Everyone who has run OKRs knows the five failure modes. The sandbagged target, set low enough to be certain. The vanity metric, chosen because it only goes up. The orphan objective, owned by nobody by the second month. The quarterly lie, told on a Friday afternoon. And the grading conversation, which is not a grading conversation at all but a negotiation with the person who will decide your bonus.
Organisations attack these as five problems and buy five remedies. Training for the sandbagging. A metrics taxonomy for the vanity. An RACI for the orphans. A better cadence for the lie. A calibration committee for the grading.
They are one problem.
In OKR, the goal and the work live in different systems. The objective lives in a planning tool. The work lives in the CRM, the ledger, the repository, the ERP, the warehouse system. Between them sits a human being who, at some point in the period, types a number from one into the other.
Every failure mode lives in that gap. Sandbagging is possible because a human sets the target in a system with no memory of what was achievable. The vanity metric survives because nothing downstream depends on it being the right measure. The orphan objective persists because the planning tool has no idea whether anyone touched the work. The quarterly lie is simply what happens when a number is typed by the person it judges.
Here is the part that matters. For fifty years this was not a design flaw, it was a constraint. Computing a goal from the systems of record was genuinely impossible for most organisations — the data was not connected, and the capacity to act on it was not elastic. So a planning layer with typed numbers was the best available arrangement, and the failure modes were the cost of it.
That constraint is gone. The numbers exist. They are connected. The capacity to act on them is elastic and cheap. Which means the quarterly ritual is no longer a constraint. It is a choice, and it is now a fairly expensive one.
What replaces the objective is a mission with four fields. Intent — what we are actually trying to cause. Measure — read from a system of record, never typed. Level — the threshold, set against what the measure has historically done rather than what someone hoped. Route — what happens, automatically, with a named owner and a governed approval, when the level is missed.
Three of those fields most organisations can do today. The fourth is where systems fail, and it is worth being precise about why: Route requires you to decide in advance what happens on the downside, which is a much harder political act than setting a target. A target is a wish. A route is a commitment made while you still have the option not to make it.
Ask any goal you hold right now. Where does the number come from? Who decides when it moves? What happens next?
Almost everyone fails the third.