@johncutlefish's blog

Check out Iterate.fm, my podcast with friend and coworker Tareq

Making OKRs Easier…

Published: June 16, 2019

1 dpzntykSX7iohPMBwaln3Q

I frequently meet teams struggling with OKRs.

When I dig deeper, I often find that the team is having to “reinvent the wheel” each and every quarter. There’s a mad rush in the final days of the quarter to fine-tune objectives, “come up with” key results, and make sure it all “ties together up and down the organization” (while simultaneously doing actual work). In short: painful and disruptive, or as an engineering manager once described to me “planning masochism”.

The culprit is that teams are often trying to dream up results for “projects” instead of forecasting their impact on longer-standing missions, or they are part of an elaborate multi-leveled “cascade” of quarterly goals. Both of these patterns are a result of treating quarters as lengthy organizational delivery sprints vs. a cadence of introspection. Organizational delivery sprints impose way too much conformity onto teams.

What is the alternative? The alternative is to model your product strategy (or [insert team here] strategy) such that it has more longevity, and persists for at least a couple quarters. For example, at my current job at Amplitude Analytics we have a North Star of Weekly Learning Users, with a number of key inputs (e.g. Broadcasted Learnings, Activated Accounts, etc.). The framework is relatively stable and is reasonably solution agnostic, in addition to being a good encapsulation of our product strategy. This leaves teams free to “plug in” to a flattish model without a big quarterly upheaval. Yes, teams need to forecast the impact of their bets, but they aren’t starting from scratch in terms of figuring out how it fits into the bigger picture.

So…if you are finding OKRs to be a big hassle, consider how you have communicated and modeled your product strategy. The passionate problem solvers on the team-level will appreciate the rigor and clarity of a well-modeled (and flat) model.