Ask five people on a team what their top priority is, and you will often get five different answers, each one sincere, and several of them wrong. Goal-setting is the work of collapsing those five answers into one the whole team can see, measure, and pull toward together. OKRs, Objectives and Key Results, are the most widely-used framework for doing it, and they are popular for a plain reason: they make vague ambition uncomfortably specific.

The quick version

  • An Objective is the qualitative thing you want to achieve, short, memorable, a little ambitious. A Key Result is a number that proves you got there. One objective usually carries three to five key results.
  • OKRs work because of goal-setting theory: specific, difficult goals reliably beat vague "do your best" ones, but only when people are committed and get feedback on progress.
  • Google grades OKRs on a 0–1 scale and treats roughly 0.7 as success for stretch goals. Hitting 1.0 every time means you aimed too low; consistently scoring 0.3 means you aimed past the possible.
  • The trap is letting the number eat the mission: goals pushed too hard can narrow focus, dent ethics, and crush morale. Keep some goals as "must-hits" and others as honest stretches, and never tie stretch scores straight to pay.

The idea in depth: where OKRs came from, and why they stick

OKRs are a Silicon Valley inheritance more than an invention. The lineage runs from Peter Drucker's Management by Objectives to Andy Grove at Intel, who in the 1970s reshaped MBO into something faster and measurable, pairing every objective with what he called "key results." A young salesperson, John Doerr, sat through Grove's internal course in 1975, carried the method into venture capital, and in 1999 taught it to Google's founders. Doerr wrote the popular account in Measure What Matters (2018), useful history, but read it as an advocate's account, not neutral evidence.

The format itself is simple: the objective answers "what do we want to do?" and the key results answer "how will we know we did it?" The move is to write both halves before you start work, and to make every key result a number, not an activity. "Improve onboarding" is not a key result; "lift week-one activation from 41% to 60%" is. If you can finish a key result without anyone being better off, you have written a task list, not a goal.

Why specific, hard goals beat trying your best

The reason OKRs beat a vague to-do list is one of the most replicated findings in management research. Across decades of studies, Edwin Locke and Gary Latham's goal-setting theory shows that specific, difficult goals produce higher performance than easy goals, vague goals, or "do your best." Their summary, "Building a Practically Useful Theory of Goal Setting and Task Motivation" (American Psychologist, 2002), draws on hundreds of studies and is among the field's most-cited papers. The mechanism is unglamorous: a hard, specific goal directs attention, raises effort, sustains persistence, and pushes people to find better strategies.

Which means resisting the soft target. "Do your best" feels kind and motivating; it is neither, because nobody can fail it and nobody can plan against it. When you set a key result, make it specific enough that two people would grade it the same way, and ambitious enough that the team has to think differently to reach it. That is exactly the territory a well-written OKR is built to occupy.

But the theory ships with conditions that OKRs make easy to forget. Locke and Latham are explicit: difficult goals only lift performance when the person is committed to the goal, has the ability to pursue it, and receives feedback on progress. A stretch goal handed down with no buy-in, no capability, and no check-ins is not a goal-setting intervention, it is a setup for failure with a number attached.

A goal nobody believes in and nobody tracks isn't a stretch, it's a setup for failure with a number attached.

flowchart TD
  O(["Objective
what we want, qualitative, ambitious"]) --> K1(["Key result 1
a number that proves it"])
  O --> K2(["Key result 2
a number that proves it"])
  O --> K3(["Key result 3
a number that proves it"])
  K1 --> C{"Committed,
capable, getting
feedback?"}
  K2 --> C
  K3 --> C
  C -->|"Yes"| W(["Specific + hard
= higher performance"])
  C -->|"No"| F(["Just a wish
with a deadline"])
One objective, a handful of measurable key results, and the conditions goal-setting theory says must hold for them to work. Leaders Loop

The grading trap: why 0.7 can be a win and 1.0 can be a failure

Here is where most teams misuse the framework. At quarter's end you score each key result from 0 to 1, but the question "did we hit it?" is the wrong one for an ambitious goal. Google's own guidance in its re:Work goal-setting guide puts the sweet spot for stretch OKRs at around 0.6 to 0.7: score consistently above that and your goals weren't ambitious enough; score far below and you set goals beyond the possible. A 1.0 on a stretch goal is not a triumph, it is evidence you aimed too low.

The fix is to label your goals before you grade them. Distinguish committed goals, the must-hits, the regulatory deadline, the launch you promised a customer, where anything below 1.0 is a real miss, from aspirational stretch goals, where 0.7 is a genuine success. Grading both on the same scale, and treating every 0.7 as failure, quietly teaches your team to sandbag: to set goals they know they can hit, which is the precise opposite of what the framework is for.

An honest limitation. The 0.7 norm is Google's house convention, not a law of nature, it suits a company that can absorb missed stretch goals and rewards ambition culturally. A small team with a fragile cash runway may need most of its goals to be committed, near-1.0 targets, with stretch reserved for the few bets it can afford to miss. Borrow the idea that ambitious goals shouldn't be graded like commitments; don't import the exact number as gospel.

