Blog

The Power of OKR Implementation: Cascading Corporate Strategy for Success

How OKRs connect strategy to the operating quarter: the difference between objectives, key results and initiatives, why strict cascading and bonus-linked targets break the method, and the cadence that keeps it alive.

3 September 2023

16 min read

The OKR hierarchy: strategic direction, annual objectives and quarterly team OKRs

Home

Blog

OKRs are among the most adopted and most abandoned management practices of the last decade. The adoption is easy to explain: the method is simple to describe, it has a credible lineage from Intel through Google, and it addresses a real problem, which is that strategy written at the top rarely changes what happens on Tuesday morning.

The abandonment is also easy to explain. Most implementations fail on a small number of specific mistakes, and those mistakes are predictable enough to be designed out in advance. This article covers the structure of the method, the distinctions that determine whether it works, and the failure modes that account for most of the abandonment.

What OKRs actually are

An objective is a qualitative statement of what you intend to achieve, written so that anyone reading it can tell why it matters. A key result is the measurable evidence that the objective happened, expressed as a change in a number between a stated start and a stated end.

There is a third element that belongs in the conversation but not inside the OKR itself. An initiative is the work you believe will move the number: the project, the campaign, the hire. Initiatives sit in the plan beneath the OKR.

Confusing key results with initiatives is the single most common implementation error, and it quietly hollows the system out. Consider the difference:

Launch the redesigned onboarding flow is an initiative. It can be completed on time while changing nothing about the business.
Raise seven-day activation from 34% to 50% is a key result. It can only be achieved if something actually improved.

A set of key results made of initiatives produces a system that reports activity while telling leadership nothing about outcomes. The test is simple: if the item can be marked complete by finishing work rather than by moving a number, it is not a key result.

OKRs and KPIs are not the same instrument

This distinction is worth stating plainly because conflating the two is what produces the fifty-item OKR list that nobody can act on.

KPIs are health metrics. They run continuously, they have acceptable ranges, and they tell you whether the business is operating normally. Gross margin, uptime, churn, cycle time, cash runway. You do not achieve a KPI; you monitor it.

OKRs are a change agenda. They name the small number of things you intend to move this quarter, and they are retired once the change has happened. A KPI drifting out of range is a good reason to open an OKR. It is not itself an OKR.

Organisations that convert their whole metric dashboard into OKRs end up with a system that has no priorities in it, which is the opposite of what the method is for.

The layers, and what belongs in each

Strategy operates on longer horizons than OKRs do, and it helps to name the layers separately rather than calling everything an OKR.

Strategic direction spans three to five years. It states where the company will compete and how it intends to win. This is strategy, not an OKR, and it should not be scored quarterly.

Annual objectives translate that direction into what has to be true by the end of the year. These are the bridge between the strategy and the operating cycle, and they are usually owned by the executive team.

Quarterly OKRs are where the method proper applies. These belong to teams, they are set for a single cycle, and they are scored and retired at the end of it.

Labelling a five-year ambition as an OKR is a category error with practical consequences: it cannot be scored, so it is never closed, and its presence on the list teaches everyone that items on the list do not have to be completed.

Why strict cascading fails

The word cascading has done real damage to this method. It suggests a mechanical process in which each level decomposes the level above and hands fragments downward.

In practice, strict top-down cascading produces three problems. It is slow, because each level must wait for the one above to finish. It produces objectives written to satisfy the level above rather than the customer, because that is who the author is writing for. And it discards local knowledge, because the people who understand what is actually achievable in a given system are the people operating it, not the people two levels up.

The healthier pattern is alignment rather than cascade. Leadership sets direction and constraints. Teams propose the objectives they will take on and how they will measure them. The two are reconciled in a short negotiation, and roughly half the final content originates below the executive layer. This takes longer to run the first time and produces targets people believe in, which is the only kind that gets hit.

Individual OKRs deserve particular scepticism. They make sense for senior individual contributors with genuine autonomy over outcomes. Applied to everyone, they generate enormous administrative load and convert a prioritisation tool into a performance management system, which brings us to the most consequential mistake.

Never tie OKRs to compensation

The moment a target determines someone's bonus or rating, the rational behaviour changes. The sensible move is no longer to set an ambitious target and pursue it honestly; it is to negotiate the easiest target that will be approved, and to report progress optimistically.

At that point the organisation has lost the thing that made OKRs worth running, which is an honest quarterly picture of what is and is not moving. Andy Grove kept the two systems apart at Intel. Google's own guidance repeats it. Any implementation that quietly routes OKR scores into the review cycle will get sandbagged targets within two cycles, and the leadership team will be the last to know.

Performance evaluation is legitimate and necessary. It should draw on a broader picture than a quarterly score, and it should run on its own track.

Committed and aspirational

Two types of OKR coexist, and problems arise when the type is left unstated.

Committed OKRs are expected to be delivered in full. Missing one is a failure that requires explanation and usually a change in resourcing.

Aspirational OKRs are set deliberately beyond what is known to be achievable. Reaching roughly seventy per cent is a good outcome. Their purpose is to force a different approach rather than a harder push at the existing one.

When the type is not declared, teams default to treating everything as committed. They then miss an aspirational target, are treated as having failed, and set conservative targets in the following cycle. Two rounds of this and the system produces nothing but safe numbers. Declare the type when the objective is set.

Scoring

Each key result is scored from 0.0 to 1.0 according to progress between its start and target values. The objective takes the average of its key results.

