Data and design
DYNAMIC UX BUILT ON CONTEXT AND BEHAVIOR
Note: This is reposted from Medium as I migrate my tech/product writing.
User experience is important. This is platitude in 2013, but it’s worth noting design wasn’t getting its due in the same way a decade ago. It took multiple incumbents with technological head-starts being disrupted by younger companies with stronger design. The number of different professions under the “UX” umbrella has ballooned in the last 10 years.
So, what is design? Some people say everything is design (Looking at you, Tim Brown). But I think that’s a non-answer: when you frame the debate that way, solving anything becomes “designing a solution.” This is overly general to the point of being ineffective.
In the software and web world, we have two sides: experience design (focused more on user interface elements and products that the end-user directly touches/clicks) and system design (data interfaces, back-end services and architecture).
The future of UX in consumer tech is heavily personalized and context aware interfaces. Dynamic interaction flows that re-organize based on user activity, history and peripheral information (weather, calendar, payments apps, etc). These flows will appear differently to different users at different times. This paradigm shift is what has increased focus on dynamic, interactive prototyping: the notion of a static interface is going away. In its place, we’re creating personalized interfaces that are architected around opportune context (as defined through session metadata, like geolocation and time, as well as historical activity data).
The shift towards dynamic, highly-personalized interfaces means that front-end experience design is more reliant on the system design than in the past. Instead of making a database call just to figure out how to personalize a social feed or product suggestions field, we’ll be relying on user data to figure out what features and content to display, how to organize them, and what to suppress.
Never use the “Discover” tab on Twitter’s mobile app? It won’t retain 25% of bottom menubar real estate. Never click on Spotify links in your Facebook Feed? They’ll be suppressed (this already sort of exists). Love the “Radar” view in Dark Sky? The app will go straight there on launch. Compulsively check the “Who’s Viewed My Profile?” page on LinkedIn’s mobile app? That list will be in your home feed, instead of buried 3 taps away. Increases in back-end efficiency allow for machine learning to re-jigger these algorithms in near real-time, helping preserve discovery and keep the experience fresh and effective.
Companies have been using this kind of logic to organize and populate discovery features for years. I’m suggesting that algorithms of this type will start using data inputs to determine how interface elements are visually organized. Interaction flows will change depending on your interaction history with the application.
This isn’t a bold prediction, it’s already happening. Industry leaders are urging designers to learn to create interactive prototypes and move away from static design. Dave Morin has said “AI is the new UI.” Design and engineering are working in closer partnership (as long as pesky PMs don’t get in the way). We’re making new tools to support this kind of interaction design. With front-ends built with increased reliance on the back-end, good design is going to depend on good data.
Some consumer companies collect huge amounts of relatively unstructured data on users, batch it, and figure out how to make it actionable and leverage it into functionality later. Some companies make inaccurate assumptions about user activity, and act on data incorrectly. Shared accounts are a great example of this: Netflix viewing suggestions and Amazon product suggestions start to get laughably bad. This leads to bad design and will be even more problematic for end users going forward.
Facebook News Feed is a good example. They have done a huge amount of work in personalization. However, they make a poor data assumption: that my social graph (my 1285 Facebook friends) represents my actual social interests and connections. That assumption might work for some users, but for me, it makes the experience of News Feed awful. It’s not engaging and is cumbersome to interact with: I have to sift through all kinds of irrelevant content and have trouble making sense of what appears where, when and why.
Engaging personalization experiences are contingent on the hygiene of your data.
Proliferation of mobile devices has led to a huge increase in behavioral inputs— you’re sending more information on how you interact with software. Advances in database technology, distributed and cloud computing, and RESTful APIs make these insights accessible and actionable at scale, in real-time and across platforms.
Dynamic UX means more data experiments, user testing, and iteration. It means a constantly-shifting layout, like responsive design problems on steroids (look at the bright side, once you implement you won’t have to worry about re-formatting for all the different Android viewports!). It means greater cross-collaborative efforts between groups (design, engineering, PMs, research, data). It also implies what has become a ubiquitous adage: anyone making software needs to have technical literacy and the ability to make data-driven choices. These are exciting challenges, and I think they’ll lead to more delighted users.
Good data yields good design. Good experiences lead to happy users.
Thanks to Gabe Leader-Rose, Sam Cunningham and Albert Nichols for edits.