OKR vs KPI: the difference and how to run both in one company

The difference is not in the formula but in the purpose. A KPI is a metric with a threshold, held for years and owned by one person. An OKR is a pair of an objective in words and three to five measurable results, rewritten every quarter and deliberately set above what the team can comfortably reach.

The two came out of different traditions. KPIs grew out of management reporting and settled into their present shape after the Balanced Scorecard by Robert Kaplan and David Norton (Harvard Business Review, 1992): four perspectives, dozens of metrics, a regular review. OKRs grew out of the system Andy Grove ran at Intel in the 1970s and went mainstream after John Doerr brought them to Google in 1999.

That difference has a practical consequence. Scoring a metric you hold inside limits at 0.7 makes no sense: it is either within the band or outside it. The reverse holds too. An objective where 0.7 counts as success cannot feed a bonus, because the team will lower the bar in the first quarter.

The same number can be a KPI or a key result. “Defect rate 2%” sits on the plant manager’s dashboard for years, and that is a KPI. Worded as “cut from 2% to 0.8% by the end of the quarter”, it becomes a key result. The metric did not change. What the team does with it did.

Two things decide it: what you are doing with the metric, and how many functions have to touch it. Holding it inside limits with one person accountable gives you a KPI. Moving it, where no single department can do it alone, gives you an OKR. The two middle cells close without quarterly goals at all.

The “move it, one owner” cell is where fake OKRs come from most often. A manager wants the work to be visible and writes it up as a quarterly goal. What comes out is an objective with a single executor that nobody discusses at check-ins, because there is nobody to discuss it with.

The “hold it, several functions” cell looks like an OKR and is not one. Delivery time that depends on the warehouse, logistics and support at once stays a service level metric: the norm is in the agreement, and while the teams hold it, a quarterly goal adds nothing. It appears in the quarter someone decides to move that norm.

A KPI has no quarter. The number gets taken weekly or monthly for years, and the threshold is revisited when the process itself changes. An OKR is cut into periods: every quarter ends with a score and a retrospective, and the next set of objectives is written from scratch.

This is where the most common organisational mistake comes from: the team gets quarterly goals while the review stays annual. An objective with no check-in every week or two turns by March into a list of intentions nobody has opened since January. The mirror image happens as well. A KPI the whole team picks apart every week eats exactly the time the objectives needed.

The sequence takes one working day per team. First you write down what already exists, then you mark what you are moving, and only then do you write objectives. Doing it the other way round, objectives first and a dashboard reconciliation later, produces duplication you discover at grading, when it is too late to change anything.

1. 01Write out every metric the team already reports, with its owner, threshold and data source. Expect between eight and thirty lines, half of which nobody opens.
2. 02Cross out the ones nobody acted on in the past two quarters. A metric that prompted no decision is not a KPI, it is a reporting habit.
3. 03Mark the ones you are genuinely moving this quarter. Do not mark more than two or three: a team carries two fronts at once, and the third gets no attention.
4. 04Turn the marked ones into key results in the “from X to Y by date” shape. Take the starting number before the quarter begins, not from memory.
5. 05Write an objective above them in words, saying why the change matters. If the objective reads as a list of the same metrics, it is not written yet.
6. 06Leave everything else on the dashboard as it was. Those metrics go nowhere and keep signalling when the process leaves its limits.

A quarter later, start again from step one. Metrics you moved go back into the KPI set with a new threshold: the one you actually reached.

Failures at the seam between the two systems come down to a handful of plots. Every one of them is visible at planning, if you know what to look for.

• The dashboard moved into the objectives. The team takes fifteen metrics and calls them key results. Focus is gone and the check-in becomes a reading of the table.
• The objective got tied to a bonus. The moment an OKR score feeds compensation, the team sets a bar it will clear for certain. The stretch goal disappears in the first quarter, and the 0.7 scale loses its point.
• The metric became a target and stopped measuring. Goodhart’s law in Marilyn Strathern’s wording (1997): when a measure becomes a target, it ceases to be a good measure. A support team given a key result on tickets closed will close them faster and worse.
• A key result with no starting number. “Improve conversion” cannot be graded: nobody knows where it started. Take the starting number before the quarter begins.
• The two sets live apart. The manager watches KPIs in the monthly report, the team runs OKRs at the check-in, and neither side knows about the other. A quarter later it turns out the objective was pulling the metric down.
• KPIs got renamed OKRs and nothing else changed. Same thresholds, same annual review, a new word in the header row. Two quarters later leadership concludes the method does not work.
• Objectives cascaded top down. Each level slices someone else’s number into parts, and the team ends up with a key result it cannot influence. That is a KPI again, wearing a different interface.