← Writing

Wizard-of-Oz Experiments: Shipping Learning Without Shipping Code

Wizard-of-Oz Experiments: Shipping Learning Without Shipping Code

Wizard-of-Oz means the user experiences the product as if it’s automated, but behind the scenes it’s manual.

Why it’s powerful: it lets you validate value and workflow before investing in full engineering.

Great use cases:

  • “AI” features where quality needs real-world validation
  • workflows that depend on integrations and data access
  • new automations where trust and control are key
  • complex internal tools

How to run it well:

1) Make the user experience real

Users should interact with something that feels like the product:

  • a form, a UI, a chatbot, a dashboard Avoid “email me and I’ll do it.” That tests a different behavior.

2) Keep scope narrow

Pick one job and one cohort:

  • one report type
  • one segment
  • one integration The goal is learning, not coverage.

3) Measure the right thing

You’re not measuring “delight.” You’re measuring:

  • would users rely on this?
  • does it fit their workflow?
  • what do they need to trust it?
  • where does it break?

4) Document requirements as learnings

The output is:

  • what users expected
  • what they didn’t trust
  • what edge cases mattered
  • what “control” features are required (preview, undo, approval)

Wizard-of-Oz makes you look fast and strategic, because you’re using engineering time responsibly.

Interview line:

“When the value is uncertain but potentially high, I’ll use a Wizard-of-Oz to validate workflow fit and trust before we build. It reduces wasted effort and increases launch confidence.”