A bit like finding a husband? Success factors for implementing segmentation analysis in social research studies

Written by: Alexandra Cronberg

Introduction

Segmentation analysis is gaining popularity in social research. While it has long been used in market research, this analytical approach can also add value in social research contexts. Specifically, it can help providing an understanding of different needs and motivations among sub-groups in a target population. Consequently it can help donors and agencies tailoring their programmes and interventions and thus increasing the likelihood of success.

There is much to be said about adopting segmentation as an analytical tool into social research. Yet it is also important to recognise the differences between social and market segmentations. This helps to both apply the tool appropriately and to set the right expectations early on. In this light, the blog post here will talk about the main differences between market and social segmentations, and what to bear in mind to ensure segmentation studies are successful in social research.

Now, you might wonder where that husband comes into the picture? Well, bear with me for a moment, but you can think of each segmentation solution as a potential partner. It will all become clear.

Hands holding seeds 2

Examples of segmentation studies

At Kantar Public we have conducted a number of segmentation studies over the last couple of years, including the following projects:

  • Segmentation of the adult population in India, which is one of the countries where open defecation is a major concern. The segmentation explored and helped gain an understanding of people’s toilet acquisition behaviour, drivers, and barriers. The segments identified were Regressives, Conservatives, Prospectives, and Progressives.
  • Farmer segmentation in Tanzania and Mali to understand which African farmers are open to new behaviours. The segments identified were Contented dependents, Competent optimists, Independents, Frustrated escapists, Traditionalists, Trapped.
  • Segmentation of young women and girls at risk of HIV in Kenya and South Africa. This study aimed at understanding risk factors that increase young women’s vulnerability to HIV infection based on behavioural, attitudinal, and demographic variables. The analysis led to segments such as teenage girls just starting to explore sex and relationships; young women in traditional marriages; girls with boyfriends who always use condoms (except when they don’t); and girls with steady boyfriends and sugar daddies on the side.

These projects give a flavour of what social segmentation solutions may look like. The studies have helped our clients to better target their interventions based on the specific needs and drivers of each segment, hence illustrating the value of applying segmentation analysis in a development context.

What is meant by ‘segmentation’ and what should it look like?

Before we move on to the differences and success factors, let’s agree on what is meant by ‘segmentation’. The word segmentation is sometimes used to simply denote splitting a population into sub-categories and presenting analysis by variables such as gender or age group. While this is indeed one type of segmentation, ‘segmentation analysis’ generally refers to sophisticated statistical techniques to segment people based on carefully designed questions and topic areas, and on patterns in the data that are unknown prior to the analysis. Segmentation can be based on a wide range of factors such as socio-demographics, beliefs, attitudes, behaviour, needs, and individual emotional traits. It is this type of segmentation we are concerned with here.

The aim of segmentation analysis is to have segments that are as distinct as possible from each other, while the people within each segment should be as similar as possible. The segments should also be easily identifiable in the population from a practical point of view. Furthermore, a successful segmentation should offer insights, some ‘ah ha!’ experience, and be intuitive enough to strike a chord with the client and stakeholders. If not, the segments are unlikely to gain traction.[1]

How do market and social segmentations differ?

Moving on to the differences between market and social segmentation studies, there are two main differences which I will talk about here.

Firstly, the outcome variables – that is, the factors on which the segmentation is based on – may be less clearly defined in social segmentations than in market ones. While market segmentations generally focus on segmenting the target population on the basis of a single outcome variable and a single behaviour – purchase of a product – social segmentation studies tend to be more complex than that. Social ones often (a) look at multifaceted and socially sensitive behaviours and (b) often try to explain multiple behaviours which each is affected by a different set of drivers and barriers.

As mentioned above, one of the benefits of using segmentation analysis in development is that programmes and interventions can be tailored according to the specific needs and behaviours of the target population. The key outcome variables for a programme may indeed be dependent on the findings from the segmentation analysis. This means that outcome variables may not actually be known or clearly defined at the beginning of a project.

In the context of young women at risk of HIV, there is a multitude of behaviours that lead to increased vulnerability. Risky behaviour may stem from lack of willingness to go out of one’s way to get a condom, lack of confidence to insist on condom use, or the keeping of multiple and/or concurrent boyfriends, to mention but a few. These behaviours, in turn, may be related to opportunities and socio-economic factors. There may also be physical barriers, such as inaccessibility to places providing free condoms, or lack of money to buy them. These factors can all feed into the segments, which subsequently reflect a variety of risk factors and population profiles. The intervention could focus on any one, or more, of these risk factors and drivers.

