Measure What Matters - John Doerr

This book is not the book I thought it was. For some reason, I though it was a book on how to quantify things you want to track. I think that would be an awesome book to have. I like quantifying things. There's a backlash against overquantification - the important things are hard to measure, and things that are easy to measure get watched even if they aren't helpful. And numbers never tell the whole story. And that's before we get to the fact that focusing on a metric leads to it being gamed. On the other hand, this sounds like a good reason to get good at building quantitative metrics - working out how to measure the right thing, control its abuse, and look at the results in context.

Quantitative measurements of progress to a goal push you to be honest with yourself. If you miss, you might have words to justify the miss, but at least you're clear that it's a miss, rather than papering over with some hand-waving. Sticking a number on it helps with coordination and discussion, not unlike SLOs.

As such, it's a disappointment that this book is not about that!

What it is about, though, is how OKRs are great. "Objectives and Key Results" are a way of planning your company/team/personal goals. The big whats of what you want to do, with measurable milestones. Yes, the quantitative element is in this scheme, it's just the book doesn't tell you how to do it. It's a big-picture book. It's got all the usual bits - how OKRs enforce discipline by making you highlight a small set of things you really need to get done over a shopping list you won't complete, the red/amber/green status tracking, etc. It's not new to me, but it's nice to see it laid out straightforwardly.

The reason it's not new to me is because OKRs are the bread-and-butter of Google. While the system was assembled by Andy Grove at Intel (see High Output Management) it was taken on with gusto at Google. Indeed, the book includes an appendix on "How OKRs are done at Google" that I recognised word-for-word from an internal document!

Yet... internally, it's not that simple. In an SRE team, priorities can (and do) change at the drop of a hat, and there's a large operational burden besides "building cool new things". SRE teams have wide scope, looking after large systems that a correspondingly-sized dev team would only develop a small fraction of. As such, there are a lot of plates to spin, and a lot of partners to coordinate with. OKRs have a tendency to turn into a shopping list of the here and now. So, it's nice to have a reminder of how we aspire for them to be!

John Doerr is the reason Google has OKRs. He's been a bit of an OKR disciple, having learnt about them under Grove at Intel, and then spreading them across the industry as a VC at Kleiner Perkins. He knows everyone. The book has praise from Bill Gates, Al Gore and Sheryl Sandberg, a foreword from Larry Page and a chapter from Bono!

OKRs can be explained briefly (the Google OKR description is a short appendix), so the book fleshes it out from a number of angles. There's the history of OKRs at Intel and beyond, descriptions of the major things OKRs provide (focus and commitment to priorities, aligning and connecting teams, providing accountability and encouraging stretch goals), combined with illustrative examples across the industry. Many are big names (Sundar Pichai talking about Chrome), but there are chapters focusing on start-ups, too (presumably to justify that OKRs can be used at all scales). The book's a few years old, so it's interesting to see where those companies have gone, like modest success for MyFitnessPal and, well, failure for Zume pizza. Overall, though, it's an impressive exercise in appeal to authority. ;)

I think in some ways, this is a particularly interesting aspect of the whole thing. OKRs are in danger of being a methodology. The OKR approach came from the management genius of Andy Grove. He constructed it to suit his needs. He thought hard and generated the system. Applying any system without thinking is dangerous. OKRs are applied across the industry in a way that looks surprisingly uniform, in part by appeal to the fact they're being applied everywhere else. Are they being applied thoughtfully?

Despite the eye-rolling at management fads, it does feel like OKRs are a pretty stable and useful approach, certainly compared to the fads that occur on the engineering side! Perhaps OKRs aren't supremely optimal, but they're pretty solid.

As an aside, the book's dedicated to "Coach" Bill Campbell, of Trillion Dollar Coach. Books like this remind me of how small Silicon Valley really is at that level, with Intel, Google, Apple and a bunch of VCs at the core of this tight clique. I can't tell if this dynamic is good because of the companies it created in the group, or bad because of the companies denied opportunities by being outside.

This sense of ambiguity comes back in an example in the book showing that OKRs can be used for good or evil. Early on, it's explained how OKRs drove Intel's response to Motorola's 68K, a coordinated sales and marketing assault to push the 8086 ahead of Motorola's technically superior 68000. And that's why we have the x86 dominated world now, with all the inefficiencies that brings. Maybe OKRs are a net force for evil after all. :p

I wanted to end the review there, but I feel I should say a little more about the book. It's well-written and the examples are pretty interesting. It's in the "spin it out" school of management books on topics that could be kept short, but the way it's done means I don't really begrudge that. At this point it's old news for a decent fraction of the tech world, but if you're not up on OKRs it's a pretty good read. I'm still looking for a book on how to quantify difficult metrics, though.

Posted 2022-09-09.