Segmentation & Targeting
The Considered_ approach to behavioural innovation Part 02
In the latest piece in this series, we focus on the second element of the Considered approach to behavioural innovation — segmentation and targeting: ‘Who’ are we designing for?
Did you miss Part 01? You can find it here👇
What is segmentation?
Segmentation involves dividing the population or community into sub-groups based on factors that they have in common. There are several ways in which a population can be segmented:
- Geographic
- Demographic
- Attitudinal
- Needs/Assets-based
- Psychographic tec
However, given that we are focussed on a behavioural approach to innovation, we are primarily interested in building segmentation models based on behaviours*.
Why is it important for Behavioural Science?
Using segmentation in this way allows you to do three things:
- Optimise resource: target activity to maximise impact with available budget.
- Maximise effectiveness: tailor intervention to the specific needs of particular groups.
- Promote equality of outcome: ensure that sub-populations aren’t over- or under-served**
Practical application & examples
The following example is taken from a project focussed on missed rent payments.
The horizontal axis represents the number of times a customer has missed a payment, whilst the vertical axis represents how long it took for the organisation to recover the debt. These serve as measures of Frequency and Severity respectively.
Within these two dimensions, we can see how distinct patterns of behaviour emerge:
- Segment 1 for instance doesn’t appear on the chart as these customers have never missed a payment.
- Segment 7 on the other hand have missed payments on several occasions (high frequency) and have taken a considerable period to repay the debt (high severity)
- Segment 2 rarely miss a payment (low frequency) and if they do, they catch up quickly (low severity).
- Segment 3 regularly miss payments (high frequency) but repay their debt quickly when they do (low severity).
Target & Tailor
Immediately we have a powerful tool for targeting and tailoring intervention: an appropriate and effective intervention for customers in Segment 7 should clearly be different from an intervention designed for customers in Segment 1 or 2, given the extent to which their behavioural patterns and level of risk differ.
In very basic terms:
- Segment 2 may only need an automated text message reminder to prompt payment.
- Segment 7 may require a face-to-face visit and a hand-delivered letter.
- Segment 3 may respond to information about how many times they’ve missed a payment compared with their fellow citizens.
In this way, we tailor the design of each intervention to the specific behavioural pattern (maximise effectiveness) and align the intensity of intervention with the level of risk presented (optimise resource).*** In doing so, we work harder to assist those in greater need, whilst still making provision for those in less need (promote equality of outcome).
Using this approach, one of clients saved £1/4 million in recovery costs by reducing the amount intervention delivered to low-risk customers.
Simple, yet powerful
To underline how simple yet powerful behavioural segmentation can be, consider the following example taken from a project focussed on missed appointments:
- Repeat compliers: attended all appointments.
- Repeat offenders: have missed at least 4 consecutive appointments.
- Waverers: a mix of failed and successful appointments.
If an appointment comes due for a customer in the Repeat Offender segment, why would you send them the same letter as a customer in the Repeat Complier segment? We’ve worked with social housing providers and local authorities across the UK and all of them do!
Challenges & Limitations
Whilst the logic of behavioural segmentation is simple, the practical execution can be complex. Firstly, it relies on the relevant data being available. Ideally, you’ll be using existing data that is collected by operational or adminstrative systems as a matter of course. Depending on the target behaviours this could include:
- CRM systems
- Payment / recovery systems
- Appointment management software
- Service interactions and outcomes
If the relevant data isn’t available then undertaking primary research could be an option — normally large-scale surveys with representative samples. However, not only is this time- and cost-prohibitive for many projects, it also relies on self-reported behavioural data rather than more accurate observational data. On the positive side, the primary research can be used to collect attitudinal data as well as demographic and behavioural to introduce further nuance into the model.
In most cases segmentation models are based on data snapshots and are therefore descriptive tool based on historical behaviours. This means that:
- They can quickly become out-of-date and therefore inaccurate.
- They can’t respond to future changes in behaviour.
- They aren’t designed or equipped to be used as a predictive tool.
It’s also important to remember that targeting based on group characteristics raises a host of ethical questions. This is another reason to focus on the use of behavioural segmentation specifically — rather than demogrpaphic — and to target activity based on those behaviours rather than other factors that may be statistically associated with those behaviours.
One-size-fits-all interventions waste resource and exacerbate inequalities. Here we have seen how behavioural segmentation creates a powerful framework to overcome these limitations and empower designers to respond to the specific needs of various population subgroups.
Please clap 👏👏 if you find this post helpful. Thanks!To be continued…
Within the Considered_ approach to behavioural innovation, segmentation goes hand-in-hand with qualitative insight — understanding the Why behind the What and the Who. This will provide the focus for the third piece in this series: Behavioural Drivers — ‘Why’ are/aren’t people doing what we’d like them to? Watch this space!
Footnotes
* Once the main behavioural segments have been identified, we would normally undertake further analysis to identify any demographic or other behaviour factors that are more likely to be associated with each pattern. It’s possible that some behaviours are more/less associated with different demographic factors, in which case intervention can be targeted and tailored more specifically.
** The Considered approach to behavioural innovation was largely forged working on public health programmes across the UK. In this domain it has been repeatedly shown that one-size-fits-all (untargeted / untailored) interventions disproportionately benefit those members of the population who are already more empowered, comfortably resourced and more pre-disposed to undertaking the target behaviour. This results in the widening of health inequalities.
*** As we’ll see in Part 03 of this series, the use of qualitative research goes hand-in-hand with behavioural segmentation to deepen understanding of behavioural drivers and inform intervention design.
About Steven Johnson
Steven is a writer, speaker and freelance consultant specialising in the practical application of behavioural science.
Over the last decade he has been responsible for leading insight and design on a vast range of award-winning behaviour change projects addressing health, sustainability and social housing issues.
Operating across the UK, Europe and North America his work has received recognition from Cannes Lions, D&AD, IPA, AHC, How-Do and the German Design Council.
He served 3 years as an NED with global advertising body D&AD, working with the leadership team to galvanise the global creative community as a force for positive change in relation to sustainability, health and inequality.
When not working Steven focuses on the infinitely more daunting behaviour change challenge presented by his two young daughters. He is married to Helen and lives in the Ribble Valley, Lancashire.
