Tech Deserves Better than OKRs

As somebody who works in the tech industry, I have a problem: My brain rebels against objectives and key results (OKRs), which have become the dominant way tech companies (including Duolingo) manage performance for teams within the company. If you haven’t worked in an “OKR culture” company, OKRs are an extremely ritualized and formalized way of setting and talking about goals. In the book Measure what Matters, the self-described “Johnny Appleseed of OKRs,” John Doerr, defines them as follows:

An OBJECTIVE, I explained, is simply WHAT is to be achieved, no more and no less. By definition, objectives are significant, concrete, action oriented, and (ideally) inspirational. When properly designed and deployed, they’re a vaccine against fuzzy thinking—and fuzzy execution. KEY RESULTS benchmark and monitor HOW we get to the objective. Effective KRs are specific and time-bound, aggressive yet realistic. Most of all, they are measurable and verifiable. (As prize pupil Marissa Mayer would say, “It’s not a key result unless it has a number.”)

OKR proponents credit this simple process of defining goals and supporting metrics with “driving clarity, accountability, and the uninhibited pursuit of greatness” (Doerr, again). Sounds fantastic! Then why does my brain lock up every time OKR season returns at Duolingo?

My first aggravation is how much time is spent trying to force ideas into the precise format required by OKRs. Doerr writes, approvingly:

Google CEO Sundar Pichai once told me that his team often “agonized” over their goal-setting process: “There are single OKR lines on which you can spend an hour and a half thinking, to make sure we are focused on doing something better for the user.” (Emphasis added)

The verb “agonize” seems right to me. The genre comes with so many rules! Are your objectives “aggressive yet realistic; tangible, objective, and unambiguous”? Make sure each objective fits on one line, otherwise it probably isn’t “crisp” enough. Are the key results “measurable milestones” that describe “outcomes, not activities”? (Confusingly, although you are supposed to describe outcomes and not activities, key results about “inputs” are somehow good…) Do you have the required number of objectives and key results?

The final product of all this agonizing is an admirably concise description of what each team is striving to accomplish. However, because it is so concise, when it comes time to review OKRs, the best that most people can do is figure out if the rules of the form were obeyed. For example, Measure What Matters includes the following table to help people understand weak, average, and strong OKRs:

The table does a good job teaching the reader the rules: As you progress through the columns, the key results become quantifiable (which move them to “Average”) and then include a mix of input and output results (thus getting them to “Strong”). However, how are we supposed to evaluate if these key results actually make sense? Almost nobody reading the book has any idea what it takes to win the Indy 500. The format is so terse that there’s no room to explain why these key results are the right one, so all you can do when reviewing them is see if they follow the rules. Lewis Carroll’s Jabberwocky fits the rules of poetry, but it’s also famously nonsense.¹

OKRs also suffer from misplaced emphasis. They assume that the biggest problem team members face is knowing if they have met their goal, when the reality is that it’s more vital to think through how the team will meet its goal. It’s not as if you can just “manifest” success into existence if you define it precisely enough!

What I hunger for instead of OKRs is strategy. At least, I hunger for “strategy” as defined by the book Good Strategy, Bad Strategy by Richard Rumelt. (This is a popular book among Duolingo managers.) According to Rumelt, every good business strategy has the same “kernel” that consists of three things:

1. A diagnosis of the main problem in front of the business,
2. A guiding policy that defines the overall approach to overcoming the problem, and
3. A set of coherent actions — basically, “what are you going to do?”

OKRs let people gloss over that last bit, and Rumelt thinks that ignoring the “what are we going to do?” question is a common recipe for bad strategy. This passage from the book could have been a subtweet of Measure What Matters:

When the “strategy” process is basically a game of setting performance goals—so much market share and so much profit, so many students graduating high school, so many visitors to the museum—then there remains a yawning gap between these ambitions and action. Strategy is about how an organization will move forward. Doing strategy is figuring out how to advance the organization’s interests.

Circling back to the beginning: I’m not saying that OKRs cause teams to create bad strategy. I am saying, however, that OKRs permit bad strategy. They provide a warm, cozy environment where it can thrive undetected like black mold behind drywall.

I have many theories of varying degrees of cynicism about why, in spite of these problems, the OKR format has conquered the tech industry. I’ll leave you with the least cynical: In an unpredictable world, checklists provide comfort. OKRs are fundamentally about checklists. Measure What Matters essentially gives you a checklist for how to create good OKRs and conduct OKR reviews. Your OKRs themselves, once written, are your checklist for success. Cross off all your key results and you will achieve your objectives! In contrast, Rumelt’s Good Strategy, Bad Strategy not only has no checklists, it actively argues against them: A section of the book warns how “template-style strategy” leads to “pious statements of the obvious presented as if they were decisive insights.” Strategy involves thinking through chains of cause-and-effect and is best explained with narrative. It’s harder to do and harder to reproduce. But if I’m going to spend time “agonizing” over creating something, that seems more valuable and worthwhile.