- User Experience
- User Interface
- User Testing
UI for AI
There are lots of big claims about AI, but it’s really just a collection of statistical models and computing techniques. Designing for AI means reconciling these complicated models with the needs of the marketer. At Lytics, our data scientists make predictions about future user behavior based on a ridiculous amount of user data. Marketers get value out of these predictions, but in order to trust them they need to understand the models behind them. That’s where the audience insight report comes in.
The project started with a rough wireframe that essentially gave form to a set of API endpoints. This was an excersise to see what data was already available, and how it might be visualized.
This wireframe served as a conversation starter for gathering more feedback, and what we heard was that many parts of the intial wireframe were too technical. A histogram that showed the “significance” of user attributes didn’t really mean anything to a marketer.
So we abstracted the report a bit futher. “Significance” became an index, so marketers could better see the relative importance of an attribute in a model. I also considered introducing data visualizations that would be immediately faimilar, like a map.
The final design
The final design takes a couple cues from existing Lytics reporting (like the line chart). But more importantly it surfaces actionable, human readable suggestions. These suggestions are generated from the like of “significant” attributes, and they let marketers get value out of this report extremely fast.
Lytics users can invite colleagues to view any insight report. It turns out that this interaction is fairly complex, so I turned to mapping it out as a user flow diagram. These diagrams are one of the best ways to keep everyone on the same page once development begins.
The after party
Internally and externally, the insight report has been well-recieved for bringing API-only features into the web UI. It allows marketers to explore previously arcane data-science, and to take immedaite action on their findings.