Intelligent Prompts Research

This is a KAI Social Research contract intended to be let through the Research and Insights DPS.

HMRC has previously used digital prompts to 'pop-up' messages to infuence customer behaviour in digital filing journeys with the aim to reduce errors made by customers submitting returns. The work involves presenting nudges to taxpayers as they make submissions, highlighting potential specific mistakes, general potential problems, and directing them to relevant guidance/support. New technology now allows for personalised 'intelligent' prompts, drawing on wider HMRC data sources which should be more impactful than basic prompts. Previous research supports the value of basic prompts, but we lack detailed understanding of the 'premium' and customer experience likely to be generated by intelligent prompts.

HMRC's Digital Prompts (DP) Programme trialled some personalised examples for customers submitting their 20/21 returns. While these trials have provided preliminary quantitative data to begin to understand whether personalised prompts are effective, several limitatons during the trial(s) have prompted the need for qualitative research to hypothetically explore the extent to which inaccurate personalisation of messaging undermines trust between the taxpayer and HMRC. Qualitative research is required to draw out the comparison between blunt (basic) prompts and intelligent ones to explore whether the positive behavioural changes is from an intelligent prompt compares to a blunt (basic) prompt. Intelligent prompts are generally five-times more expensive to deisgn and build, and so understanding how much more valuable (if at all), the intelligent prompt might be, will have a big influence on the fudning needed and the scale of the application of the different options within a fixed budget.

The original research design included eye-tracking with taxpayers completing a portion of the live tax return followed immediately by in-depth interviews by a 50/50 split - 50% exposed to a digital intelligent prompt and 50% exposed to a blunt (basic) prompt. However, due to issues surrounding the hiring of a suitably qualified contractor with eye-tracking expertise and sample feasibility that has arisen from existing intelligent prompt trials and subsequent quantitative research/evaluation, this approach will utilise hypothetical questions with taxpayers. Presenting examples of blunt (basic) prompts and intelligent prompts, therefore making the obtainability of the sample more feasible and representing value for money.