I have a scale for how much measurements matter for each category of products:
DACs: 100%
Amplifiers (headphone and speaker): 80 to 90% due to variability of available power. Hard to internalize how much power is available/enough without listening tests.
Speakers: 70 to 80%
Headphones: 50 to 80% (measurements too variable)
@amirm, this is a great answer and almost exactly what I tell people.
The underlying issue is that some classes of products have distortion mechanisms that are difficult to test. These tend to stem from the system being stressed to their limits, which is common in transducers (speakers and mics). In a DAC, you’re operating at 3V maybe, with minimal current requirements. If you do the simple stuff right, it’s pretty easy to hit good specs, and the performance isn’t affected by particular signals. Measurements almost perfectly correlate to real world performance. I trust the AP way more than my ears in this case.
Amps can get stressed with high current and voltage swings. Headphone outputs are often overlooked by the pro industry, and often underperform at low and high impedance loads. Hitting current limitation of an amp is clipping and sounds terrible. And it can be easy to hit if all channels are driven hard. Measurements still rule here, but it makes sense to listen for issues that may pop up with multi-channel signal dependent current draw.
Speakers, for me, are one of the most challenging to correlate measurements to enjoyment, but I still measure everything. It is the first cut. If measurements are terrible, the speaker won’t sound good. Full stop. If the measurements are mediocre, but not terrible, it could still outperform a better measuring speaker in practice, so it stays in the running and gets listened to. If the measurements are perfect, that’s a real good start, but not a guarantee. Take a 5” 2-way design. We measure that at reasonable level, say 86dB, like Amir. It might measure great, but it’s unlikely that you won’t see transient peaks at higher SPL level even if you listen at a quiet “slow” SPL level. And what happens to the midrange if that 5” driver is really moving due to loud low frequency content? The midrange doesn’t stay clean. It distorts like crazy. Not an easy thing to measure because you’d have to measure IMD at all frequency combinations, at all mixed power levels, to see a complete picture. And I’m not even sure how you’d visualize that in a graph.
So going to that JBL that measured mediocre, but sounded good. It’s a large diameter woofer with a compression tweeter. The woofer doesn’t have to work as hard to produce a given output volume, and the tweeter isn’t breaking a sweat.
In my personal experience, I’d weight speaker measurements roughly like this:
Axial Frequency Response: 30%
Polar Response: 20%
Intermod Distortion at 100dB: 30%
THD@100dB: 20%
And those last two, especially IMD, are rarely measured, and even more rarely at that level. So while that JBL may only be mediocre in the top two, it’s likely crushing the bottom two. And with that weighting, I’d likely prefer that JBL over a speaker that was excellent in the top two, but poor in the bottom two.