What goes in must come out: forgiving Google
1 min read

What goes in must come out: forgiving Google

Mike Dong

Why is Google shutting down so many services that some of us love? Wave, Buzz, Reader, the list goes on. They frequently launch interesting new projects and “get” us in that respect, why are they failing to maintain the ones that we use?

First, a market leader like Google does not double its resources every year anymore. Projects compete for talent and rack space among other inputs. There is a zero-sum game within companies and more so in the short term.

When resources are fixed, closure is the price of innovation.

Google can come up with new experimental products only because they are prepared to retire existing ones. In fact, companies that are more “faithful to their customers” and keep products around longer are less innovative and probably more conservative in the first place.

This goes back to a result in queueing theory called Burke's theorem. I first came across it during a rainy Cambridge morning, where our Applied Probability lecturer remarked on its fascinating simplicity.

In a queue in equilibrium, under some special conditions*, the process of customer arrivals is indistinguishable from the process of customer departures.

Replacing customers with products, for a stable company (factoring out growth), innovation and closure will have matching rates. To be innovative, a company must be prepared to ruthlessly reallocate from underperforming** products.

It is time to support Google, not because we expect them to only release products that they can commit to keeping around, but because of how respectfully they handle closures: with export options and an early heads-up.

Thanks for reading, you should follow me on Twitter here.

* — One model would be an M/M/m queue with Poisson arrivals. Meaning “product ideas” occur randomly at a fixed rate and are launched concurrently with at most m products at a given moment.

** — Under an individualized utility function for each company.