Insurtech, remote testing and the future of voice

This week we'll go through some of our findings from testing voice skills from home, our thoughts on the future of voice after the Voice Global event and 3 things we've found interesting in insurance recently.

Have a great week and enjoy,

Hi Mum! Said Dad


What we've learnt from remote testing

As Amazon featured voice skill builders, we do a lot of work on voice. As a platform with streams of new users, use cases and capabilities, it's crucial to do good user testing and research on voice. Best practices or 'established' patterns cannot yet be taken as given - they'll evolve and update over the coming years, in the same way that today's mobile apps don't look the same as they did back on the iPhone 4 era. During the lockdown, we needed to make sure that our processes to define and design voice experiences didn't suffer, but now working remotely. Given that in-person testing was out of the window, there were three main options:

1. Stop Doing User Testing

It is an option to go without user testing, at least at first. If a skill is simple enough, or has best practices that have been properly proven out and explored already, then forgoing user testing to get a skill into the market quickly can work, as long as you give yourself room to iterate on the skill later, whether through an analytics strategy that can identify improvement areas, or through user feedback in reviews. A fair number of our skills have worked like this, or started out like this and then brought in testing later once the demand had proven itself.

Where this doesn't work is if your improvement areas are likely to involve big design/development efforts that quickly eat up the earlier savings on user testing. For example, a skill based around a famous voice or brand with lots of professionally-voiced content is very hard to adjust without making lots of wholesale changes - testing up front to make sure the flows make sense saves on having to go back and re-record later - measure twice, cut once.

2. Remote Unmoderated Testing

Remote unmoderated testing allows skill builders to put prototypes or live skills in front of users without needing a facilitator or moderator present to guide them through the experience or ask them questions. Companies like Userzoom let skill developers set up tasks and surveys to get users to go through. The advantages here are speed and ease- we can set up a test and leave it running effectively on autopilot, and then react to the data all at once.

For web and mobile products, this can work at almost any stage, particularly for usability tests where you'll be testing very specific flows that are around understanding rather than appeal e.g. "Do first-time users understand how to set up an account?". A lot of the testing we do, however, is around appeal, which often requires quite a bit more polish to measure accurately (especially if you're expecting to use custom voiced content) than is achievable the early stages of a project. We'd recommend against unmoderated testing for concept/appeal testing at early stages - its at its best for on the fly continued usability testing and quick experiments.

3. Remote Moderated Testing

Remote moderated testing means having one or more members of the team in the session to moderate - interviewing the participant to get deeper context and understanding of their potential usage of a product, and allowing us to probe deeper on bits that are relevant for an individual session, rather than overwhelming every participant with an overly-lengthy survey.

For voice in particular, having a moderator is perhaps most helpful when testing early stage prototypes that you're expecting to iterate on quickly. It lets us explain how to use a prototype with an incomplete user experience, probe on throwaway comments like "Hmmm, that's cool" that often belie some of the most powerful insights and get a true sense of how users interacted with a product beyond the limited scope of task completion rates and survey rating scales.

Voice testing can be difficult remotely - we've been fortunate that the skills we've been working on testing lately have mostly had fairly simple, predominantly linear paths (reducing the likelihood of having to jump in as a moderator and recover broken user journeys) and short interaction times lending themselves well to stop-start 'demo, then feedback, then repeat' loops for testing sessions. Testing this way does have advantages over in-person testing that mean we're likely to recommend it in the future in some cases - having a handful of webcam-off observers in a zoom call is much less intimidating to participants than being outnumbered in person or in front of a 1-way mirror, and testing days have generally run smoother than in-person with fewer dropouts and no traffic-driven late arrivals, and niche user groups are far more accessible if you don't need them to be able to attend physically.


The future of voice is for everyone

VOICE Global, a self-proclaimed 24-hour livestream at the forefront of conversational development, design, and artificial intelligence platforms, was a virtual event held last month with ambitious plans to live stream content on the topic of voice technology for an entire day… and it almost achieved those lofty heights, were it not for some technical issues.

Check out some thoughts and reactions on human centred voice design, accessibility and expressing brand personality through voice from our design team on on the full blog post.


3 bits of insurance tech we're into right now


Proactively engaging with technology presents an opportunity for the insurance industry to reconnect with customers and to re-emphasise their role as protectors rather than as black boxes of admin tasks and frustration. We're seeing a lot of interesting stuff happening, from huge names to tiny, here's a few bits of innovation we really like at the moment.

Big brands doing tech well

State Farm's Drive Safe and Save feature echoes the functionality of tracker box insurance without the need for a manually installed black box to the car - letting users get discounts for safer driving, and crucially feedback on each trip a driver makes. There are some qualms around one-size-fits-all approaches to rating the safety of driving, perhaps unsurprising when 73% of US drivers think they are better than average.

This takes the monitoring and long term incentives of lower premiums that offerings like InsureTheBox in the UK achieve through a physically installed device; dramatically lowers the effort and time costs of getting set up,  and offers users direct feedback and scores down to individual drives immediately after drives - whereas box insurance drivers have typically had to look through clunky, slow-to-update web portals to find out what their driving is doing.

Unicorns on the horizon

Lemonade's model of taking a flat fee, paying out quickly and generally trying to be a 'feel good' insurer to do business with is laudable, and their surging recent IPO suggests markets believe in them too. One of the things we've always liked is their focus on UX from the first impression; in an industry that so commonly is plagued by beige, unintuitive form pages, their light aesthetic and measured chatbot-form-hybrid approach makes everything much easier.

By asking the questions in simple, highly visual pages, they reduce cognitive load for users but also  lending credibility to the quote you eventually get as users know what factors it is considering.

New and niche markets served well

It's not just large incumbents and mega-funded challengers that are using and responding to tech well. Companies like Flock (Drones) and Zego (Private hire e.g. uber drivers) are showing some of the opportunities to capitalise on specialist markets not just in offering the insurance and adding an option to their forms, but developing a a whole user experience around that makes it clear that they're the place to go for it.

Flock in particular has its own UX for a bunch of features and utilities that are unique to drone pilots, down to single flight risk assessments based on location, rainfall probability, windspeed etc. to give customers the power to really understand what's going into that insurance model and price.