Copy of Creating a Segmentation

creating a segmentation

Session 2

The purpose of this session is to describe the steps of conducting a segmentation in more detail and provide an example of an advanced segmentation effort developed to inform malaria SBC programming. The case study we will review is of pregnant women at risk of malaria in Cameroon, Côte d’Ivoire, and Malawi. This session will describe the segments identified through advanced segmentation analyses, as well as their defining characteristics and the factors influencing their behavior.

Learning Objectives

List the steps and resources required to conduct an advanced audience segmentation with a quantitative foundation and complementary qualitative research.

Describe an advanced audience segmentation conducted among pregnant women at risk for malaria.

List key distinguishing factors of each segment.

Describe tools to segment members of a target population to tailor SBC interventions.

Steps to Create a Segmentation

Session 1 introduced the concept of audience segmentation, advanced audience segmentation, the role of advanced audience segmentation in malaria SBC, and the steps required to conduct advanced audience segmentation. Session 2 will describe each step required to conduct advanced audience segmentation (steps 1-7) in more detail. (Steps 8-11 are covered in Session 3.)

  1. Identify the priority behaviors
  2. Define the target population
  3. Enlist key stakeholders
  4. Develop research questions
  5. Select a dataset
  6. Define segments
  7. Refine your segments
  8. Develop intervention elements
  9. Pilot solution
  10. Evaluate & Refine
  11. Adapt & Scale

Step 1. Identify the Priority Behaviors

Audience segmentation provides great insight into how certain behaviors vary across a population and can be used to identify and prioritize groups for social and behavior change. To leverage saudience segmentation, start by identifying the priority behaviors. For example, what malaria-related behavior needs to be addressed in your local context? What is the desired behavior change that will result in improving case management or reducing malaria incidence?

Below is a list of several malaria prevention and control priorities. Use these as a starting point for brainstorming the desired social or behavior change for your segmentation.

  • Use an ITN correctly every night. / Increase correct and consistent use of ITNs
  • Attend ANC early and frequently / Increase early and frequent ANC attendance
  • Accept and use IPTp / Increase acceptance and use of IPTp
  • Seek quality care for fever within the same day or next day of fever onset / Increase prompt seeking for fever
  • Adhere to national malaria case management and malaria in pregnancy guidelines / Increase adherence to national malaria case management or malaria in pregnancy guidelines.

Step 2. Define the Target Population

Second, determine which population you want to encourage the desired behavior change within. This population may be particularly vulnerable to malaria or impacted by malaria. Examples of populations of interest to focus on malaria behavior change include pregnant women, caregivers of children under 5, health providers, farming communities, and more. Priority populations for malaria interventions are most commonly pregnant women and caregivers of children under 5.

Step 3. Enlist Key Stakeholders

Next, identify individuals, including experts in the research methods required to conduct segmentation analyses. It is important that stakeholders are aligned to the objective of the audience segmentation identified in steps 1 and 2, as they will be closely involved at different points in the process.

It is important to include a mix of stakeholders in the process, ensuring that they have an equal voice in providing feedback and decision-making. Ensuring a diversity of voices in the decision-making process is vital to designing inclusive solutions as they have direct impacts on people’s lives. Flip each card to learn about who to involve in the segmentation process and their potential roles. Consider the functions that already exist on your own team, then fill in the gaps.


People with knowledge of national malaria trends, data analysis, or primary data collection. Examples may include researchers, statisticians, vector-borne infectious disease experts, or malaria control experts.


People with experience with in-country malaria control and elimination programs or in implementation and evaluation of interventions. Examples may include program implementers, staff from NGOs, or health personnel working directly with the population of interest.


People with influence over policy and quality improvement efforts for health facilities or at the health system level. Examples may include policy makers, Ministry of Health officials, or individuals from the National Malaria Program (NMP).

Step 4. Develop Research Questions

It is now time to develop 2-3 research questions to guide your audience segmentation. With your team, particularly those knowledgeable about national malaria trends, determine the main factors related to the objective that you want to better understand. It is helpful to conduct a search for peer-reviewed published articles and papers written on behaviors, beliefs, and attitudes towards malaria in your country to ensure your research questions are not duplicative and are driven by existing data of what is already known. Consider the following:

  • Which factors influence the target population’s practice of malaria-related behaviors?
  • Among members of the target population, what are the characteristics of the individuals who are most likely to practice the priority malaria-related behavior?  What are the characteristics of the individuals who are least likely to practice the priority malaria-related behavior?
  • How can the factors that influence the target popluation’s practice of the priority malaria-related behavior be addressed through SBC interventions?
  • What is the likelihood that a member of the target population will adopt the practice of the priority malaria-related behavior?

Research questions should primarily be behavioral and will act as the outcome variables for your analysis. Examples of research questions from the case study presented in this course are listed in the next lesson under Key Questions & Analysis.

Step 5. Select a Dataset

To conduct an audience segmentation using quantitative techniques, a survey dataset is required. At baseline, this dataset should have the following parameters:

  • Come from a fairly recent survey, administered in the last 1-10 years, with a representative sample of the target population.
  • Report data for each survey respondent individually, not in aggregate.
  • Contain data on individuals in the specific population identified in step 2 (i.e., age, gender, occupation, etc.).
  • Contain variables that can be used as proxies for the outcome variables identified in step 4.

