The COM-B model for digital intervention design
Digital intervention design and the COM-B model as an analytic framework
Table of contents
In one of our recent articles, we considered the role of behavioural science in the design process. What if you’ve built a digital intervention and you want to better understand whether interaction with that intervention tells you something about your target audience’s behaviour? It’s to help answer that question that we look at the COM-B model. We want to understand if it’s a good candidate as a comprehensive analytical framework for digital intervention design.
Why would the COM-B model be a good candidate as an analytic framework for digital intervention measurement?
When we talk about the COM-B model we’re talking about the constituents of behaviour defined as:
- Capability – one’s ability to complete a task.
- Opportunity – whether the environment and resources available to the individual are sufficient for them to complete the task.
- Motivation – are they motivated to complete the task, either because of rewards, habits or an intrinsic sense of value associated with the outcome.
For this to work as a more complete model we want to use the six constituent sub-components of the COM-B model that are used in the more extensive literature. This is to give us more comprehensive insight and because interventions in the real world rarely, if ever, act on one component exclusively.
The model and it’s sub-components we consider are:
- Physical capability – the physical skills, strength, cognitive capacity to complete a task
- Psychological capability – the reasoning skills, information processing capacity and other mental processes required to complete a task
- Social opportunity – the norms, rules and social ‘environmental’ factors that enable someone to complete the task or see the task completed near them.
- Physical opportunity – is the lived environment the person completes the task in, organised in such a way as to make the task possible.
- Reflective motivation- intrinsic motivation components such as a belief in a goal or that an outcome task ends positively.
- Automatic motivation – the behaviour is rewarded or can be factored into a habit that the user follows without “thinking”
As you can see from this breakdown the COM-B model covers many elements of human behaviour and in a systematic way – that’s useful for us because we’re talking about digital interventions and computation is systematic. We should be aware that there can be complications resulting from overlap between the components of the model – for example whether an action is driven more by someone’s psychological capability and their reflective motivation can be difficult to unpick – is someone unable or unmotivated to do a complicated task can pose a challenge. To reduce the risk of this sensible identification of which component it being addressed can reduce the problems that might arise.
We think, therefore, that COM-B is a good candidate due to its systematic nature. How then can we consider it as an analytic framework?
For purposes of this article we’ll consider that B = p(RM) + p(AM) + p(SO) + p(PO) + p(PsyC) + p(PhyC)
Where ‘p’ is the proportion of the digital intervention designed to address the specific component of the behaviour and B is the task or action expected from the individual as a result of the intervention. (A side note, the summative components never add up to 1 because the components of behaviour can never be provided perfectly and a task is never fully completed because of the presence of these components). Therefore, we consider that the value of a digital intervention is in how substantially it can address as many components of behaviour as possible while retaining logical coherence as an intervention.
Why is that important?
Logical coherence is important because in design you have to make non-behavioural decisions. The colours used in a poster design will have an effect on how a message is received or perceived. For some users, even users in your target audience who you intend to help take action, may be put off by them, yet you still need to make that decision.
This is what we mean by logical coherence: the set of design choices, taken as a whole, that make up your intervention. Fundamentally these cannot address every component of a behavioural target. They must necessarily exclude some people or you may never produce anything, or you may end up producing something that is too confusing or multi-faceted, or too nuanced – leaving it incomprehensible by the target audience as you try to address all behavioural components at once without design.
So if we agree at this point that COM-B makes up a good candidate as an analytic framework of some potential, what next?
The goal then is to make it digestible in a software format. One model we can use for that is the Object, Verb mantra of many analytic services used by Object relational databases on the web. This is a useful analogue, it’s been used successfully for years and years. It also maps neatly onto the COM-B model which if considered as “Behaviour = Object” and “Intervention = Verb” allows us to consider that potentially each component of the COM-B model is an object that has the potential for a verb assignment.
Where ObjectName is the component of a web page, email link or activity the user is engaged in. It could be a post on a forum, the click of a card with a specific piece of information on it, or the completion of a set task in an app.
‘rm’ is reflective motivation – let’s say it’s an action that would only be completed if someone had the required level of intentional motivation towards the action. Let’s say it’s completing a self-analysis in the app itself.
‘interacted’ is of course an identifier for the form the engagement action took, this could be clicking on a link, emailing a support ticket, or any other digital ‘action’ that could take place.
So now we can see that COM-B model can be developed into a tracking model for software uses, how might it scale?
As you can glean from the description so far, at the individual level you are able to gather interesting user insights by observing the degree to which engagement with parts of your system is accounted for by the COM-B components interacted with. In an app example you might be able to observe that users were more likely to engage in the new behaviour if they showed the reflective motivation to interact with some content, while also reading information provided by you to them, thus developing their psychological capability.
The fact that this is intuitive and observable at the individual level should allow it to scale up to larger audiences, you may for example be able to determine whether or not different groups require different additive elements of the intervention’s design in order to be more likely to complete the behaviour. This information may get further insight once demographic, or social group signifiers are accommodated for in the analytic model.
Of course it’s very unlikely that categorically causative relationships will be able to be observed. However that doesn’t limit your ability to draw useful inferences, especially where those inferences allow you to simplify the intervention down to it’s minimally useful scope, reducing cognitive load and reducing the risk of sludge or noise.
We hope you’ve found this analysis insightful, to find out more about how you can apply it to any behaviour or people change programme you’re conceiving please do get in touch.
Data-driven behaviour design: how to uncover the real impact of your learning efforts Measuring something is the key to understanding its success. But are you measuring the things that really matter? Can you see the
KPIs, Personal Development Plans and employee objectives are all forms of…
Best at Digital is delighted that Amex GBT have been recognised for its outstanding work on compliance, being awarded Compliance Program of the Year at this year’s Excellence in Compliance Awards. A well deserved win