One of the readers who preferred to remain anonymous wrote in to ask:
In my team we decided to have a look at the emotional impact accessibility issues have on our customers. But we quickly realised we can't measure this in a good way. We tried to do user interviews and rank emotions on a scale, but it turned out to be too subjective and the data was fairly inconsistent. How can we measure emotions after all?
That's true. There's no good way to measure the frustration, anxiety and exclusion users feel when they hit a roadblock. Maybe the emotional toll leads to them abandoning your product entirely, and you'll never know. This is rarely captured in standard metrics.
But a measurement isn't an exact science. It's one thing to measure your height and waist. You need exact numbers there, otherwise the pants won't fit you. That's a measurement for practical purposes. They need to be exact or they're useless.
Then there's the kind of measurements that are useful in making decisions. Do you invest time and resources to implement this feature or that feature? You'll want some numbers to work with, maybe how many users requested each feature, how much monetary value will each bring in, etc. But do these need to be exact numbers? No, not at all.
For the purposes of making decisions, you can think of a measurement as an observation you can make to quantitatively reduce uncertainty. I first read about this in Douglas Hubbard's How to measure anything. He argued that for a measurement to be valuable, all you need is to have a way to observe a reduction, not necessarily an elimination, of uncertainty. In a sense, measuring a thing is a probabilistic exercise, because certainty is not the goal.
To measure emotions, you can start to make observations, even indirect observations, of what you can see about the problem in such a way that you get somewhere in the ball park of being fairly certain solving it will make a difference. There is a lot of value in quickly estimating emotional turmoil caused by inaccessible features. And if you know something is a big problem, because you observed it as such, it's a good enough measurement to go back to the stakeholders and argue for the need to find a solution.
Think about it. Does it matter if it's a big problem that affects most your customers in some way or if this problem is a 8.4124551 on a scale of 1 to 10 affecting 74.321233% of your user base? Those decimals don't matter. Neither does if it's a 7 or an 8. There's no good way to reduce people and emotions to a number.
The simple fact that it's pervasive enough that you can observe it often should be enough of a measurement. The goal isn't to measure something, but to make an informed decision of where to invest your limited resources.