Using behavioural science to inform ESFA content decisions.

We’ve worked with the Cabinet Office’s Behavioural Insights Team to help employers use their levy funds and make better decisions.

We don’t overuse motivational language as this can be interpreted as patronising or inappropriate1. But we are trialing informative messages at particular times, and we’re measuring their impact.

Using loss aversion

Levy funds are no longer available after 24 months if levy-paying employers don't spend them. We use the term 'sunsetting' to describe this because people tend to be highly loss-averse.

We make these losses explicit by labelling sunsetting funds as ‘Funds you could lose’. And we use the call-to-action ‘Use funds now’ in relation to these expiring funds to encourage the use of these funds. We make these messages more effective by being precise about when funds expire - so we say, for example, ‘in July 2017’ rather than ‘in 2 months’.

To encourage desirable behaviour we should make calls-to-action salient, easy, and timely2. We show these sunsetting messages when users are checking their finances and making forecasts. We make the messages very visible, and aim to make the processs of adding new apprentices simple and easy.

Overcoming the endowment effect

People give more value to the things they already have - this is known as the endowment effect3. There’s potential for users to see their apprenticeship account as though it was a bank account - they may see accumulating funds to be a good thing, and become reluctant to spend their funds.

To overcome this instinct, we try to reframe the spending of funds as a gain to the employer. So, we show the employer ‘Co-investment you've gained’ and ‘Incentives you've gained’ to make it clear that when employers invest in apprentices it’s a gain for them, rather than a loss.

Overcoming the equilibrium effect

One possible behaviour in the apprenticeship service is for the employer to aim towards spending all their levy, but not to spend any more than they have available to them.

With the ultimate goal of the apprenticeship service being to encourage the long-term take-up of apprentices, we are looking at ways to show users the benefits of spending more than just their levy funds, and encouraging them to go into co-investment.

We are looking at showing timely messages in the forecasting tool to remind employers that government will pay 90% of all additional apprentices. Again, this reframes the hiring of apprentices as a gain.

Humans are notoriously bad at forecasting. They tend to be overconfident and overweigh small probabilities4. We’re designing the apprenticeship service forecasting tool with these behavioural tendencies in mind, to help employers make better decisions.

Reducing cognitive load

Cognitive load is the amount of thought you need to exercise to complete a specific task. In the apprenticeship service there are a number of processes going on at any given time and this information can overwhelm users.

Two forms of cognitive load are ‘intrinsic’ and ‘germane’.

Intrinsic cognitive load refers to the information in a particular piece of information or instruction. To reduce this, we chunk content up into easy-to-understand messages, which are easier for employers to act on. For example, the Welcome Wizard guides employers through 4 straight forward steps to set up their account.

Germane cognitive load refers to patterns of thought. People find it easier to recognise and learn something new if they can discern a pattern within it. In the apprenticeship service, we show vast amounts of information in our activity feed, but by using a formulaic and consistent syntax in the messages, users are quickly able to identify which information to pay attention to, and which bits they can ignore.

Reducing choice paralysis

Most people’s working memory is around 4 pieces of information. If we show more than this it can lead to them paying less attention. It’s therefore important we prioritise information for users.

We deliberately only show 4 tasks in the alerts list on the apprenticeship service homepage. This is a manageable number of options to read and process. Also, we present an even number of options because odd numbers of options tend to create a bias towards the 'middle' option of the list.

Creating desirable friction

In certain areas of the apprenticeship service we need to tell users important information, without this information they might struggle to create accounts or spend funds on apprentices. However, at best people only read 28% of the content on a screen5.

In fact, ESFA user research has showed that users often don't pay attention to many of the messages on screen - they simply click green buttons and get as far as they can through the site until they get stuck.

To tackle this we make users slow down by putting important information in an unusual format, for example a numbered list or bullets, against an unusually coloured background. This friction or ‘disfluency’ causes users to slow down and pay attention to the information.


Users tend to pay more attention to personalised messages - the messages become self-relevant and users are motivated to invest time in them. Personalised messages can also increase the perceived competence of our users.6.

However, some users may take offence to personalised messages, especially if they sound overly familiar. So, we’ve decided to avoid using first names in the service for now. But it’s something we’ll test going forward.

Social comparisons

We haven’t yet done any work using social comparisons. But in theory, comparing employers to other companies in their sector could be a positive nudge to encourage them to take on apprentices.

For example, when employers look to hiring apprentices, we could show messages such as “85% of companies in your sector hired more than 20 apprentices this quarter”.

The urge to conform is a powerful human tendency. And in a commercial context, the compulsion to gain advantage may play a role here.

Humanising the data

Currently in the forecasting designs, we only calculate the potential gains of spending funds in terms of the government funds that the employers gain. But we could look at humanising the spending of funds.

For example, in the forecasting tool we could show employers how many apprentices they’d be able to hire based on certain spending decisions.


(1) Black, A. and K. Stanbridge, (2012), ‘Documents as ‘critical incidents’ in organization to consumer communication’., Visible Language, 46, 3, pp. 246– 281

(2) Nudge principles that can encourage desirable behaviours See the Behavioural Insights Team’s 2014 report EAST: Four simple ways to apply behavioural insights, available at simple-ways- to-apply- behavioural-insights/
People are more likely to do stuff if it’s easy to do. 2 There is a wide range of evidence on this point. On default pension contributions, see Madrian, B., and D. Shea ;The power of suggestion." Quarterly Journal of Economics (2001): 18-116. On simplifying applications for tuition financial assistance, see Bettinger, Eric P., et al. "The role of application assistance and information in college decisions: Results from the H&R Block FAFSA experiment.; The Quarterly Journal of Economics 127.3 (2012): 1205-1242. On giving hospital patients clear instructions to prevent re-admission, see Jack, Brian W., et al. &;A reengineered hospital discharge program to decrease rehospitalization: a randomized trial." Annals of internal medicine 150.3 (2009): 178-187.
Make it salient 3 Relevant evidence: on the general principle of using salience, see Dolan, P., Hallsworth, M., Halpern, D., King, D.,&; Vlaev, I. (2010). “MINDSPACE: Influencing behaviour through public policy” Institute for Government and Cabinet Office. On using post-it notes to enhance survey response rates, see Irish Revenue (2013). “Survey of Small and Medium Sized Business Customers” publications/business-survey- 2013.pdf. On making nutritional coding systems effective, see Mehta, R., &; Zhu, R. J. (2009). Blue or red? Exploring the effect of color on cognitive task performances. Science, 323(5918), 1226-1229.
People are more likely to do something if you prompt them at the right time 4 Relevant evidence: on getting people to wash their hands, Judah, G., Aunger, R., Schmidt, W.-P., Michie, S., Granger, S. &; Curtis, V. (2009). Experimental pretesting of handwashing interventions in a natural setting. American Journal of Public Health, 99(S2), S405–S411. On increasing payment for court fines, Haynes, L., Green, D. P., Gallagher, R., John., O. &; Torgerson, D.J. (2013). Collection of delinquent fines: An adaptive randomized trial to access the effectiveness of alternative text messages On getting people to make honesty declarations on forms, Shu, L. L., Mazar, N., Gino, F., Ariely, D., &; Bazerman, M. H. (2012). Signing at the beginning makes ethics salient and decreases dishonest self-reports in comparison to signing at the end. Proceedings of the National Academy of Sciences, 109(38), 15197-15200.

(3) People ascribe more value to things they already possess 4 Morewedge, Carey K.; Giblin, Colleen E. (2015). "Explanations of the endowment effect: an integrative review". Trends in Cognitive Sciences. 19 (6): 339–348.

(4) Humans are notoriously bad at forecasting. Over confident and overweigh small probabilities. See i.a. Haran, Uriel, and Don A. Moore. "A better way to forecast." California Management Review 57.1 (2014): 5-15. Makridakis, Spyros, Robin M. Hogarth, and Anil Gaba. "Why forecasts fail. What to do instead." MIT Sloan Management Review 51.2 (2010): 83-90. According to mental accounting, people may inappropriately divide the same pot of money into different categories.
Thaler, Richard H. (1985). "Mental Accounting and Consumer Choice". Marketing Science. 4 (3): 199–214.

(5) Harald Weinreich, Hartmut Obendorf, Eelco Herder, and Matthias Mayer: "Not Quite the Average: An Empirical Study of Web Use," in the ACM Transactions on the Web, vol. 2, no. 1 (February 2008), article #5.

(6) Personalisation and scarcity of time can influence behaviour. See the Behavioural Insights Team’s EAST framework,