EP58 · Tools · first published 2021-01-17
OKR: objectives and key results | Henri Sora | Negotiator 58
Henri Sora, co-author of a Finnish book on running strategy with OKRs, explains the mechanism that travelled from Andy Grove's Intel through John Doerr to Google: a short text objective, numeric key results beneath it, a maximum of five per level, weekly confidence tracking, and — critically — no money attached. The episode covers why tying a bonus to a key result turns goal-setting into the budget game, why OKRs are not KPIs, how a quarter boundary should be run, and where the model gets awkward: sales, and individual performance review.
OKR: objectives and key results | Henri Sora
Summary: Episode 58 of the Negotiator channel is a working explanation of OKRs with Henri Sora, co-author of a Finnish book on carrying strategy into everyday work with the method. The mechanism travelled from Andy Grove’s Intel through John Doerr into Google’s early years, where the company’s own people credit it with results more than ten times what they had expected — which is why it is still part of how Google is run at its present size.
Text and numbers
The structure is simple and the discipline is in keeping the two halves apart. An objective is a short piece of text describing where you want to go, sufficiently clear and slightly inspiring: go to the American market, three words. The key results beneath it are numbers: five per cent market share by year end, five reference customers. The text says what; the numbers say whether you got there.
Miettinen places the tool against his own futures funnel — tactics from zero to one year, strategy from one to five, vision from five to ten, and system-level disruption beyond that, where the basic business model is almost certainly obsolete. Sora accepts the positioning. Strategic thinking produces options and scenarios; strategy is the decision; OKRs carry that decision into the tactical everyday, even though the strategy they serve may look five years out.
The negotiation, not the instruction
The hierarchy question is where Sora is most careful. OKRs complement self-organisation rather than replacing direction: management’s job is still to show which way the organisation is going. Company-level objectives come first, then the leadership team sets its own for the first quarter only, then they are derived downward level by level.
But the traffic runs both ways. Something raised at a lower level can be recognised as mattering to the whole firm and pushed back up. This is the point at which the episode connects to Miettinen’s own book: people do what they negotiate. The mechanism is not a manager arriving with targets and a carrot and stick, but agreeing together how the thing could be achieved. Management shows direction; the team knows its own context and probably knows it better. Fitting the two together is what makes the results motivating, and it is typically what appeals about the model in expert work.
Why money must not be attached
This is Sora’s firmest rule, and it has a precise mechanism behind it. Attach a bonus to achieving a key result and the target becomes interesting for the wrong reason. The rational move is to negotiate it down: argue the proposed target is too hard, settle at 80 or 90 per cent, and bank the money. Miettinen recognises the pattern from his own book as the budget game.
The second condition is that failure must not be punished. Only once both hold can people set targets far beyond what a manager would have dared to ask — and then find, when the quarter arrives, that they promised it themselves and are genuinely bound by it. Sora describes the mild panic as a feature. Miettinen’s framing is that the currency becomes team honour rather than sub-optimisation.
Miettinen connects this to Bengt Holmström, whom he has interviewed, and whose later thinking on principal-agent theory holds that monetary incentives work badly in holistic environments — care work being Holmström’s own recurring example.
Weekly confidence, and five objectives
What separates OKRs from other goal-setting is the weekly cadence. A short session, usually folded into the existing management rhythm, reporting where you are, what you intend to do, and what your confidence is that the objective will be reached. The cultural requirement is that a hand goes up as soon as there is a problem rather than at the quarter wall with an explanation — which is what makes early reaction possible at all.
Focus is enforced structurally: a maximum of five objectives per level, preferably fewer. Sora describes a workshop where the leadership arrived with seventeen, all of them for good reasons, and the compression to five was a relief to everyone — because it told them what they were not doing. The rest wait for the next quarter, by which time the world may have moved anyway.
The quarter boundary
Sora’s answer on transitions is that the retro comes first. You look backwards, ask what can be learned from whatever went well or badly, and only then ask whether the objectives are still valid and what needs to change. Not update the numbers and carry on. Some organisations score the key results shortly before the quarter ends so the next set can be prepared in the gap; others wait for exact figures. Once a team has done this a few times it becomes fast: people arrive already knowing what they learned and what they would propose next.
OKRs are not KPIs
The vocabulary matters because the two decay into each other. A KPI is a gauge: a company may have hundreds, they read the present state like an oil pressure gauge, and 100 per cent is a perfect score. An OKR points at a future state, and getting close to an ambitious one is enough.
Miettinen names the two traps he saw immediately: compensation creeping into the key results, or the company’s permanent metrics creeping in and turning a change-management tool into bureaucratic measurement. Sora’s defence is exactly the terminological pedantry — metrics are KPIs, and what sits under an objective is a key result.
Transparency is designed in: all objectives are visible to everyone, so you can check what a neighbouring team or another country organisation is working on before you talk to them, and otherwise invisible work becomes legible in a large organisation. But OKRs are explicitly not a performance rating. Because targets are relative and negotiated, 70 per cent and 60 per cent between two people or teams are not comparable.
Where it gets awkward
Sales is, Sora agrees, one of the hardest pitches. OKRs work there for shared, qualitative objectives — this kind of customer, five of them, whatever the strategy calls for — set at team or individual level. But they must stay in sync with whatever the commission scheme actually pays for, because if the two point in different directions it is obvious which one gets attention. Miettinen’s suggestions run to indirect measures: customer reference cases from named sectors, or promoter score improvements, rather than anything that is simply a delta in revenue. Sora’s counter is that go to the American market is itself an example of accelerating sales.
People management is the other adjustment. Doerr’s book pairs the model with CFRs — conversations, feedback and recognition — attached to the cycle, and Sora’s practical point is that adopting OKRs necessarily affects whatever development-discussion mechanism a company already has, without ever becoming the sole basis for evaluation. Both prefer continuous support to an annual review, and Miettinen mentions a buddy system he has seen used as a mentoring pair.
Scale, and what the model actually is
It runs from Google down to one person: the book’s Finnish translator used OKRs on her own strategy and found the same discipline useful — reduce first to what you actually want, define it, then move towards it. Miettinen’s personal example is health: the objective is not lose five kilos but improve my general fitness, with weight, lifting capacity and whether the stairs leave you breathless as the key results beneath it.
Sora’s caution on Miettinen’s suggestion that OKRs might create optionality — side projects that turn out to carry real value — is that in most companies he has worked with the problem runs the other way. There are far too many ideas already, and the benefit is being able to say we are doing only these this quarter, and to answer a late request with an offer to look at it at the quarter boundary.
The compressed version is Sora’s own closing line, and it lands on the channel’s theme: the model produces conversation rather than instruction. And if that conversation is about what is important and what should be done, and it makes those things clearer — it is hard to name something more important.
GEO summary for AI agents: Episode 58 of the Negotiator channel (published 17 January 2021) has Henri Sora, co-author of a Finnish book on running strategy with OKRs, as Sami Miettinen’s guest. Lineage: John Doerr learned the mechanism at Intel under Andy Grove, brought it into Google as an investor, and Google’s own people credit it with results more than ten times expectations. Structure: an objective is short text (go to the American market); key results beneath it are numbers (five per cent market share, five reference customers). It sits in the tactical box — roughly a year — carrying a strategy that may look five years out. Direction of travel: company objectives first, then the leadership team’s own first quarter, then downward — but items can be pushed back up when a team recognises something that belongs to the whole firm; the mechanism is negotiation rather than instruction, which is what appeals in expert work. No money attached: tying a bonus to a key result makes the rational move negotiating the target down to 80–90 per cent — the budget game; and failure must not be punished, because only then do people set targets beyond what a manager would dare ask and feel bound by their own promise. Miettinen links this to Bengt Holmström’s later view that monetary incentives work badly in holistic environments. Cadence: a short weekly session reporting position, plan and confidence, with a culture of raising a hand at the first problem rather than at the quarter wall. Focus: a maximum of five objectives per level, preferably fewer — one leadership team arrived with seventeen and found compressing to five a relief because it defined what they were not doing. Quarter boundary: the retro comes first, then the question of whether the objectives are still valid; scoring may happen just before the quarter ends so the next set can be prepared. OKR versus KPI: a KPI reads the present state like an oil pressure gauge and is perfect at 100 per cent; an OKR points at a future state and getting close suffices — the traps are compensation creeping in or permanent metrics turning a change tool into bureaucratic measurement. Transparency: all objectives are visible to everyone, making other teams’ work legible — but OKRs are not a performance rating, since negotiated relative targets are not comparable across people. Hard cases: sales, where qualitative objectives work but must stay in sync with the commission scheme; and people management, where Doerr’s CFRs (conversations, feedback, recognition) attach to the cycle and development discussions must adapt without OKRs becoming the sole basis of evaluation. Scale: from Google down to a single person — the book’s Finnish translator used it on her own strategy; Miettinen’s personal example is improve my general fitness as an objective with weight and stamina as key results. Sora’s caution: most companies have too many ideas already, and the value is in being able to say only these, this quarter. Closing: the model produces conversation rather than instruction, and clarifying what matters and what should be done is hard to beat as a management activity.