From MBO to OKR: what got lost along the way

A clinic network, 600 people, 14 sites. In January, three days of OKR workshops: objectives, key results, a shared spreadsheet, everyone aligned. In April, at the quarterly review, 40% of the key results had no updated number, 30% had been rewritten midway, and the rest were on track because they'd been set with plenty of slack.

The meeting concluded that the team needed to 'take OKRs more seriously'. Not one of those four hours was spent asking why updating a number required someone to pull data out of three systems and type it in by hand.

01The idea: Drucker, 1954 #

Peter Drucker introduced management by objectives in The Practice of Management in 1954. The full name matters: management by objectives and self-control. The second half is almost always forgotten, and it's the half that carries the argument.

Drucker's proposal was that every manager understand precisely what contribution their unit owes the whole, agree with their superior on the objectives that follow from it, and — this is the point — have direct access to the information showing their own performance, so they can correct course without waiting for someone to point out the mistake.

Drucker was explicit about the risk: if performance information travels upward before it reaches the people doing the work, the method turns into an enforcement tool and loses exactly what justified it.

02Grove, 1983: the method gets a cadence #

Andy Grove, then at Intel, published High Output Management in 1983 with an operational version of that idea. Two questions: where do I want to go (the objective) and how do I know I'm getting there (the key results, measurable). The cycle is short — quarterly, not annual — and the numbers are few.

Grove added three things that matter in practice. A short cycle, because an annual objective doesn't survive contact with reality. Few indicators, because measuring everything is prioritizing nothing. And targets ambitious enough that hitting 100% means the target was easy — which is why, he argued, OKRs shouldn't feed directly into individual reviews and bonuses.

That's the formulation John Doerr brought to Google in 1999 and which, from the 2010s on, reached mid-size companies across Latin America — almost always through the workshop, and almost never through the information system Drucker treated as a precondition.

03What got lost along the way #

• Self-control. In most rollouts, a central team consolidates the number and presents it at the review. The people doing the work learn how they performed once it's already a slide.
• Keeping bonuses separate. Tying OKRs to pay produces the obvious behavior: conservative targets, negotiated with slack. The method starts measuring negotiating skill, not results.
• The cadence. A 90-day cycle with two reading points — the start and the end — isn't a cadence, it's a report. Course correction depends on seeing the number while there's still time.
• The link to the work. The objective lives in a spreadsheet; the work lives somewhere else. If updating one takes a data-gathering effort, it will be updated the night before the meeting, with an estimate.

04The math of a cycle that doesn't close #

The scale of that 600-person Brazilian network (figures in Brazilian reais), with 45 managers involved in the cycle. Conservative premises, stated so you can redo the math with your own numbers:

Want to see how this looks inside a real operation? Explore the platform.

The math: 45 managers × 9 hours × 4 quarters = 1,620 hours; at R$ 90 an hour, roughly R$ 146 mil a year. That spend isn't the problem — it's a reasonable investment in shared direction. The problem is when about half those hours go to collecting and formatting data instead of deciding. At that point the company is paying R$ 73 mil a year to retype numbers that already exist in some system.

05Where this stalls in practice #

1. The key result has no source. 'Increase customer satisfaction' without defining where the number comes from guarantees it will be estimated.
2. Too many objectives. Seven objectives per area with four key results each is 28 numbers per area. Nobody tracks 28 numbers; the cycle dies of excess.
3. Goals confused with tasks. 'Roll out the new system' is a deliverable, not a result. The result is what changes once the system is in.
4. Reacting to normal fluctuation. An indicator down two points isn't always asking for an action plan — variation isn't error, and treating noise as signal burns the method fast.

06What data-driven management answers #

The distance between Drucker's intent and the spreadsheet ritual is, to a large degree, an information-system problem. When the work is already recorded — ticket handled, order delivered, project closed, deadline met — the key result reads that record instead of depending on manual collection. The quarterly review stops being a consolidation exercise and goes back to being a conversation about decisions.

One warning Drucker made and Grove reinforced: a visible number is a tool for autonomy or for surveillance, depending on who sees it first and why. Choosing between those two uses is, at bottom, a question of what theory you hold about people. And realistic goals depend on knowing how much is open at once — more open fronts deliver less.

That fit between objective and recorded work is what the performance and operations modules in the platform support, and what the solutions by area show in context.

The team didn't need to take OKRs more seriously. It needed a number that existed without someone typing it.