Optimisation, choice and the art of data science
We data folk are always optimising for something. Our code, the return on an investment, warehouse stock levels, drug dosages, whatever. In many cases though, optimisation isn’t just a singular thing that can be easily arrived at.
Take Google Maps for instance. Maps will provide me with directions to a specific destination using the most optimal [fastest] route. But is that what we always want, the fastest route?
Many times I’ve found Google directing me down the UK’s numerous narrow, twisty-turny roads, some barely wider than our car in order to reach our destination in the most optimal [fastest] manner. This sort of driving can be incredibly stressful if you’re not in the mood for it as many of these roads have 60mph speed limits and extremely poor forward visibility.
What if I, the lowly user, would prefer my journey be optimised by some other measure? “Easiest drive”, “most fuel efficient” and “most scenic” spring to mind as alternative ways to optimise a route, but these are not options I’m presented with.
How about, when companies build data products like this, they put the user first and think about what sort of optimisations they might want and not just limit us to the most obvious one?
Of course, I understand that optimising a route for speed is exactly what a lot — probably a majority — of users want. However, mechanically optimising for one thing like this throws a lot of other considerations out of the window. In my specific example of the UK’s country roads, the additional cognitive load of driving these routes can put an undue burden on the driver, particularly if it’s only for a modest time saving over a much simpler route.
Perhaps instead I’d prefer the journey were 5 minutes longer, but was less stressful and only included larger roads. Or maybe I’d prefer to spend an extra 10 minutes and get to drive through beautiful scenery, past ancient ruins and other wondrous spectacles. In either case, these are not options that are currently presented to me.
So, the next time you’re building a data product and you’re “optimising” your user’s experiences in some way, you should be cognisant of what the user actually needs and how they’ll experience your optimisation. Maybe even go a few steps further and offer your users some agency by providing a selection of different optimisations for them to choose from or customise. Your users will thank you for it in the long run. If you’re a data scientist working on a product like this, speak up! This is where your role crosses into over into the art of data science and where you’ll find the ability to truly delight your users.