Fewer Goals, Scored in Public

The Core Insight

In November 1979 an Intel sales manager named Don Buckout sent an eight-page telex up the chain. Intel was losing the 16-bit market to Motorola's 68000 and Zilog's Z8000. A task force convened on Tuesday, December 4, and had a plan by Friday. The next Tuesday brought approval for a nine-part program called Operation Crush, and each of the nine projects became a company key result.

No product was modified. Intel resold the same chip on support and cost of ownership, and stopped selling to programmers to sell to CEOs. By 1986 the 8086 held 85 percent of the 16-bit market.

Andy Grove installed the system at Intel in 1971 and called it iMBOs. John Doerr arrived as an intern in 1975 and gave it the name OKRs. An objective states what is to be achieved. Key results benchmark and monitor how, and Marissa Mayer set the bar: it is not a key result unless it has a number.

Completing all key results must by construction achieve the objective, and an OKR that fails that test was designed badly. Doerr's mantra is that ideas are easy and execution is everything.

Most companies set goals once a year, keep them private, push them down from the top, and attach them to the bonus. Doerr argues for quarterly cycles, public goals, roughly half of them set from below, and pay decided in a separate conversation. A meta-analysis of seventy studies of the Drucker ancestor found 56 percent productivity gains under high commitment, against 6 percent under low.

The Framework

The book organizes the system as four superpowers, each with its own mechanics.

• Focus and Commit picks three to five objectives a cycle, with five or fewer key results each.
• Align and Connect makes every goal public, with roughly half of them set bottom up.
• Track for Accountability runs weekly check-ins, four legal moves, and honest scoring at the end.
• Stretch for Amazing sets aspirational goals nobody expects to hit in full.

The writing rules are exact: an objective is significant, concrete, action oriented, and its success must be obvious to a rational observer. Key results are specific, time-bound, aggressive yet realistic, and above all measurable and verifiable.

Key results describe outcomes rather than activities. Words like consult, help, analyze, and participate mark an activity. Publishing latency measurements from six Colossus cells by March 7 passes, where assessing Colossus latency fails. Every key result carries evidence of completion that is available, credible, and easy to find.

Objectives then sort into two baskets, and the split is the most consequential rule in the book. Committed objectives tie to metrics such as releases, bookings, hiring, and customers. They must be achieved in full inside the set time frame, and a miss triggers a postmortem rather than punishment. Aspirational objectives describe how the world looks when nobody knows the route yet, and they carry over quarter to quarter until done.

CFRs are the fifth piece, sitting underneath the other four. Conversations, feedback, and recognition replace the annual performance review.

Key Ideas

Completing the Key Results Must Reach the Objective

The test for a committed objective is one question. Is it reasonably possible to score 1.0 on every key result and still miss the intent of the objective? A yes means the set gets reworked until completion guarantees the objective.

Other traps follow. Business-as-usual OKRs describe what a team achieves without changing anything. Mislabeling a commitment as aspirational means the team stops taking it seriously. Sandbagging shows when committed OKRs consume most but not all available resources. Low value objectives fail the who-cares test, since raising task CPU utilization by 3 percent helps nobody.

Two litmus tests stay cheap. An OKR written in five minutes is not good enough. Key results that all land on the quarter's last day mean the team has no plan.

Three to Five Objectives, and Each One Needs a Counterweight

Grove's first rule of hygiene is that less is more. A short list says what a company refuses as clearly as what it accepts. Among eleven thousand senior executives and managers surveyed, most failed to name their company's top priorities, and only half named even one.

Grove paired key results for effect and counter-effect, so every quantity carries a quality partner. Three new features pair with fewer than five bugs per feature in QA.

One-dimensional goals kill. The Ford Pinto shipped in 1971 under two objectives enforced with an iron hand: under 2,000 pounds and under 2,000 dollars. Safety appeared in none of its three product objectives. A one-dollar, one-pound plastic part stopped the gas tank from puncturing in crash tests, and Ford cut it as extra cost and weight. The 1978 recall covered 1.5 million Pintos and Mercury Bobcats.

Wells Fargo hit its account goals by opening millions of fraudulent accounts, and the bank fired more than five thousand bankers.

Cascading Every Goal Costs More Than It Aligns

Public goals beat private ones. Of a thousand U.S. adults surveyed, 92 percent reported more motivation when colleagues see their progress. Only 7 percent of employees fully understand their company's strategy.

Cascading everything degrades into a color-by-numbers exercise with four costs. Agility goes first, because six or seven reporting levels stretch each cycle into weeks. Flexibility goes next, because goals that cost that much to formulate never get revised. Frontline contributors lose their say, and peers across departments never connect.

Because OKRs are visible, levels can be skipped, and the healthy mix settles near half top-down and half bottom-up. Intuit's Atticus Tysen rolled OKRs to all 600 IT employees without mandating anything. Employees view their managers' OKRs more than 4,000 times a quarter, seven views per person.

Seventy Percent Is the Target, and Nobody Aims at It

Aspirational OKRs at Google are set at 60 to 70 percent attainment, in line with Grove's old standard. Performance is expected to fall short at least 30 percent of the time, and that counts as success. Aspirational failure runs near 40 percent inside Google. Susan Wojcicki's caveat holds it together: no team enters an OKR planning to settle for 70 percent.