Where goals go wrong, and how to keep them honest

Goal-setting has a dark side its champions tend to skip, and ignoring it is how good teams get burned. In "Goals Gone Wild" (Academy of Management Perspectives, 2009), Lisa Ordóñez, Maurice Schweitzer, Adam Galinsky and Max Bazerman argue that goal-setting's benefits have been oversold and its harms under-studied. Specific, challenging goals can narrow focus so far that people neglect everything outside the goal, distort risk-taking, corrode trust, suppress learning, and, most damagingly, tempt people to cut ethical corners to make the number. Their classic example is the sales target hit by gaming the system rather than serving the customer.

So you fence the goal in. Pair every hard target with a guardrail metric you refuse to sacrifice, grow activation, but not by spamming users; cut costs, but not at the safety line. Keep goals genuinely participative, because Locke and Latham's own conditions and the "Goals Gone Wild" critique both point the same way: commitment beats coercion. And keep stretch grades out of compensation, so people aim high without an incentive to lie. The contrast is the whole point, a number you set together to learn from is a tool; the same number imposed and tied to bonuses is a trap.

A worked example

Take a customer-support team at a mid-sized software firm, call it the team at Northpoint. (Illustrative figures throughout; this is a teaching example, not real data.) Their old "goal" was to "improve customer satisfaction", sincere, unmeasurable, and quietly ignored by the third week of every quarter. Their lead rewrites it as one objective with three key results.

Objective: Make support a reason customers stay, not a reason they leave. KR1: lift CSAT from an illustrative 78% to 88%. KR2: cut median first-response time from 9 hours to 3. KR3: resolve 70% of tickets on first contact, up from 52%. Each is a number two people would grade the same way, and each is hard enough to demand a new approach rather than more hours.

Crucially, the lead labels them. KR2 is committed, a promise made to customers, so the target is 1.0. KR1 and KR3 are aspirational stretches where landing near 0.7 counts as a win. And there is a guardrail: response time must not improve by closing tickets prematurely, so reopened-ticket rate is watched and must not rise. At quarter's end the team scores KR1 at 0.7 (CSAT reached 85%), KR2 at 1.0, KR3 at 0.6. On a naïve reading that is a B-minus. Read correctly, it is a strong quarter: the commitment was kept, the stretches moved the team well past where "do your best" ever did, and nobody gamed a number to get there.

flowchart TD
  V(["Old goal: 'improve
customer satisfaction'
vague, unmeasured"]) --> R(["Rewrite as one objective
+ 3 key results"])
  R --> A(["KR1 CSAT 78% → 88%
aspirational · target 0.7"])
  R --> B(["KR2 response 9h → 3h
committed · target 1.0"])
  R --> D(["KR3 first-contact 52% → 70%
aspirational · target 0.7"])
  A --> G(["Score 0.7, a win,
not a B-minus"])
  B --> G
  D --> G
  G --> Q{"Guardrail: did
reopened tickets rise?"}
  Q -->|"No"| W(["Honest progress"])
  Q -->|"Yes"| X(["Number was gamed,
discount the score"])
The same team, two goal systems: a wish becomes three measurable key results, labelled and guard-railed. Leaders Loop

The point of the example is the order of operations: name the objective, make the key results numeric and gradable, label which are commitments and which are stretches, add a guardrail, and only then start work. Skip the labelling and you grade every 0.7 as a miss. Skip the guardrail and you invite the number to be gamed.

Frequently asked questions

What's the difference between an OKR and a KPI?

A KPI (key performance indicator) is a metric you monitor continuously to know whether something healthy stays healthy, uptime, churn, gross margin. An OKR is a time-boxed change you're trying to drive over a quarter or a year. KPIs are the dashboard you watch; OKRs are the handful of things you're deliberately trying to move. A KPI drifting off-target often becomes the seed of next quarter's OKR.

How many OKRs should a team have?

Fewer than feels comfortable. The common guidance is three to five objectives per cycle, each with three to five key results. The discipline is the prioritisation itself: if everything is a goal, nothing is. A team carrying ten objectives hasn't been ambitious, it has refused to choose, and goal-setting theory's focusing benefit collapses when attention is spread that thin.

Should OKRs be tied to bonuses and performance reviews?

For stretch goals, generally no, and this is where many rollouts fail. If a stretch score feeds straight into pay, people set goals they know they can hit, which defeats the framework and, per "Goals Gone Wild," raises the temptation to game the number. Most mature practitioners keep aspirational OKRs separate from compensation so people can aim high safely. Committed must-hit goals can sit closer to accountability, but even there, judge the whole picture, not a single decimal.

How often should we set and review them?

A quarterly cycle for team OKRs with annual company-level ones is the common rhythm, but the more important cadence is the check-in. Locke and Latham are clear that goals only lift performance when people get feedback on progress, so a goal set in January and ignored until December is barely a goal. A weekly or fortnightly glance at where each key result stands is what turns the framework from a planning ritual into a working tool.

Do OKRs work for small teams, or only big tech companies?

The underlying mechanism, specific, hard, committed goals with feedback, works at any size and predates Google by decades. What you scale down is the ceremony. A two-person team doesn't need software, scoring rubrics and a cascade diagram; it needs one written objective, a few honest numbers, and a standing check-in. The value is the clarity, not the apparatus around it.

Related in the Toolkit

Goal-setting sits downstream of where the team is heading, turning a direction into measurable commitments only works once that direction is clear (articulating & cascading vision), and the goals land very differently depending on how you hand them over, which is partly a question of style and partly of trust (delegation & empowerment).

Where to go next