Understanding Your Customer with Empathy and Data

Note: this is the post accompanying a talk I gave at WordCamp Asia in 2023. If you are interested in seeing the slides themselves, check out my slides from the event or just the empathy map worksheet.


If you’re trying to reach customers with a product you have or a service you have, you probably want to be able to make data informed decisions. Making decisions using data can help us be better at understanding a broad set of customer needs without having to interview every single individual customer. The basic idea being that every business that wants to effectively scale up the number of customers they have needs, needs to be able to scale up their ability to listen to their customers. If you want to keep tabs on customer satisfaction, on how effective your product is for customers, or anything else, and you are growing quickly, one of the best ways to do that is to figure out what data points to measure.

If your method of listening to customers doesn’t scale, then the more you scale, the harder it will be for you to hear your customers.

Many people think of using data around customer satisfaction and customer behavior data as a simple thing. People think you can just look at how customer behave draw seemingly obvious conclusions based on the data, let that inform design, and you’re done – easy right? Well it might sound easy, and sometimes it is, but usually there’s a lot of hidden problem that are difficult to identify unless you understand the data in context of how it fits to real people. Here’s an example of what I mean:

1950s the US Air Force was trying to improve the way their aircrafts would handle and wanted to answer the question “have pilots on average changed size since the first world war”. Lots of problems with this approach. Even the question itself contains lots of bias: average pilot. As it turns out, there is sort of no such thing.

Anthropologist Gilbert Daniels was hired to determine dimensions of pilots and found that there actually is no such thing as an average pilot – literally. Out of 4,063 pilots he examined, not a single airman fit within the average range on all 10 dimensions, and less than 3.5% of pilots would be average sized on three key dimensions he chose. In other words, if you make a cockpit for an “average pilot” you will make a cockpit that fits literally zero pilots. It sounds like a paradox. But this is now a often cited case study in the problem of using averages.

The solution they found was actually to completely change the approach. Instead of trying to make an average cockpit and find average sized humans in all relevant dimensions (which doesn’t exist), the Air Force instead determined they needed to find a way to make cockpits customizable for each pilot. They needed adjustable seats, adjustable helmets, etc. Cars still use a lot of those same designs for adjustable seats a half century later.

Getting back to the main topic here, that’s just an example of how using data can product better and unexpected results. No one went into that thinking they were going to disprove the entire methodology. They thought they would confirm or deny if the average pilot was getting bigger or not. But when faced with new information, the strategy was adjusted and a solution was found. This is actually a data success story.


Here’s another data success story from Ancestry.com

UX+PM=BFFs by Josh Penrod and Kendall Hulet at Front 2017

Here’s a lighter story about how spaghetti sauce customer satisfaction data did not fit the mold, and how prego was able to rapidly change their strategy by understanding how the data fit the customer.

Choice, happiness and spaghetti sauce | Malcolm Gladwell

It’s a fun story, and the point it illustrates is true. If you want to understand customers, just observing data isn’t enough. Just trying to make a bell graph and guessing something in the middle is right, isn’t actually right. You have to understand how it fits to customers, and sometimes it doesn’t, so the model has to change.

Further reading:

Read more: