OKR examples from Google's early years
Andy Grove brought OKRs to Intel; John Doerr brought them to Google in 1999. A look at the OKRs that shaped Google's culture and what they teach about writing your own.
OKRs at Google are the canonical case study — partly because John Doerr documented them, partly because Google's growth made them the most-cited operating system in tech. The examples below are the ones Doerr describes in Measure What Matters and that have been confirmed publicly by Google executives over the years. They're useful both as templates and as warnings.
OKR 1 — Sundar Pichai on Chrome (2008)
Objective: Build the best browser
Key Results:
- Reach 20 million 7-day-active users by end of 2008
- Reach 50 million 7-day-active users by end of 2009
- Reach 100 million 7-day-active users by end of 2010
What's right about this: a single, sharp metric (7-day-active users) repeated three times with escalating targets. Easy to track, hard to game. The target was explicitly ambitious — at the time Chrome had zero users.
The team missed 2008. They hit 2009. They beat 2010 (~120M).
Lesson: missing year 1 of a stretch OKR isn't failure — it's information. Google didn't fire the team. They re-set for the next quarter and continued.
OKR 2 — YouTube on watch-time (2012)
Objective: Reach 1 billion hours of daily watch-time
Key Results:
- Improve the recommendation algorithm by [specific metric]
- Reduce video load time to under [N] seconds
- Expand video catalog by partnering with [N] new creators per quarter
What's right: the objective is a single bold number. The key results are specific actions the team could control. The team didn't promise watch-time directly (a lagging metric outside their full control); they promised the inputs that would drive it.
YouTube hit 1B daily hours in 2016.
Lesson: when an objective is a lagging metric (revenue, retention, watch-time), the key results should be the leading metrics you control.
OKR 3 — Larry Page on a stalled project (~2008)
This one is famous for what it warned against. Larry was reviewing a struggling team's quarterly OKRs and noticed:
- Objective: "Maintain product quality"
- Key Results: "Reach 99.5% uptime", "Resolve 90% of support tickets within 24h"
His response, paraphrased: "This isn't an OKR. This is operational floor. You should be doing all of these every quarter. What's the change we're going to make?"
The team had quietly converted their quarterly bet into routine KPIs. The objective was vague; the key results were ongoing standards. Larry sent them back to write real OKRs.
Lesson: this is the most common OKR failure mode. An OKR has to commit to change, not to maintenance. If at the end of the quarter the world looks the same, the OKR wasn't an OKR — it was a status report.
OKR 4 — Google Search ranking team (early 2010s)
Objective: Make search results meaningfully better for ambiguous queries
Key Results:
- Improve click-through on top result by 5% (measured: yes / no)
- Reduce abandoned searches by 10% (measured: yes / no)
- Launch ranking improvements based on [specific signal] (measured: yes / no)
What's right: each key result is binary — at quarter-end you either hit the number or you didn't. No fuzzy "made progress" judgments. The team either improved CTR by 5% or they didn't.
Lesson: well-formed key results have a single test of success that's defensible without explanation. "Improve X" is not a key result. "Improve X to Y by [date]" is.
What these have in common
Looking across all four:
- Objectives are 1 sentence, qualitative, ambitious. "Build the best browser." Not "Improve Chrome's market share by 12% via the Q3 features roadmap and the Q4 marketing campaign while maintaining team velocity."
- Key Results are quantitative and binary. A real KR can be checked off as hit / missed without interpretation.
- 3–5 KRs per Objective. Fewer and the bet is too narrow; more and the focus dissolves.
- Ambition produces ~70% achievement. Google's culture explicitly rewarded missed-but-stretched OKRs over hit-but-conservative ones. Teams that hit 100% were suspected of sandbagging.
What Google got wrong (and learned)
Doerr admits OKRs at Google went through several wrong turns:
- Year 1 (1999): too many OKRs per team. Teams were drowning in 12+ objectives per quarter. They cut to 3–5.
- Years 2–3: managers used OKRs as performance reviews. This destroyed the willingness to stretch — if missing your OKR meant a bad review, you'd write easy OKRs. They explicitly decoupled OKRs from promotion criteria.
- Years 4+: too much bureaucracy around grading OKRs. They simplified to a 0.0–1.0 score per KR, calculated once per quarter, with a simple readout.
The version Doerr describes in Measure What Matters is the one that survived 25 years of iteration. Other companies adopting OKRs would do well to skip the wrong turns and start with the simplified version.
Run your own
The full method, including how OKRs differ from KPIs, is in the OKRs vs KPIs Academy guide →. The OKR catalog entry has more templates. Start a canvas to draft your own.