, we advocate conducting qualitative research as if it were a stand-alone module rather than to mainly support quantitative research as it always helps unearth novel and useful insights.
3. Sync-up business thinking and the multi-variate analysis in quantitative research
Often, there is an analysis team that generates a set of segmentation solutions using some variant of cluster analysis. The business team then selects a solution, interprets it for business relevance, and develops recommendations. Even if there is a certain degree of back-and-forth iteration, it does not substantially alter the final solution. The business team treats what the analysis team provides as a given and works with that which lead to all kinds of challenges. For instance, a client showed us a segmentation that combined prescription behaviour and attitudinal data. When we looked at the data, it was obvious to us that despite having names and descriptions that suggested attitudinal differences, the solution was driven dominantly by prescription data, and the segments were basically light, medium, and heavy prescribers of a certain drug class. The analysis team, through their “proprietary model”, let the variables with the maximum variation drive the segmentation with very little business judgement. Though, a mistake like this seems pretty basic, it is often not picked up because of the disconnect between the analysis and the business strategy team.
More significantly, when the interaction between the business and analysis teams is seamless, the ability to experiment and create effective segmentations increases dramatically. In the hospital antibiotics study, we first followed an “unsupervised learning” approach and conducted cluster analysis without a dependant variable, as is common practice. We felt the segments were not as sharp and insightful as what we had observed in the qualitative research. We switched to an approach where the segmentation was developed to explain a dependant variable which we chose to be the number of prescriptions of other novel antibiotics for early use made by an individual. There are specific techniques suitable for such analysis, e.g., latent class regression. We were able to improve our segmentation significantly in terms of business value and intuitiveness.
At Insight Dojo, business problem solvers are the ones who do the hands-on analysis. But, even if that is not the case, a strong link needs to be forged between the business thinkers and the analysis team.
4. Profile the segments deeply
Often the game-changing insight comes from the intersection of the segments and another variable. For instance, when we were segmenting oncologists for a Non Hodgkins Lymphoma (NHL) condition, the belief amongst doctors and our clients was that by the time patients reached third line, physicians improvised and what they prescribed was unpredictable.We identified three priority segments and then tested what they would prescribe to a set of pre-defined patient profiles. That gave us our Aha! insight – we could predict accurately what each of those segments would do when they encountered a specific patient profile. This was an example of a real insight when the discovery of a hidden pattern completely replaces the belief that physicians are simply improvising.Based on this we could identify the “sweet spot” for our client’s therapy in terms of target physician segment, patient profile, and unmet need. We could also identify the subset of important competitors for the “sweet spot” and sharpen the positioning of the client’s therapy. Segments need to be profiled thoughtfully on many other aspects, e.g., suitability for opportunities along the treatment pathway, their channel preferences, and other critical variables.
Profiling segments deeply is a non-trivial point. This usually results into a longer questionnaire. The value of adding profiling variables must be traded-off with the loss in quality due to respondent fatigue. Furthermore, one needs to ensure relatively large sample sizes to dissect the individual segments. This can be challenging and expensive in the context of physician research. That said, it is a lot more cumbersome and expensive to do separate studies and try to link the segments derived from one study to the data collected in another study.
5. Pressure test, embed broadly, and segment physicians through the life cycle of a drug
Having conducted a segmentation, it is critical to pressure test it with various stakeholders, especially those who are customer-facing. The criteria used relate to statistical robustness, e.g., replicability, business relevance, e.g., the size of the opportunities supported, and ease of implementation. Customer-facing stakeholders, in particular, must find the segmentation intuitive based on their experience, even if the segmentation itself is an entirely new way of looking at the customer.
Segmentation studies are usually conducted in preparation for the launch of a drug. , and the typical target users are the salespeople. To embed the findings, typically the segments are brought to life in a workshop, and then a quantitative typing tool is created to identify segments with a short set of questions that can be implemented by the salesforce.
However, for substantial business impact, segmentation must be used at all stages of the lifecycle of a drug. For instance, segmentation can be used at the early stages of drug development, e.g., whilst recruiting physicians to participate in clinical trials. Understanding a physician’s priorities can increase the likelihood of their participation in a programme. It is important to highlight that often the type of segmentation needs to change based on the business priority and stage of the drug. In the case of the women’s sexual health drug described above, the “voice of the women” segment could be discovered only after our client’s drug had been launched – the very act of making this new drug available made it possible for this segment to express itself. If our client had continued to use the segmentation that they had developed at the time of launch they would have missed the opportunity to engage this highly valuable segment.
It follows that users of segmentation cannot be only salespeople. We have trained commercial teams, medical teams, and C-Suite executives because all of them, in some capacity, interact with physicians. In one instance, the CEO was so inspired by the value of the segmentation training that he got us to run a session with the entire board.
Having such a broad set of stakeholders as users requires that the segments can identified in different settings. A CEO whilst meeting a Key Opinion Leader is unlikely to administer a questionnaire to collect data for a typing tool. In addition to creating the conventional typing tools, we have developed guides and trained our clients to identify segments in informal conversations.
Following these recommendations should significantly improve the business value of your physician segmentation. Contact us if you have further questions or if you want us to help you on this topic.