Running It When It Is Already Broken

What to do this week

Change one thing, and make it the weekly check. Every key result gets a number, written down, in the same place, said out loud to the team.

Leave the goals alone. Rewriting them in week one confirms the suspicion that this is all churn, and the goals are rarely the problem anyway.

Most repairs start in the wrong place

The instinct with a struggling system is to improve the process. Better offsites, tighter objectives, a tool.

The more common situation is that a goal-setting instrument is being used outside the conditions where goal setting works at all. Five questions test that, and each one traces to evidence rather than to consulting habit. They are worth running before anything is changed.

Which system are you actually running? If three leaders describe it differently, that is the documented state of the field rather than a failing of the team. That makes it safe to raise early. See How the Planning Stack Fits Together for why the vocabulary is genuinely unsettled.

Is the work too complex for the instrument? Goal-difficulty effect sizes weaken as complexity rises, to ".48 for the most complex tasks versus .67 for the least complex tasks" (Locke and Latham, 2002). The critics put it more sharply. Goals "cause the most harm in complex, natural settings when outcomes are interdependent, monitoring is difficult, and cheating possible" (Ordonez et al., 2009). That describes most knowledge work.

Does anyone hold conflicting goals? A team carrying quarterly objectives, a roadmap commitment and a service target is not being asked to prioritise. It is being asked to fail at something and choose which.

Is the system wired to performance review? If it is, people will negotiate the number down before the cycle starts, and they are right to.

Can the team measure an outcome at all? Where they cannot, key results quietly become task lists whatever the training said. The signature is a number that sits at zero for weeks and then jumps to complete.

The number everyone has wrong

A specific confusion causes more damage than any process flaw, and it dissolves on contact with the source texts.

The convention says aim for seventy percent. The man usually credited with the method wrote something far more aggressive. Objectives should be set so high that even someone pushing hard "will still only have a fifty-fifty chance of making them" (Grove, 1983).

The research literature is more demanding still. Characterising the goal-setting research, Sitkin and colleagues describe a challenging goal as carrying "a nonzero (typically, 10 percent) probability of attainment" (Sitkin et al., 2011). A stretch goal, in their definition, is seemingly impossible.

Line those up and the seventy percent is the softest of the set, not a midpoint: 70, then Grove's 50, then 10, then near zero.

It is also not the same measurement. Seventy percent is an expected score, meaning how much of the goal you land. The other two are probabilities of hitting the goal at all. Comparing them directly is a category error, which is part of why the number travels so easily.

Where it does come from is one company's internal practice, and there it applies only to goals explicitly marked aspirational. Goals marked committed are expected to be met in full, with a miss requiring an explanation.

Most struggling systems have merged those two categories, applied the softer number to everything, and then attached the result to performance ratings. Separating them again is the cheapest high-value change available, and it costs one conversation.

If they are reaching for stretch goals, stop

A team that has missed its targets often proposes more ambitious ones. This is the worst available move, and it is predictable.

Stretch goals help organizations with strong recent performance and spare capacity, and harm those without either. They are "paradoxically, most seductive for organizations that can least afford the risks associated with them" (Sitkin et al., 2011).

A struggling client proposing a harder target is that paradox in motion. The evidence supports easier goals this cycle, not harder ones, and says so on the same grounds that would otherwise recommend stretch.

This is not an argument against goals. The critics stop well short of that. They recommend treating goal setting as "a prescription-strength medication that requires careful dosing" and close supervision (Ordonez et al., 2009).

Expect the cascade to be broken, because it usually is

When goals do not connect to work, the instinct is to blame the teams writing them. The evidence points a layer up.

The best available empirical study drew on 47 interviews and 512 survey responses inside one large organization. In it, "more than 50% of managers reported that they are not very effective at translating OKRs" downward (Butler et al., 2023).

That is the normal outcome rather than a local failure, and it changes what to fix. The cascade depends on every middle manager performing a translation most of them report doing poorly. That is a design problem, not a discipline problem.

Week one

Six moves, in order. Five of them are removals or clarifications.

1. Do not rewrite the goals. Whatever is on the board stays for this cycle.
2. Start the weekly check immediately. Same time, one page, every key result with a number and a confidence score. This is the only addition on the list.
3. Sever the link to ratings, out loud, from leadership. Sandbagging stops when the reason for it is removed, not when it is asked to stop.
4. Find the step functions. Any key result that jumps from zero to complete is a task. List them without fixing them, because the list is the teaching material for the next setting session.
5. Give the cycle an owner, and have leadership attend rather than receive. The sector that succeeded with this kind of system had "permanent employees overseeing the entire process" (Pavlickova et al., 2022). Smaller organizations had the most serious problems. If nobody's calendar has the cycle on it, nobody is running it.
6. At the next setting session, change exactly one thing. Split committed from aspirational. That single move repairs the seventy percent confusion, makes a miss meaningful again, and costs nothing.

What to remove

Subtraction buys more credibility with a sour team than addition, and none of these removals touches anything evidenced.

Individual goals. They collapse into task lists and pull the whole system toward performance review. Company and team level only.

Mechanical cascading. Every level waits for the level above, so a quarter is partly gone before teams begin. The translation step is one most managers report doing poorly anyway. Publish the company goals, let teams write their own against them, then check the connection.

