Is THIS ART OR SCIENCE?

Market segmentation is a necessary step before creating any integrated marketing communications plan in order for it to lead to brand equity!


Michalis A. Michael
Izabella Jaskolska

When referring to segmentation most experienced researchers think of factor and cluster analysis. In conducting ad-hoc factor or cluster analysis a great deal of subjectivity (and creativity) flows into the "equation" in terms of selecting the "best" factors and cluster solutions for doctors or patient data. Hence the title of this paper 'Is this art or science?' The same question can be asked in a more concrete way: Is this (segmentation) subjective or objective? Like most things in life the answer is not black or white. There is a large grey area and most definitely our answer will be found there. No surprise here. You do not have to wait until you reach the end of this paper to know the answer to the question: it is indeed a combination of art and science or a combination of subjectivity and objectivity. However the real issue here is how subjectivity can influence segmentation and by extension the marketing plan which in its turn affects the growth or success of the brand.


MARKET OR CUSTOMER SEGMENTATION


Market segmentation was first described during the 1950s, when product differentiation was the primary marketing strategy used. In the 1970s and 1980s, market segmentation began to become more popular as a way of increasing sales and obtaining competitive advantage. In the 1990's, direct marketing professionals used many sophisticated techniques, including market segmentation, to reach prospective customers with the most tailor-made proposition possible.

Segmentation is often regarded as science because it utilizes sophisticated statistical techniques to arrive at derived segments that are meaningful for developing differentiated communication strategies. The "science" designation, however, is often controversial because there are so many different methods that might lead to different results, and also the same method applied by different people might result in different findings and conclusions.

Today it is common wisdom that segmentation is a necessary step before a successful marketing effort. The word marketing is key here because a few years back this was a word reserved for the FMCG and some other sectors, but certainly not for pharmaceutical companies in the ethical drug business. This has definitely changed in the last few years, not only in western economies, but also in tier 1 Central Eastern European countries (such as Poland, Hungary and Czech Republic). One might say we are in the middle of it all. It is now becoming clear that the brand is important also in the ethical drug business, and communications that build brand equity are a necessity. What makes things a little complicated, however, is the fact that in the case of the pharma industry it is not the end user (patient) who decides what to buy but the physician. The debate about who is the ultimate decision-maker continues to rage, but we believe that the patient has at least an influence on the decision about what drug to prescribe. Thus not only physician segmentation but also patient segmentation can be instrumental in devising a successful marketing plan.

There are many different ways to classify or to segment a market. The main differentiator between segmentation schemes is often related to whether they are based on existing factual databases or if primary research is utilized in order to collect descriptor variables from the prospective customers. What is applicable, when, and how, is a huge subject, the discussion of which would take us out of the framework of this paper. Let us however impose these thoughts to you:

As already explained, factor and cluster analysis is not all there is when thinking "market or customer segmentation". Classification algorithms is another way of doing it and quite in fashion lately among multinational market research agencies, especially when they are packaged and presented as a turnkey solution. In such a case respondents are classified in predefined segments based on some objective criteria (see the example in figure 1).

In this example customers are classified in Brand Relationship Segments. The question should always be: how can this segmentation help me grow my business (by implementing a successful marketing plan)?

Growth - Commitment (relationship) + potential to buy or buy more. If we only stick to relationship then we can target customers who need and are inclined to buy the product but what about their ability to purchase it?

The ideal segmentation should tell a company which prospective or existing customers to target with specific communications through specific media and with what specific core message in order to induce higher spend on the brand.


Click to enlarge



Figure 2 shows four predefined segments geared towards growth (potential) and commitment (relationship). A customer with high relationship and high potential (to spend more on the brand) is called "Dynamic". One with high relationship but low potential is called "stable" or core. On the other hand low relationship but high potential is labelled as "Challenging" and low relationship combined with low potential as "Reluctant".

Imagine the power of such a chart when combined with a detailed description as to who the "dynamics" are (demographics), where to find them (lifestyles) and what to tell them (attitudes toward statements) in order to induce growth.


