Fourteen months ago we went live with our first minimal viable product (MVP) – after a fast and furious alpha and beta we were offering real people a very real remote sexual health testing service. Service development director, Chris Howroyd, reflects on how we've optimised the service to support users and clinicians - before and after launch.
The word ‘optimisation’ was mentioned in every other sentence – embedded in our psyche in a passionate bid to improve the experience of both end-users and clinicians alike. Our famous user journey funnel became my god, I virtually bowed down to KissMetrics every morning, hours were spent analysing mini cliff-edges (points along the order journey where users fall out/choose not to proceed).
I had grand plans of immediately running several AB tests and collecting feedback en masse, but ultimately neither happened.
Letting the funnel look after itself*
The six words above will be tantamount to heresy in most people’s eyes, but effectively that’s what we did for the first six months - for two very good reasons. Firstly, we didn’t have any major cliff-edges, and those we did see, we could explain. Secondly, we had more pressing optimisations to consider, which weren’t directly related to the end-user’s initial order journey. And thirdly, quite simply, our founding MVP had immediately been validated and we were keen on pressing on with our second.
*this title is heavily caveated, in so far as it only applies to healthy funnels – within the first month, our order completion rate sat at 97%. It has flat-lined ever since due to some militant protectionism (see lessons). Obviously, if you start with a journey rife with cliff-edges then this approach would simply be ludicrous.
Clinicians are users too
We call our back-end (the secure app-come-portal which stores and manages all user data and result notifications) ‘Admin’. Although its UI is not nearly as sexy as our public facing site, you could argue that the UI and UX of Admin is more important, as one misguided action or error could potentially result in a user receiving the wrong test results.
Working closely with clinicians, we have relentlessly optimised Admin to use data more intelligently, increase its vision and scope, and remove risk (principally that of potential human error).
To date, I would estimate that we have spent over 80% of our time optimising our Admin app, simplifying, integrating (very smartly) and automating functions and features.
Results interfaces with our lab, results management, safeguarding flags, risk assessment flags, and comms with users (about their test kit and sexual health) have all enjoyed a huge amount of optimisation, which translates into a better experience of the end-user.
Protect and build on what you have
The old adage applies; if it ain’t broke, don’t fix it. I have fiercely protected the order journey from feature creep (despite risking upsetting some of my colleagues). Unless a new feature adds clear value to the user (or is clinically essential), then it’s been pushed back. I guarantee the team will be sick of me saying ‘and what does the end-user get out of that?’
Don’t (just) be a slave to conversion
Consider the entire UX, beyond the order journey.
We suffered from users mistakenly/inadvertently selecting the wrong address when choosing their delivery address from a drop down, which resulted in users not receiving their kits and us receiving about four returns a week courtesy of the Royal Mail. We’ve since removed the drop down address menu (which will please the guys at GDS) and ask users to hard write their address in instead. The result, no more undelivered returns (to date).
Be prepared to pivot (and admit that you were wrong)
We’ve tried and failed to connect to a number of other services through APIs which have promised a lot, but have failed to deliver. Shortcuts very rarely result in a net gain, for you or users.
Data is king, but it is always countered by (and benefits from) a dose of common sense
Whilst the likes of Google Analytics and KissMetrics will spit data at you on an hourly basis, resist the urge to act upon it without considering the wider context. On countless occasions, alternative, qualitative insights have persuaded us not to rely on raw quantitative data alone.
Know when to stop and bide your time
Optimising without a respectable understanding of the matter in hand is dangerous and, ultimately, a waste of time. After all, if your assumptions aren’t based on anything other than a hunch, then they’re likely to be wrong.
Don’t be afraid of concierging (again), even if you believe that you’re out of MVP mode
In many ways, reverting back to a concierge (doing things in a very basic, manual way) service feels rather odd having developed what appears to be well defined and ‘operational’ model, but resist the urge to develop add-on features without thorough testing.