With complex segmentation studies such as the one of young women at risk of HIV, the analysis is often an iterative exercise where solutions are scrutinised and re-scrutinised as part of the process. In fact, you could say it is a bit like finding a partner or spouse with whom you want to settle down: you might need to meet a few potential partners before you even fully realise what it is you are seeking. Now, some researchers estimated the ideal number of partners to date before settling down is as high as 12![2]

Turning the attention back to segmentation, the multitude of outcome variables and the often complex associations between behaviours, attitudes, and needs further mean that segments produced in social segmentations are unlikely to be as neat as standard market segments.

As for your potential long-term partner, no segmentation solution is perfect. It is thus a matter of deciding what the most important traits are, and focusing on those. Although we may dream of extremely well-differentiated segments, each consisting of highly homogenous groups, we are unlikely to observe such a pattern for the full range of relevant variables. For example, among our young women, social norms and touch points turned out to be less differentiating than behaviour to protect oneself against HIV and also experience of abuse.

On this note, it is worth highlighting the importance of including a sufficient number of behavioural variables in the segmentation. While behavioural variables may not necessarily be more differentiating than attitudinal ones, they tend to have more practical value for identifying the target groups in the population at large. It is therefore important to ensure a sufficient range of relevant behavioural variables are covered.

Success factors

Having talked about segmentation analysis in broad terms, and the main differences between market and social segmentations, we can summarise the learnings for successful social segmentations as follows:

  1. Define as clearly as possible the element(s) (behaviours, attitudes etc.) on which you want the segments to vary, while acknowledging the complexities in social segmentations. Identifying the right segmentation variables is critical for successful segmentations. However, lack of a single outcome variables, and multifaceted relationships between behavioural, attitudinal and demographic variables mean segmentation analysis may involve an iterative process of finding the most suitable solution. It also means that segments may not be as clearly defined as standard market segments.
  2. Make sure the segments are easily identifiable in the population and, if necessary, tilt the balance towards behavioural factors. As for any segmentation, whether in market or social research, it is important that segments are identifiable in the population at large. How will the target groups be reached in practice? Behavioural variables tend to be more useful for this purpose, but this is dependent on the nature of the intervention.
  3. Allow time and resources to find the optimal segmentation solution. Two or three iterations are unlikely to be enough, so it is important to allow sufficient time for analysis. Finding the right segmentation solution is indeed a bit like finding a spouse. None is perfect, and it is only after meeting a few potential partners that one better knows what to settle for.
  4. Align expectations early on since the resulting segments are unlikely to be as neat as standard market segments. In light of the points above, it is important to acknowledge the differences between market and social segmentations, and the expected outputs. Have, and set, the right expectations from the start and segmentation solution will invariably become a smoother exercise.

Social segmentations have immense potential to add value and insight to programme designs, in particular to better understand the needs and drivers across different sub-groups in the target population. Bear in mind the points above, and you will maximise the chances of finding a set of segments that will succeed in making you happy. Perhaps not forever after, but at least until your next programme.

 

[1] I won’t go into the technical details of segmentation here, but it is worth noting that there are several different statistical methods of conducting segmentation analysis. One common analytical approach is Latent Class Analysis (LCA), which for example was used for the HIV related-project. The segmentation analysis is typically used to produces outputs for several different segmentation solutions such as solutions for 3, 4, 5, 6 and 7 segments. When deciding which solution to use, we normally look at the segments based on the segmenting variables and also by cross-tabulating the segments against other variables in the questionnaire. Pen portraits can then be produced of the different segments and to help decide which solution is the most useful ones.

[2] http://www.bbc.co.uk/programmes/p02hl73h

From snoring camels to product diversification: A gendered analysis of internet participation in Ghana, Kenya, Nigeria and South Africa

Written by: Alexandra Cronberg

It is hard to find anything that offers so much hope and potential as increased internet access across Africa. The internet offers a whole new world of information, ideas, tools, and ways of connecting people as well as providing sources of entertainment and distractions, certainly with silly kittens and camels galore. Importantly, it offers revolutionising ways of accessing and delivering services, including vital ones such as finance. Recent discussions with jua kali, or informal sector producers in Kenya, showed enormous potential to diversify their product lines provided they had access to and knowledge of the internet. Enabling people at the bottom of the pyramid, who currently have little or limited internet access, to make use of all of this will be life changing.