SEGMENTATION IN HEALTHCARE vs. OTHER BUSINESS SECTORS AND THE LINK TO BRAND EQUITY

The challenge for multinational healthcare companies is to find the right balance of "global" and "local" strategies, tactics, decisions, etc. Being able to transfer "global" to "local" means cost savings, in general because each geography does not have to re-invent the wheel. On the other hand there is an enormous risk in doing this in the case when "local" is very different from "global".

In segmentation matters there are two possibilities:

  1. Each (or some) of the involved countries have the same number and type of segments in common (or have some of the segments in common), although the size of the international segments may differ. We refer to this as an international segmentation approach/scheme.

  2. The involved countries have no common segments. This means that the segments are not comparable in terms of the nature of the segments: i.e. the identified segments differ structurally on the variables on which the segments are identified. We refer to this as a country-specific segmentation scheme.

This is what is applicable when it comes to segmentation planning at Eli Lilly and possibly in more broader terms in the Healthcare sector:


It is very interesting that in a parallel situation to the Doctor-Patient decision making process for prescription of ethical drugs, like the Retailer-Shopper, the questions this paper handles were seemingly answered years ago. Have they been answered really? Let us examine this parallel:

We cannot repeat often enough that the underlying question before every segmentation should be: How is this segmentation going to help me achieve growth? A breakdown of the steps to the ultimate goal (growth) to make it more simple is:

  1. Find them (define the target population in broad terms)
  2. Segment them and choose (targeting)
  3. Tell them (marketing and communications)

If the segmentation was conducted among the shoppers of one specific retailer chain or even at the store level, in the case of retailers and shoppers, then there would be no need to segment the retailers (who are the customers of the FMCG Manufacturers). However this is not a very economic thing to do (to make a segmentation for the shoppers at each retailer separately), thus retailer segmentation should be used in conjunction with shopper segmentation.

It is widely accepted that the Healthcare industry can learn how to market their products from the FMCG industries but it might as well be that here we have a marketing angle where Healthcare is more advanced than FMCG: Double source segmentation to more successfully reach a single target and extensive use of Direct Marketing or Permission Marketing. Some FMCG sectors, such as Tobacco and Alcohol with the ban in advertising, are in a very similar situation with Healthcare companies in terms of marketing and there is a lot they can learn from Healthcare.

Brand Equity in FMCG, although a controversial subject - since it is very difficult to put a "price tag", in other words a financial value to a brand name - can somehow be calculated in tangible (current/future sales) and intangible (good will) terms. A very good way to estimate the brand equity of a specific brand in "dollars" is to conduct a relationship segmentation combined with the potential to buy more as described in the first chapter of this paper. If we assume that the life cycle of a product is ten years, for example, then we can forecast the sales to each customer segment based on their current spending on the brand and on the assumption that they will switch segments as a result of an integrated marketing plan. The IMC will aim to move more brand users to "Dynamics" (see previous chapter) and non-users to users. Consequently not only Brand Equity can be estimated using Segmentation but more importantly Segmentation is the right start in building it (brand equity).


CASE STUDY

Eli Lilly decided to carry out a market segmentation in Poland for a new drug they planned to introduce. For the purposes of this paper we will call this drug "Drug A" and the disease for which it is applicable "Disease A". Since Drug A was already in use in other countries. a "global" segmentation of patients and doctors was already available. This global segmentation was the result of segmentations done in 10 - 12 countries.

It was decided that the market segmentation in Poland should include physicians based on primary quantitative research (carried out in August 2002) using factor and cluster analysis and also patients as seen through the eyes of the physicians using primary qualitative research (carried out in November 2002).

Factor and Cluster analysis was used for the physicians. Qualitative Segmentation (FGIs) was used for the patients because this was explicitly requested by Eli Lilly.


GLOBAL PATIENT SEGMENTATION


The global segmentation of patients reflects how the patients are perceived by the doctors. This segmentation consists of a two-dimensional approach: a crossing of the physician types and the patient types (see figure 3).


Click to enlarge


The patient types are described in detail (see table 1).


Table 1 - PATIENT PROFILES, CONTINUED

Profile #7
[Disease A (Symptoms A)]: Patient...
  • has a long history of illness
  • is compliant and highly functional
  • is employed
  • has symptoms which are well controlled
  • current visit is a regularly scheduled evaluation
Profile #8
[Disease A (Symptoms B)]: Patient...
  • exhibits periods of grandiosity
  • has no depressive symptoms
  • has decreased need for sleep
  • is euphoric and elated

  • is hypersymptom
  • has symptom H, K
  • has symptom L
  • visits due to an incident
Profile #9
[Bipolar (Symptoms C)]: Patient ...
  • experiences symptom C
  • exhibits a symptom F and M
  • has symptoms X
  • reports difficulty in activity A

  • exhibits symptom Y
  • is less functional
  • visits due to incident
  • reports having difficulty with activity A
Profile #10
[Disease B (Symptoms D)]: Patient ...
  • has long history of episodes
  • is male and overweight
  • exhibits symptom R
  • has symptoms P, Q
  • has symptom Z

  • has symptoms D, E
  • symptom I
  • has symptoms P, Q
  • visits due to deterioration and poor symptom control
  • has received an injectible drug in the past
Profile #11
[Symptoms E]: Patient ...
  • is between 60 and 70 years of age
  • has symptoms G, H
  • is burdening caregiver
  • is beginning to decline
  • is being seen because of reason D

Source: Eli Lilly


In crossing the physician segments with the patient segments they treat, we create a new segmentation based on patient needs and physician activities with them. This approach can result in grouping some \ patient segments as well as some physician segments to form fewer segments which provide actionable information to the marketeers. This is a global tool, used by Eli Lilly, but the question here is whether this Global Patient Segmentation is valid for the Polish market? Do the primary and secondary targets (identified in the Global Patient Segmentation) they have the same importance in Poland?

This was exactly the objective of the research/segmentation done in Poland: to learn if a different segmentation should be used due to local specificities.


Segmentation of Physicians using Primary Research in Poland

Although segmentation was the main aim of this study, it also provided an overview of the situation among physicians. Information was obtained about Disease A, the patient types and consultation patterns, the diagnosis and treatment patterns, and evaluation of the key products available on the market for this disease. Prescribing potential among the examined doctors was also investigated.

Study design: 200 face-to-face interviews were conducted with physicians using a paper questionnaire. The interview was quite long, about 50 - 60 minutes; the interview flow, reflecting its complexity, is shown in figure 4.


Click to enlarge


The interviews were spread across all Poland, taking place in 20 major cities - centers of medical treatment.

The basis of the segmentation was 50 statements describing the habits and attitudes of physicians. Some of the statements were quite general, touching the physicians' more general attitude towards their life and work, and some were quite detailed, touching their day-to-day attitudes towards medical representatives and their prescribing patterns. A few examples of the statements are shown in table 2.

The statements originated from Lilly's experience in the countries, from Lilly's local Polish team and from Synovate's experience from medical research mainly in Poland. Defining the statements is an area where a lot of creativity can be applied.

The assumption was that probably not all statements would be necessary to differentiate the doctors' opinions, but on the other hand we were expecting that many would not be appropriate based on our previous experience.

Cluster Analysis can be either performed on the factors extracted after a factor analysis or directly on a set of independent variables.


Click to enlarge

Factor analysis was used to identify the core number of statements. Only a few statements were excluded after this step from the subsequent analysis. The reason was that those statements were highly correlated with other statements, which described the doctors' attitudes better.

Here are some of the statements that were excluded:

  1. I believe, that for the majority of patients with Disease A, leading a normal life under control of drugs is possible.
  2. In treating disorders I rely on opinion leaders (leading specialists) in my field.
  3. If a specific drug is not showing the desired efficacy, I first try to use other dosage before switching a patient with Disease A to another drug.
  4. In treatment of disorders I also use psychotherapy.

The next step was to run K-means Cluster Analysis using SPSS. The set-up was done for two, three, four and five clusters. The task now was to "subjectively" decide what number of clusters was the optimum for our objectives. The assumption was that more than five clusters are not reasonable from the practical point of view. When the results for different splits where generated, the subjective opinion and / or creativity of the Synovate analysts was employed.

For each option there were prepared short descriptions of the segments based on the key features reflected by the statements on which the particular segment had significantly higher or lower mean score (see table 2). The initial segmentations were subject to evaluation of the researcher and the statistician regarding two aspects: logic and consistency of the key features of the segment as well as probability of existence of a group with the given profile in reality (confrontation with knowledge about the market). Clusters had also to show meaningful differences between them. Additionally the profiles of clusters were compared regarding their consistency with parameters not used for segmentation - socio-demographic parameters (i.e. length of practice, number of patients with Diseases A and B).

