1. Googler wants a promo. They know they need to "launch" to get promo. They can't get a promo if they just improve or maintain something that already exists.
2. Googler gets together with other googlers and puts together something they could launch. Nobody cares if it makes any sense.
3. The thing launches. Googler gets their promo. It is now pointless to spend any effort on maintaining something that you can't "launch" again. Googler moves on.
4. Project dies and gets shut down after a while because no sane person would touch maintenance work with a 10 foot pole.
9 out of 10 projects at Google go through this lifecycle, pretty much verbatim. This is why you have three different Google-provided messsengers on Android, and why stuff launches with great fanfare and then quickly runs out of steam. You get the behavior you reward. This is what Google rewards.
Well if Google tried rewarding maintenance work, would people who do hard maintenance work and reduce tech debt actually be rewarded? Instead I think the result would be people gaming the new system and doing nothing while claiming to solve hard maintenance problems. Managers, as usual, would reward the wrong people. At least with the current system, google gets something kinda measurable.
Managers are only very tangentially involved with "rewarding" employees at Google. They do not promote them: it's the responsibility of the employee to apply for a promo, though manager has to support. And promotion decision is not made by the manager either: it's made by a committee that has never even seen the employee and knows nothing about their work other than what they choose to put in their "promotion packet", as well as peer feedback. IOW your manager can torpedo your promo chances, but they can't unilaterally "promote" or "reward" you per se, aside from some fractional bump in the yearly perf score.
So employees have to support the assertions they make in the promo packet with evidence. I suppose you could support the assertion that you've done great maintenance work on something that's valuable to the company, and show off the metrics, PRs (or "CL"s in Google terminology), monetary impact, and things of that nature. But Google culturally doesn't really give a shit about this kind of thing, even if it's otherwise impactful. A few people manage to do well maintaining code at scale, but by and large my advice to Nooglers would be to stay as far away as they can from anything that's not contributing to a launch in some _easily quantifiable and attributable_ way. Or even if it does contribute to a launch, stay away from things they personally can't take credit for. Otherwise you'll end up working on something that someone else "started", and that someone will take credit when it's done.
In other places without committees most of the same rules still apply. Instead of the committee, the employee needs to justify the promo to the manager+skip(s) and external assessors from other teams often need to also approve promos.
I find a lot of the most impactful work isn't easily measured by metrics or monetary impact. These metrics are always deceptive and if you are doing honest maintenance you may encounter many additional issues that make you look like a troublemaker. The people who get promoted often cause disasters and then fix the disaster while recording metrics. Putting in hard work to make stuff run as expected looks like doing nothing to people unfamiliar with the low level technical details.
Arguably Google's system is fairer, because the committee doesn't know you, so their decision has to rely on the facts in the packet (and their cursory validation of such facts against other evidence). IOW, you don't get a promo just because you brownnose your boss or skip, or butter up some bosses that you know will be present when the decision is made - they aren't on the committee, and they aren't the ones making the decision.
Even though this system is objectively fairer, it _feels_ shitty nevertheless, I suppose because it's much harder to game through the mechanisms outlined above. You have to game it in other, much more labor-intensive ways, such as "work with" people who are a level or two above you (of which there are very few in remote offices, so you better be in Mountain View for that), sit in important meetings, "show leadership", virtue signal, etc. In fact if you do all of these things really well, you could skip the actual productive work by switching to the management track. If you're not super good at this, you have to do them in addition to your day job, and your promo prospects become tenuous at best beyond level 5.
Regarding metrics, there's a universal law at Google: you can't improve what you don't measure. This is part of the reason why Google sucks at UX for example, and rocks pretty hard at ads and search. The former is not measurable. The latter is.
"You get the behavior you reward. This is what Google rewards." From what I have heard, I think you're right. Promotions and kudos flow from launching new products & services, but not in maintaining them.
Yes, it convinces me there needs to be a 3rd stage to the old axiom: 1) Ideas are easy but 2) execution is hard and (new) 3) Maintenance is boring and lacks prestige but you'll fail without it.
1. Googler wants a promo. They know they need to "launch" to get promo. They can't get a promo if they just improve or maintain something that already exists.
2. Googler gets together with other googlers and puts together something they could launch. Nobody cares if it makes any sense.
3. The thing launches. Googler gets their promo. It is now pointless to spend any effort on maintaining something that you can't "launch" again. Googler moves on.
4. Project dies and gets shut down after a while because no sane person would touch maintenance work with a 10 foot pole.
9 out of 10 projects at Google go through this lifecycle, pretty much verbatim. This is why you have three different Google-provided messsengers on Android, and why stuff launches with great fanfare and then quickly runs out of steam. You get the behavior you reward. This is what Google rewards.