How to Wizard of Oz Your Application

Wizard of Oz prototyping is the process of making something manual appear automatic. The term comes from the novel and film, where the wizard turns out to be a man behind a curtain.

So what does this have to do with software? A big component to Agile and lean development is to minimize your build to learn ratio. That is, to build as little as possible to learn as much as you can.

This can be easier said than done, so let’s look at some real-world examples.

If I build it, will they come?

I attended a great talk by Angelique Rickhoff the other day where she discussed UX research at Solstice Mobile.

Angelique was tasked with designing an internal app that gave salespeople information on clients right before a meeting. At this point there were two unknowns: Would people use the service and what content was valuable?

Her first step was talking to potential users, but there’s always a difference between what people say and what they actually do. So Angelique’s team didn’t build anything. Yet. Before each meeting, they manually scraped information to compose their messages.

These messages contained links to things like “recent tweets” and the “latest earnings report.” They could have easily embedded this content, but there was a reason they linked to it: so they could track what was clicked.

By analyzing email open rates and what content was actually viewed, Angelique and her team determined what was and wasn’t worth building.

Ideas to get you thinking

Bringing data into your app

Integrating with third-party APIs and parsing various formats is incredibly time consuming. And you might not know what users need support for.

Instead, have users upload their data (XLS, CSV) and tell them their data is being processed while you work to manually create items (users, products, whatever) in the background. This helps you prioritize what to actually build support for.

Testing an algorithm

Take the dating app Coffee Meets Bagel, where users are matched with one potential date each day. I’m sure they have an automated system now, but I’d be surprised if they didn’t start by reading profiles and matching users by hand.

Even if your algorithm is too complicated to fake, running the code yourself let’s you do a sniff test on the results before showing them to the user.

What else?

Have any Wizard of Oz examples of your own? I’d love to hear them.