I've seen it attributed to Peter Drucker as well.
I think the general notion that it is more difficult to improve things that you can't/don't measure is fairly accurate.
To be somewhat pedantic:
The difficulty of improving a thing does not change based on you measuring it. If I, say, measure the response time of a service, it's still just as difficult as it was before to improve the response time of the service. Furthermore, any improvement you do make will not change based on you measuring it. If I do improve the response time of the service, it will be just as improved whether or not I measure the change.
So, measuring does nothing but allow you to quantify the improvement you have made. It doesn't make improving a thing easier, and it doesn't make improvements better.
That said, there's plenty of scenarios where impartial measurements are critical. If you don't know what to improve, broad measurement across the entire system can provide an answer. If there's indecisiveness about which action will provide the best improvement, then impartial measuring will provide an answer. If you don't trust your perception, or cannot perceive the result, then impartial measuring can also provide an answer.
Measurements will usually encourage attention and effort to change in a way that non-measured things are unlikely to.
On that basis, I'd argue that even things that are very difficult to measure quantitatively would benefit if a suitable proxy was found that could be measured.
Of course, you do have to be careful that your proxy is suitable!
That said, it helps me keep budgets more. If I don't check mint periodically through the month, making sure I'm on track for goals, then I have a tendency to overspend.
Qualitative doesn't imply unmeasurable.
If something is unmeasurable, there is no way to tell if it's changed.
If you can tell if something has changed, and you have no feelings about it, you can measure it, but not optimize it, because you can't force it towards your preference.
If you can tell if something has changed, and you know how you feel about it, you can always just say "I feel better about this since last time" or "I feel worse about this since last time." There you go, you can graph how you think of the thing over time, and steadily increase how much you like it.
But maybe you're right in the area of emotions and feelings. I think humans know inherently what is optimal when it comes to social situations, but we're still trying to codify exactly what that is. For instance, we know that people work better with sunlight and plants in the office. We could do an informal measurement of sorts simply by asking around, do you like this?
But ultimately I think that's what they are doing here, trying to figure out what works and what doesn't.
Does anyone else think the phrase "you can't improve what you can't measure" is nonsense? I think that most improvements are qualitative and inherently unmeasurable.
I'm not downplaying the importance of measuring, but we have to be careful when trying to reduce qualitative changes down to metrics. Too many times I've seen businesses or teams try to measure qualitative changes, and then spend so much time optimizing for their metrics that they qualitatively regress. Measurements are not magic bullets that remove the need for good intuition.