The segmentation option, which was finally accepted, presented the highest differentiation between segments and the highest consistency within clusters as well as consistency with expectations and relevant market knowledge of the researcher. The choice was done according to the statements that best described the clusters from the client's point of view, i.e a 5-point scale indicates the degree of agreement of the physician with the statements. So the statements (attitudes) were the basis for the optimum cluster number selection. The three-cluster split was judged to be the optimum; each cluster was then characterized by many socio-demographic characteristics. So the group called later Disease A Experts had a higher average number of Disease A patients in comparison to the group of Disease A Non- Experts. Disease B-Focused had a higher average number of Disease B patients in comparison to the other clusters.

Although it is difficult to give recommendations that are generally applicable, we believe that it is very useful from a practical point of view to have some guidelines. During the process of selecting the clusters it is very important for the agency to have a very detailed understanding of client needs. The market segmentation results need to lead to a market segmentation strategy. This encompasses more than the numerical and statistical results of using the techniques described in the previous pages. To ensure the relevance of the identified market segments, and to ensure that the identified market segmentation fits with the organizational strengths and weaknesses (SWOT - analysis), and could be fully implemented in the marketing planning of the clients' company. We recommend working with multiple segmentation bases since this increases the extent to which we are satisfying our segmentation criteria. After initial analyses we recommend discussing these preliminary findings extensively with the several groups within the client's company: i.e. the market researchers, marketing, etc.

The names of the three clusters we used are:

Admittedly we have not been very creative in selecting fancy names for the clusters, but these were the names that best described them.


Click to enlarge

Detailed description of the segments was provided (see figures 6 - 8), in terms of:


Click to enlarge


Click to enlarge

To provide one example of a differentiator of each segment:

The first two segments consist together of about 80% (40% of the doctors each), and the last one consists of about 20%.

The third segment, although the smallest in doctors' population, is the most important when taking into account the Disease A specific needs. For example, these 20% of doctors treat 30% of Disease A patients and prescribe 36% of all Drug A packages prescribed to Disease A patients (see figures 9 and 10).


Click to enlarge


Doctors belonging to different segments have a different perception of their reality. In particular they perceive drugs available on the Polish market differently. For example, see in figure 11 how differently Drug A is perceived by our three segments.


Click to enlarge


On the horizontal axis we plotted the importance of attributes (regardless the drug) and on the vertical the association with a specific drug (Drug A in this case; this varies from drug to drug).

For comparison purposes figure 12 shows how different the picture would be, had we not done a segmentation. Figure 12 alone could lead us to design a different IMC than the one we will have as a result of the segmentation.


Click to enlarge


Verification of the Global Segmentation of patients through the Eyes of the Physicians using Primary Qualitative Research

We conducted six focus group discussions with physicians in three cities. Ideally we should have liked 20 focus group discussions to get a more quantitative feel in the results. The fact is that Global Segmentation was available and financial constraints made us choose for the lower number of groups. The important element here was the usage of the physician segmentation from the quantitative phase when recruiting participants for the groups. Namely for three groups (one at each location) we recruited physicians from the "Disease A Experts" segment and for the other three, physicians from the "Disease B Focused" segment (equally split in terms of location).

The physicians were selected to participate in different groups based on certain statements differentiating the most between the segments. There were nine such statements selected, however experience has shown it would be even better to have more statements, say closer to 20 for easier differentiation.


The detailed objectives of this study were to:

Upon completion of this exercise, local market researchers should be able to confirm that these patient segments are real, exist in the Polish market, get a rough feel for their relative size and should be able to modify the patient segment descriptors in a language that Polish physicians best understand. This exercise should also provide a useful insight into the local physicians' behaviour towards managing / treating Drug A and the reasons for their therapy choices for different patient segments. This information can be utilised at the stage of message development to ensure the message accurately describes the solutions for the targeted Patient segments needs.

During the recruiting process the physicians were identified by recruiters using a specially designed questionnaire. In the questionnaire there was a section with some statements which helped assign each physician to one of three segments (see example in table 3).