Tooling, for now. The study above concluded that "tracking, measuring and setting goals is hard work, regardless of tools used" (Butler et al., 2023). That finding comes from authors whose employer sold goal-setting software at the time, which makes it more credible rather than less. A shared document everyone can see beats a system nobody opens.

The grading ceremony. Keep a score for the retrospective. Nobody has measured whether the grade itself changes behaviour, so it does not deserve the meeting it usually gets.

One thing to tell them early

In large companies, implementation of this kind of system "had been successful and improved over the years through 5-year cycles" (Pavlickova et al., 2022). Improvement arrived over years, on five-year policy horizons rather than quarterly ones.

That is a finding about large organizations, and it is the closest thing to a timescale anyone has published. Read against it, a team two quarters in and frustrated is early rather than failing. Saying that out loud is worth more than most of the techniques here. It converts a felt failure into an expected stage, and buys the patience every other change needs.

The same review found the most serious problems in smaller organizations. It credits the sector that succeeded with "having permanent employees overseeing the entire process" (Pavlickova et al., 2022). At small scale the constraint is rarely the design. It is whether the cycle is anybody's actual job.

Key Takeaways

• Change the weekly check first and the goals last. The goals are rarely the problem, and rewriting them in week one confirms the team's worst suspicion.
• Run the five questions before the repair. Most struggling systems are using a goal-setting instrument outside the conditions where goal setting works.
• The seventy percent matches no literature. The originator said fifty-fifty and the research says ten percent. The company the number comes from applies it only to goals marked aspirational.
• A struggling team proposing harder targets is the stretch-goal paradox in motion. The evidence recommends easier goals this cycle.
• Assume the cascade is weak at middle management, because most managers report being not very effective at the translation it depends on.
• Subtract before adding. Individual goals, mechanical cascading, tooling and the grading ceremony can all go without touching anything evidenced.
• Tell them improvement arrives over years. Where large organizations succeeded with this, the gains came over five-year horizons, so a team two quarters in is early rather than failing.

Sources

• Locke and Latham — Building a Practically Useful Theory of Goal Setting and Task Motivation (American Psychologist 57(9)), 2002. goal difficulty effect sizes (d) of .48 for the most complex tasks versus .67 for the least complex tasks. View source · verified 2026-09-24 · primary
• Ordonez, Schweitzer, Galinsky and Bazerman — Goals Gone Wild (Academy of Management Perspectives 23(1); HBS working paper 09-083), 2009. we argue that goals cause the most harm in complex, natural settings when outcomes are interdependent, monitoring is difficult, and cheating possible. View source · verified 2026-09-24 · primary
• Andrew S. Grove — High Output Management (Random House), verified against the full text, 1983. objectives should be set at a point high enough so that even if the individual (or organization) pushes himself hard, he will still only have a fifty-fifty chance of making them. View source · verified 2026-09-24 · primary
• Sitkin, See, Miller, Lawless and Carton — The Paradox of Stretch Goals (Academy of Management Review 36(3)), 2011. nonzero (typically, 10 percent) probability of at. View source · verified 2026-09-24 · primary
• Sitkin, See, Miller, Lawless and Carton — The Paradox of Stretch Goals (Academy of Management Review 36(3)), 2011. stretch goals are, paradoxically, most seductive for organizations that can least afford the risks associated with them. View source · verified 2026-09-24 · primary
• Ordonez, Schweitzer, Galinsky and Bazerman — Goals Gone Wild (Harvard Business School working paper 09-083; later published in Academy of Management Perspectives 23(1)). The saved copy is the working-paper draft and carries no year in its own text, 2009. managers and scholars need to conceptualize goal setting as a prescription-strength medication that requires careful dosing, consideration of harmful side effects, and close supervision. View source · verified 2026-09-24 · primary
• Butler, Zimmermann and Bird (Microsoft) — Objectives and Key Results in Software Teams (arXiv:2311.00236, November 2023; listed in the ICSE 2024 programme among the authors' contributions; the specific track is unconfirmed). Quotes verified against the arXiv preprint, which carries no date in its own text, 2023. more than 50% of managers reported that they are not very effec. View source · verified 2026-09-24 · primary
• Pavlickova, Mojzisova and Pocsova — Hoshin Kanri Process: A Review and Bibliometric Analysis (Processes 10, 1854), 2022. The healthcare area was very successful in the implementations of Hoshin Kanri due to having permanent employees overseeing the entire process. View source · verified 2026-09-24 · primary
• Butler, Zimmermann and Bird (Microsoft) — Objectives and Key Results in Software Teams (arXiv:2311.00236, November 2023; listed in the ICSE 2024 programme among the authors' contributions; the specific track is unconfirmed). Quotes verified against the arXiv preprint, which carries no date in its own text, 2023. tracking, measuring and setting goals is. View source · verified 2026-09-24 · primary
• Pavlickova, Mojzisova and Pocsova — Hoshin Kanri Process: A Review and Bibliometric Analysis (Processes 10, 1854), 2022. implementation in large companies had been successful and improved over the years through 5-year cycles. View source · verified 2026-09-24 · primary