Because segmentation is intended to describe segments and identify the drivers or influencers of behavior, your chosen dataset should also contain variables that are related to the outcome variables you identified in step 4. Flip each card to see examples of variables that your chosen dataset may include. Each type of variable has a different purpose for your audience segmentation dataset.

Demographic Factors


Characteristics of a population that have been categorized by distinct criteria, typically a vital or social measure.

Behavioral Factors


The way in which someone behaves towards the identified objective.

Attitudinal Factors


A way of thinking or feeling about someone or something, that is sometimes reflected in a person’s behavior.

For malaria, high quality datasets such as the Malaria Behavior Survey (MBS) or health facility survey datasets can be used. Both of which were utilized for the case study presented in this course. You may also consider exploring the Demographic & Health Surveys (DHS) datasets to see if they would work for your audience segmentation.

Types of variables that may be present in datasets


Use case

Best for simple segmentation or when combined with other variables as an additional descriptor.

Example characteristics

Age, location, gender, religion, number of children, rural/urban, literacy/numeracy, socio-economic status, household income, level of education, employment status or type, health history/risk factors, marriage status, etc.


Use case

Helpful for determining group actions and behaviors. Best when combined with attitudinal variables.

Example characteristics

Seeks information regarding diseases through specific channels, uses social media, goes to health center for illness or for preventive treatment, seeks care with traditional practitioners, involved in community activities, uses certain tools or methods (e.g., bed nets, diagnostics), etc.


Use case

Use when trying to understand rationale for behaviors. Best when combined with behavioral variables.

Example characteristics

Trust in authority (government, health institutions, health care professionals) and perceived: access to resources, safety or effectiveness of treatment, consequences of disease, social norms (e.g., believes others practice certain health behaviors or not), role of fate/divine will, etc.

Step 6. Define Segments

At this point, it is time to analyze the data and define the segments. Quantitative segmentation analysis occurs in a 4-step process. Click each step to learn more about the process. It may be helpful to partner with your national statistical institute or a research firm to conduct these analyses.

1. Identify factors that are the main influencers or drivers of the research questions.

Step 1

Run a Chi-squared correlation analysis to identify which variables showed a strong correlation with your research questions.  This will identify the key drivers of the behaviors described by your research questions. These key drivers will be your primary covariates in the analysis.

2. Conduct a quantitative segmentation analysis and identify opportunities for positive behavior change​.

Step 2

Analyze the entire dataset using one of the standard statistical methods for segmentation, which include cluster analysis and latent class analysis. These techniques help identify commonalities and trends among groups based on the variables in the data and selected outcome variables. This analysis will produce different options of segments in your dataset.

3. Review each segmentation model and determine the final model.

Step 3

Review model results of the segments identified in your dataset. Use the Bayesian information Criterion (BiC) indicator, to determine which models are statistically significant, with a lower BiC indicating a better model. Additionally, try and aim for a model with three to six segments. Finally, consider other factors to choose the best model:

  • How do the outcome variables and drivers of the outcome variables differ between various groups?
  • Does each group have clear differences or are the differences barely noticeable between some of the groups?
  • Does the model tell a strong story about different groups within the population and their attitudes, behaviors, and beliefs around malaria?
4. Write a segment “persona” for each segment in the final model.

Step 4

Once a viable segmentation has been chosen, write a “persona” for each segment that describes the key distinguishing characteristics of the segment identified in analysis. Choose a name for each segment as well. The descriptive personas should be brief and easily digestible for dissemination and feedback. It primarily describes the segment in terms of the outcome variables and/or drivers of outcome variables.

Step 7. Refine Your Segments

The segments identified through quantitative analysis can be further refined using qualitative data collection and by speaking directly with individuals from each segment to gather additional information about the factors that shape the behaviors most characteristic of their segments.

Segment Identification Questionnaire

First, determine which segment individuals in the target population can be categorized as using a segment identification questionnaire. This will enable you to recruit individuals to participate in interviews, focus groups, or workshops for their segment.

A segment identification questionnaire can be developed using a Chi-squared automatic interaction detector (CHAID) algorithm in the R Studio or SPSS programs. This algorithm considers each variable used in the segmentation analysis as well as the final determined segments and identifies the variables that were most influential in forming the segments. This set of statistically significant variables may be asked in a brief, ordered quiz to members of the target population to determine which segment they can be categorized as.

Complementary Qualitative Research

Finally, conduct qualitative research with members of each segment to develop a more comprehensive understanding of their behavioral drivers. Using the segment identification tool, invite members of the target population to participate in segment-specific focus groups or one-on-one interviews about behaviors and attitudes surrounding the identified objective. Consider partnering with a local research firm to recruit participants, facilitate the interviews or focus groups, and analyze the findings. Aim to have 5-15 research participants for each segment.

Segmentation of Pregnant Women at Risk of Malaria

Now, let us put segmentation steps 1-7 into practice with the case study in the next lesson.

Click the “Mark Complete” button to proceed.

Lesson Content
0% Complete 0/1 Steps