Customer interviews are messy, and that’s okay.

My favorite customer interviewing framework I learned from Josh Porter with Nelson Joyce and the UX Sisters. It’s a condensed and simplified version of the Scientific Method. The main idea is to decompose a simple “choice” into three buckets:

• Observations
• Inferences
• Conclusions

What were the things the customer saw (observations) which led them to believe something (inferences) and then to act on it (conclusion)?

If you can find patterns of observations that lead to an inference, you can better predict what will lead to a desired inference. If you can find patterns in inferences that lead to a conclusion, you can better predict what will lead to a desired conclusion. The stronger the patterns you find, the more confident you can be in your predicted approach.

This isn’t necessarily about figuring out what people want. It’s about being more predictive about what a larger potential audience might do based on your current audience’s behavior.

Keyword: predictive. Not speculative. That’s a different thing for a different day.

One way or another

There’s typically two ways you can go about doing these interviews. Reverse engineering from the conclusion, or building into the conclusion. Each has their own benefits and drawbacks.

Conclusions > Inferences > Observations

This approach feels safe because the conclusions are typically the easiest things to identify. Then it’s relatively straightforward to ask variations of “why did you do that?” and “when did you do that?” and “how did you do that?” to break the conclusions down into related inferences and observations.

Unfortunately, this approach gets harder and harder the further away you get from the conclusion. People don’t usually spend their energy evaluating and capturing their inferences and observations. So most of the time they aren’t able to give specific answers to specific questions. An enormous amount of time can be soaked up asking “why?” from a bunch of different angles to break down a single point. This is normal. It happens to everyone. It’s not because you don’t have a good enough question, sometimes there just isn’t a clean answer.

Two things to remember with this approach:

• People can burn out quickly being asked a bunch of questions they don’t have immediate answers for, so break it up with chit-chat and jokes if you notice any fatigue
• Pushing into more specificity can reveal uncomfortable biases and embarrassing stereotypes, so be ready to let a line of questioning go if you think it might go there

Observations > Inferences > Conclusions

This approach feels more difficult because it’s impossible to know ahead of time if an observation will eventually tie back to a relevant conclusion. You will go down several paths that lead to a conclusion you weren’t intending to investigate. Don’t get discouraged. You’re starting from the least obvious place and it requires a little faith.

Fortunately, it gets easier and easier as you go. Once you have a few relevant observations and inferences to a conclusion, they become helpful reference points to anchor other lines of questioning. Each productive line of questioning, makes the next one more effective. With practice you get better at spotting dead ends finding new lines of questioning.

Two things to remember with this approach:

• It can seem invasive and rude if you don’t give people the necessary context for extremely specific questions, so give regular reminders why you’re questions are so granular
• Switching lines of questioning before a reaching a conclusion can be disorienting, so make sure you work transitional chit-chat into each new line of questioning

It’s messy. That’s okay.

Once you’re effective at conclusion-to-observation and observation-to-conclusion lines of questioning, play with them together. When one direction starts to feel unproductive, try finding it from the other direction. Often times you won’t be able to connect them, but it might lead into another productive line of questioning.

This can leave open a lot of loose ends.

Just find a pattern or two.

Make a mess.

It’s okay.