It's probably more like a year with actual support and updates, two years where it's left to whither and slowly die, and then a year of notice before sunsetting it.

Yeah. By that logic MS has eternal support for everything, because MS has a legendary hardcore approach to backwards compatibility that they will release a product, push it hard, and then abandon development on it... but still have it "officially supported" for a decade while it bitrots into a nightmare... but never tell their users "stop using this, it's deprecated".

As much as I hate Google sunsetting products, I hate even more the companies that keep their zombies shuffling along in a state of undeath forever.

That makes no sense. Nobody is forcing you to use an abandoned product. So you're obviously better off with the Microsoft model than the Google one.

A good example is VB6, the runtime is still supported in Windows 10 -- and necessary for one our largest vendor products to run -- yet the VB6 IDE only runs in XP. Our business would have been really screwed if they just dropped support for it. We are, of course, working to move to a different product and have the luxury of time to do that.

Or worse when an old product kinda pseudo gets merged into a new product, in a way that makes both of them worse. Better just to cancel the old one and move on.

4 years for a product is abysmal from one of the richest companies in the world. This is Google not some garage shop with 2 employees.

I have no idea what the metrics are, but to continue my devil's advocacy, would the counterpoint here be "average lifespan of a marginal product that doesn't get acquired by $BIG_COMPANY"?

> but couldn't this be read with the opposite conclusion as well - that the average time that Google keeps marginal products alive is 4 years

Averages are misleading when the distribution is bi-or-multimodal.

So are you saying that it couldn't be read like that?

Think about how many Google products have lasted much longer than four year: Search, Maps, Gmail, Drive, AdWords, etc. This skews the average.

What would be more helpful is a median, or percentiles (eg 75% of Google products only last X years, 25% last Y months). I imagine that would be much more representative of Google’s fail-fast experimentations.

FTA: The average lifespan of a discontinued Google product is 4 years and 1 month.

"Living" products aren't skewing anything

Id be interested in seeing the median lifespan. I feel like it's possible there are long lived products bringing the average way up, although I could be wrong without diving in further.