OKRs were designed inside companies with thousands of employees and mature measurement. Startups adopt them from a blog post, and the result is predictable: five objectives for a fifteen-person team, key results that are actually a task list, a spreadsheet nobody opens after week three, and a retro where everyone agrees to "do OKRs better next quarter." I've watched this cycle from inside, and as a founder I've run the fixed version at my own companies.
What startup-sized OKRs look like
- Three objectives at most. A startup is already one big objective, and quarterly OKRs decompose the current chapter of it. At WisOwl AI, a quarter's objective might be as blunt as "prove recruiters will pay", and everything else is subordinate.
- Key results are outcomes you can't fake by being busy. "Ship the referral flow" is a task. "Referred users are 15% of weekly signups" is a key result. The test: could you complete it and still have failed? If shipping it while nothing improves counts as done, it's a task.
- Scoring is a conversation, not a spreadsheet ceremony. A 30-minute monthly check: what did the numbers do, what did we learn, does the objective still deserve the quarter? That cadence survives startup chaos; weekly scoring rituals don't.
- Sandbagging and moonshotting both get named. Hitting 100% of a safe target and 20% of a fantasy target are the same failure: the team learned nothing about its own capacity.
How I run OKR planning
A planning engagement is short by design: a working day with the leadership team to set the quarter's objective and key results, instrumentation checks so every KR has a live number behind it, and a monthly cadence I facilitate for the first quarter until it runs itself. I bring the outside spine that's hard to supply from inside the team: the willingness to call a task a task, and to cut objective number four.
Over ten years I've operated at both extremes: ~$50M+ ARR discipline at CaaStle and founder scrappiness at Medzin and WisOwl. Startup OKRs need to sit between the two.