Chrome shows the arithmetic of missing. The 2008 key result was 20 million seven-day active users from a base of zero, and they missed. The 2009 target of 50 million missed too, ending at 38 million. They set 111 million for 2010 and closed the year there. Sundar Pichai thought there was no way to reach the first number, and putting it out there made complacency impossible.

In February 2011 YouTube had about 800 people producing hundreds of OKRs a quarter, and less than half got done. Narrowing to three or four objectives per team fixed the focus. In September 2011 Cristos Goodrow sent an email arguing for watch time, and only watch time. The proposal was heresy, since watch time is negative for views. Winning the argument took six months.

In November 2012 they set one billion hours of daily watch time, a 10x increase dated to the end of 2016. The framing made it survivable: a billion daily hours was less than 20 percent of the world's television watch time. Susan Wojcicki inherited the goal in February 2014, behind pace, and kept it. In 2016 engineers found around 150 tiny advances, some worth 0.2 percent more watch time, and needed nearly all of them. They crossed a billion hours that fall, ahead of schedule.

Bill Gates names the ceiling. His malaria team expected to eradicate the disease by 2015, and a goal set too aspirational is bad for credibility.

Grade the Number, Then Override It

Google scores from 0 to 1.0, where 0.7 to 1.0 is green, 0.4 to 0.6 is yellow, and 0.0 to 0.3 is red. An objective scores as the average of the completion rates of its key results.

The Q2 1980 Crush OKR shows it in use. Publishing five benchmarks scored 0.6, and repackaging the 8086 family scored 1.0. The 8MHz part scored zero after a polysilicon failure. The average of 0.625 was respectable, because everyone knew how aggressively management set the goals.

Judgment then overrides the number. Thirty-five of fifty prospect calls is a raw 70 percent, and a dozen long calls yielding eight customers earns 1.0.

Doerr prescribes weekly one-on-one OKR meetings plus monthly departmental meetings. Four moves stay legal at any point. Continue is green, update revises a yellow key result or its timeline, start adds an OKR mid-cycle, and stop drops a red one.

Bonuses Attached to Goals Produce Sandbagging

The annual review costs 7.5 hours of manager time per direct report. Only 12 percent of HR leaders call it highly effective, and a tenth of the Fortune 500 has already dropped it.

CFRs replace it with conversations, feedback, and recognition. Conversations are authentic exchanges aimed at driving performance, feedback is networked evaluation among peers, and recognition is appreciation for contributions of all sizes.

The first move is divorcing compensation from OKRs. Raises get a backward-looking conversation at year end, and OKRs get an ongoing forward-looking one on five questions.

• Ask what the person is working on right now.
• Ask how the current OKRs are coming along.
• Ask whether anything is impeding the work.
• Ask what the person needs from you.
• Ask what growth their career goals require.

At Google, OKRs count for a third or less of a performance rating, and raw scores are wiped after each cycle. Adobe's annual reviews consumed eighty thousand manager hours a year, and attrition spiked every February. Its replacement, called Check-in, runs quarterly goals, regular feedback, and career development, with pay decided separately. Voluntary attrition dropped sharply.

Practical Applications

Write three objectives for the current quarter and stop. Give each one three to five key results with a number and a date attached. Then score them all at 1.0 on paper and check whether the objective is achieved.

Pair every quantity key result with a quality one. A feature count pairs with a bug ceiling. The Pinto met every objective it was given.

Label each objective committed or aspirational before the quarter starts, and tell the team which is which. Committed means 1.0 or a postmortem. Aspirational means 0.7 is the expectation.

Publish the list where anyone can read it, starting with your own. Nuna pushed OKRs to about twenty people in 2015 and failed, because some never set them and others filed them in a drawer. The retry started with a leadership team of five.

Put the weekly one-on-one and the monthly team review on the calendar. Decide raises in a different meeting on a different cadence. Expect four or five cycles before people take to it, and up to a year before adoption is real.

Who This Is For

Founders past their first ten people get the most from it. Zume adopted OKRs three weeks after the first pizza went out, at sixteen salaried employees, because asking what mattered that day produced eight different answers.

Solo operators and teams of three can keep the discipline and skip the system. Anyone hunting for strategy will find none here, since OKRs assume the strategy is already chosen.

Doerr is a partner at a venture capital firm writing about companies his firm backed, and the case studies are picked from winners. A dozen of them run the same arc: chaos, an OKR pitch, a stumble, discipline, then a hockey stick. The book carries no adopter who failed and stayed failed.

The failure mode it underplays is the one most teams hit. OKRs become a reporting tax instead of a focus tool, a quarterly ritual of writing down work already planned. Lumeris ran three cycles on paper with a high reported participation rate, and people adjusted metrics after the fact to check boxes. Andrew Cole's diagnosis was missing executive buy-in, no business rationale, and no accountability for the system itself.

The Decision

Open the goals for the quarter you are in and count the objectives. More than five at the top means the company has priorities on paper and none in practice.

Then run three checks. Score every key result at 1.0 in your head and ask whether the objective is reached. Find the quantity goal that has no quality partner. Name which objectives are committed and which are aspirational, then check whether anyone else knows.

A miss on any of the three means the quarter is already spending itself on work nobody chose. Rewrite the list this week, publish it, and put the pay conversation on a separate calendar. The rewrite takes an afternoon. The quarter takes three months.