Before launch, give samples of the product to engineers in various markets, who will use the product in different ways, and who will have different expectations of how it should work and what is important to them. Get their feedback, treat it seriously, and make changes as necessary.
For example, today I'm using a Rigol DG4000 series function generator, which has a frequency counter feature. I'm using that feature to measure the output from a board I'm testing. I want the measurement to be as accurate as possible, so I'm using a 10 second gate time and an accurate external 10 MHz reference clock.
There's no denying that the Rigol "works", but several features have already annoyed me today:
- If the external clock source fails, the Rigol switches back to using an internal reference. When the external clock comes back, it keeps using the internal reference. Result: any minor interruption in the external clock means all subsequent measurements are made using the internal clock instead, unless I happen to notice the tiny "Ext" indicator has gone out, and I manually change settings to re-enable the external clock. This is a pain.
- With a gate time of 10 seconds, the display is only updated every 20 seconds. There is no need for this; it's perfectly possible to update the display after every gate interval.
- If the output from my board stops switching, the frequency measurement doesn't ever go to zero; instead, it holds whatever the previous displayed value was. This is just plain wrong.
- The displayed frequency has a ridiculous number of figures. Right now, for example, it says the output from my board is 250.000 252 737 kHz, but over 10 seconds there are only 2,500,000 pulses to count. The last 5 digits cannot be meaningful, so why are they shown?
I've discovered all these quirks in just a few hours' worth of use. None of them would be hard to fix, but they weren't. Someone at Rigol decided the instrument worked well enough, and released it this way.
I understand that products can have bugs in them, but I only ever expect to see them if I'm doing something unusual. Anything I can find in the first few hours' worth of use should have been picked up, and fixed, before launch. More obscure bugs should be fully documented in the change log associated with the firmware update that fixes them.
Ask yourself, "would a Keysight product work this way?". If not, consider a fix.