The results of the research confirmed the existence of Global Patient Segments in all three Disease A types, which is interesting for Lilly in terms of Drug A. What is more, belonging to different physician segments (Disease A Experts or Disease B Focused) has no influence in how they perceive the Global Patient Segmentation.

The Global Segmentation of Patients happened to be equally valid in Poland for the two most important segments of physicians. This meant that Lilly could apply the Global Patient Segmentation in Poland using the same messages to both target segments which were defined in the physicians' segmentation.

If the results had shown differences among the two target segments, we would then have to create two, to some extent different, messages with regards to patients for the two target segments.

Having said all that, a quantitative approach to segmenting patients by interviewing the patients and not the doctors (like in this case) is a more accurate possibility when the disease type allows. However, it might prove difficult to find patients who can openly discuss a sensitive disease. In such a case our approach is a very appropriate one.


DESIGNING AN INTEGRATED MARKETING COMMUNICATION (IMC)

We "found" them, "segmented" them, now its time to "tell" them. This part will be based on best practice within Eli Lilly and best known practice in broader terms. The marketing plan for Drug A is in the process of being prepared, thus in this paper we can only describe what the plan could be. The Global Branding concept as well as material will be used in Poland since the global patient segmentation has been confirmed (see patient segments B to H I figure 3).

There are two main groups of marketing tactics:

Segmentation is an absolute necessity for both groups: for the former because we need to segment our customer database and assign each one to a segment, allowing us to communicate only what is relevant to them and will have an impact; and for the latter because our communication in the mass media cannot be everything for everybody. We need to target different groups with different messages.


The IMC can consist of the following elements:

Below you will find the input to the marketing communication plan for each segment, only from the physician segmentation using the primary quantitative study we conducted. In this input you will note there is not much about the patients and this is exactly why the double angle approach (i.e. patient segmentation) can offer more to an IMC. It can make the messages more distinct, targeted to the specific physician segments and impactful.

This is something Eli Lilly will do during the next few weeks and will not be explicitly discussed in this paper for confidentiality reasons.


Figure 13a - DISEASE A EXPERTS - COMMUNICATION STRATEGY RECOMMENDATION

Drug A MESSAGE - Symptom A TREATMENT

Communication recommendations


Figure 13b - DISEASE A NON-EXPERTS - COMMUNICATION STRATEGY RECOMMENDATION

Drug A MESSAGE - Group Symptoms — B TREATMENT

Communication recommendations


Figure 13c - DISEASE B FOCUSED - COMMUNICATION STRATEGY RECOMMENDATION

Drug A MESSAGE - Group Symptoms — B TREATMENT

Communication recommendations


CONCLUDING REMARKS

Among a number of available quantitative segmentation methods, factor and cluster analysis seemed to be the easiest and most acceptable way to do it, however there are two questions that might be answered in a different way depending on the situation:

In our case we took the second option and eliminated a few variables but left in over 40 variables (although in some sources you will find that more than 10 factors should be avoided) which had helped us do the evaluation. The more variables you have the more room for subjectivity (art) there is in deciding which number of segments to accept. Of course calling subjectivity: creativity (art) sounds much better and there are two levels of creativity applied here:

Additionally, creativity was applied in using qualitative research to segment the patients through the eyes of the doctors, globally and locally in order to identify possible differences. The two approaches to segmentations of physicians and patients are connected by the fact that in selecting the respondents of the groups we used the physician segmentation results in order to classify each FGD participant. The objective was achieved in finding out that the different physician segments agreed on the existence of the same patient segments, but more importantly these patient segments are the ones taken from the Global segmentation, i.e. it was proven that in Poland in this case similar messages with regards to patients can be used.

The integrated marketing plan can now be discussed having in mind both the perceptions of the physicians about the disease and the available drugs as well as the different groups of patients (in this case seven primary groups) that require a different approach by the physician when treating.


REFERENCES

Ogden, James. (1988). Creating a creative and innovative integrated marketing communication plan.
DSS Research website
Digitalplays.com website


THE AUTHORS

Izabella Jaskolska is Market Research Manager, Eli Lilly Poland

Michalis A. Michael is Managing Director Central Europe, Synovate, Poland.