Reading data like a designer
© Art in design isn’t decoration— it’s direction with emotion.
Data as a design material
Most teams treat analytics as something that happens after design — a way to measure whether a finished decision worked, reviewed in a dashboard weeks after launch. We treat it differently: as a material to design with, in the same way typography or color is a material, examined before decisions get made rather than only after.
This shift changes what the first week of a project looks like. Instead of opening a design tool, we open whatever data already exists — funnel drop-off, heatmaps, support ticket themes, search queries typed into a site's internal search bar. These sources are usually underused, not because teams don't have access to them, but because reading them as design input rather than as a reporting exercise requires a different kind of attention.
What drop-off actually tells you
A drop-off point in a funnel is often treated as a simple problem people are leaving at step three, so step three needs to be simplified or shortened. But drop-off rarely has a single cause, and treating it as one usually produces a fix that improves the metric slightly while leaving the underlying confusion untouched.
Pair quantitative drop-off data with qualitative signals from the same step
Review support tickets filed around high-drop-off pages specifically
Watch session recordings before proposing a redesign, not just after
Designing the follow-up test, not just the fix
When data reveals a problem clearly enough to propose a fix, the temptation is to ship the fix and move to the next issue. But a fix based on a hypothesis is still a hypothesis until it's been checked against real behavior again, and skipping that second look means a team can spend months optimizing based on an assumption that was only ever partially correct.
We build the follow-up check into the same project as the fix, not as a separate phase that gets deprioritized later. This usually means defining, before anything ships, exactly what number should move and by roughly how much for the fix to be considered successful a small amount of upfront discipline that makes the eventual result far easier to actually evaluate.