Or so we like to think. In reality, the picture is more complex. While internet access itself may be binary, just like the data it holds, the users are intricate, inconsistent and often contradictory human beings. Indeed, internet participation cannot be reduced to zeros and ones. A paper by Kantar Public, presented at the African ITS Conference in Accra in March 2016, sheds light on the complexity of internet engagement and the factors that underpin it. The paper, authored by Nicola Marsh, is based on analysis of a global annual study of internet use conducted by Kantar TNS in a wide range of countries[1]. This particular piece of analysis focuses on Ghana, Kenya, Nigeria and South Africa.

Gender is a key part of this picture. Fewer women than men use the internet in most African countries, and these four countries are no exception. By way of example, 19% of men in Ghana have access to the internet, whereas the figure for women is a measly 9%. In South Africa, which has the highest level of internet access among the four countries, 41% of men use the internet whereas only 29% of women do so[2]. Consequently the door to the digital world remains shut for many women.

Figure 1. Internet access by country and sex, 2012

figure-2

Source: Research In Africa, 2012

The KP paper analysed different levels of internet engagement and factors that underpin different types of usage. First, an overall “internet participation” composite score was created based on a bunch of common online activities and their frequency. The findings show that greater access for women, or indeed disadvantaged men, does not imply online engagement. In fact, the countries with higher levels of access tend to have lower levels of participation. Within the countries, men consistently have higher levels of engagement than women.

Figure 2. Mean score of internet participation by country and sex, 2015

figure-2

Source: Kantar TNS Connected Life Survey, 2015

Second, this overall score was then broken down into three main factors or categories of usage, capturing some of the nuance of internet engagement. The categories are:

  • Popular activities. This includes instant messaging, social networking, uploading photos, playing games, reading news/sports/weather.
  • Sophisticated activities. This includes mobile payments[3], streaming/downloading shows/movies, streaming music/radio, watching videos, internet banking
  • Text heavy activities. This includes blogging, visiting blogs/forums, and emails.

The gender gap is further highlighted when looking at these different categories of internet usage, with sophisticated activities having the greatest gap.

Other factors in addition to gender that lead to greater internet participation overall are younger age, better education, and higher socio-economic group. However, different life stages, defined as student status, marital status and having kids, have no consistent impact on online participation across the four countries.

Lower education and social class have less of an impact on the popular online activities. If we want to get women and people who are less well educated to participate more, the starting point should arguably therefore be data light services.

These findings show that as online participation increases and people lower down the pyramid gain access, proportionately more people engage with the internet in lighter ways. Women are often among those who are late to join the online party. Indeed, across the four countries the gender gap for internet participation is inversely related to the level of internet access.  For example, in South Africa a more similar proportion of men and women access the internet, but among those who are online, women have a lower level of participation than men.  In contrast, in Ghana where the gender gap in access is large, the men and women who do have access have more similar levels of engagement.

In sum, this analysis makes it clear that for the internet to be a truly useful tool for disadvantaged groups of people, much more ought to be done to get women in particular to develop more technical skills and online literacy, as well as solving other affordability and access issues. If not, many of the most vulnerable people will remain excluded from the digital possibilities including access to services, information, networks and ideas. While a few tentative steps online might mean people tumble into Facebook and other social networks, it is essential they don’t just get sucked into the whirlpool of singing dogs, snoring camels and other people’s dinner from which they may or may not emerge. Rather, people need to engage with more sophisticated online activities if they are to click their way onwards and upwards. A snoring camel ain’t gonna help with that.

The full version of the paper is available on request.

[1] The analysis was based on the data from the annual, multi-country survey conducted by Kantar TNS, called “Connected Life”. The survey covers technology and internet behaviours amongst internet users. All those interviewed use the internet at least once a week, and the sample for each country is weighted to be nationally representative of weekly internet users aged 16+. The data was collected between June and August 2015.

[2] Source: Research In Africa, Gillwald et al (2012), http://www.researchictafrica.net

[3] Note that in Kenya mobile payments are commonly done using Mpesa, but the level of penetration of mobile money is much lower in other countries.