Designing discovery for data products

Why discovery stalls in data projects

Most discovery rituals over-index on dashboards and tables. Stakeholders walk away with a wishlist, not a narrative. By the time engineers touch the backlog, context is already stale.

Anchor discovery on decisions, not deliverables

Gather the decisions every stakeholder needs to make. Map each decision to the evidence required. Instead of “build dashboard X,” the backlog reads “give finance a two-hour lead on claims exceptions.”

Facilitate interviews with evidence boards

I interview operators with an evidence board open. As soon as a quote or metric matters, it lands on the board with metadata. That board lives on, feeding the delivery team and the canvas embed that ships with the final project.

Sprint cadence that respects depth

I run discovery in two-week loops:

Day 1: Revisit the decision inventory. What changed? Who needs to be heard next?
Day 3: Evidence synthesis. Pull in telemetry and qualitative feedback.
Day 6: Prototype or schema sketch. Share with design/dev.
Day 10: Playback with stakeholders. Adjust bets and backlog.

What good looks like

Discovery finishes when every decision has a playbook: the data model, the automation, the explanation. Anything else is still guesswork.