For aspirational OKRs, a consistent average around 0.7 indicates the calibration is right. Scores clustering near 1.0 mean targets are being set too low. Scores below 0.4 mean they are disconnected from capacity, which is a planning failure rather than an execution one.

The score exists to inform the retrospective. It is a diagnostic reading, not a grade, and treating it as a grade returns us to the compensation problem.

The operating cadence

The method is a cycle, and each part of the cycle does specific work.

Setting, in the two weeks before the quarter opens. Draft, reconcile across teams, resolve dependencies, name an owner for every key result. A key result without a single named owner will not be worked on.

Weekly check-in. This is where the system lives or dies. It should be short and should report confidence rather than narrate status: for each key result, a confidence level, what changed since last week, and what is blocking. A drop in confidence is a trigger for a decision, not an item to be noted.

Mid-quarter review. The point at which an objective can legitimately be dropped or rewritten because the situation changed. This matters more than it sounds. A dead OKR left on the list for appearances teaches everyone that the list is decorative.

Closing retrospective. Score, then examine the calibration and the quality of the key results themselves. Most first-cycle retrospectives find that half the key results were initiatives in disguise.

The failure modes, in order of frequency

Too many objectives. Three to five per team, three to five key results each. A longer list means the prioritisation never happened.
Key results that are initiatives. Activity dressed as outcome. The most common and the most damaging.
No baseline. A target with no starting value cannot be scored and usually reveals that the metric was not being measured before.
No named owner. Shared ownership is no ownership.
Set and forget. Written in week one, reopened in week twelve. Without the weekly rhythm the method is just a document.
Tied to pay. Covered above. It corrupts the data.
Cascaded to everyone. Administrative load rises faster than value.
Big bang rollout. First cycles are poor everywhere. It is better to absorb that in a pilot than in every department simultaneously.

When OKRs are the wrong tool

OKRs suit organisations that have more possible priorities than capacity and need a mechanism for choosing. They add little in three situations: teams small enough that everyone already knows the priorities, functions whose work is genuinely steady-state operational and better governed by KPIs and service levels, and organisations in the middle of a crisis, where the correct instrument is a short direct plan rather than a quarterly cycle.

Adopting the method because respected companies use it, without the underlying prioritisation problem, produces ceremony.

Introducing OKRs

Run one cycle with two or three teams. Accept that the first cycle will be poor, because it will: key results will turn out to be initiatives, targets will be set without baselines, and some objectives will be abandoned. Absorb that in a pilot.

Expect two or three cycles before the quality of the objectives justifies the meeting time. Organisations that launch everywhere at once usually abandon the method within a year and conclude that it does not work, when what failed was the rollout.

The point

OKRs are a forcing function for prioritisation, and they connect a strategy nobody argues with to a quarter somebody can be held to. They work when the number of objectives is small, when key results measure outcomes rather than activity, when direction comes from the top but content comes substantially from the teams, when the weekly rhythm is real, and when scores stay out of the compensation system.

They fail when any of those conditions is missing, and the failure is usually blamed on the method rather than the implementation.

At go:lofty we install measurement and goal systems as part of the wider growth architecture, connecting strategic direction to the operating cadence that carries it.

Talk to us about an OKR system that survives its second quarter.

Share the article

LinkedIn X

← Back to Blog

What is the difference between an objective, a key result and an initiative?

The objective is the qualitative statement of what you are trying to achieve, written so that anyone can tell whether it matters. The key result is the measurable evidence that it happened, expressed as a change in a number with a start and end value. The initiative is the work you believe will move that number. Objectives and key results belong in the OKR; initiatives belong in the plan beneath it. Confusing key results with initiatives is the single most common error, and it produces a system that tracks activity while telling you nothing about outcomes.

Should OKRs be tied to bonuses or performance reviews?

No. Coupling OKRs to compensation is the fastest way to destroy them. As soon as a target determines pay, the rational move is to negotiate an easy target, and the organisation loses the honest picture of progress that made the system valuable. Andy Grove, who created the practice at Intel, kept the two separate, and John Doerr repeats the point. Performance evaluation is a legitimate process; it should draw on a wider picture than a quarterly score, and it should run on its own track.

What is the difference between OKRs and KPIs?

KPIs are health metrics: they run continuously and tell you whether the business is operating normally. Uptime, gross margin, churn and cycle time are KPIs. OKRs are a change agenda: they name the small number of things you intend to move this quarter and are retired when the change is achieved. A KPI that has drifted out of acceptable range can become the subject of an OKR, but the two are not interchangeable, and converting every KPI into an OKR produces a list nobody can act on.

How many OKRs should a team have?

Three to five objectives at any level, each with three to five key results. The constraint is the point. OKRs work by forcing prioritisation, so a list long enough to include everything important has already failed at its job. If a team cannot reduce its list, the underlying problem is usually that nobody has been given authority to decide what will not be done this quarter.

Should OKRs cascade all the way down to individuals?

Rarely, and never mechanically. Strict top-down cascading is the best-documented failure mode of the method: it is slow, it produces objectives written to satisfy the level above rather than the customer, and it removes the local knowledge that makes targets realistic. Healthy practice sets direction at the top and lets teams propose how they will contribute, with roughly half the content originating below the executive layer. Individual OKRs are appropriate for senior individual contributors with genuine autonomy and are usually counterproductive elsewhere.

What does a good key result look like?

It states a metric, a starting value, a target value and a date, and it measures an outcome rather than an activity. Compare two versions of the same intent. "Launch the onboarding redesign" is an initiative: it can be completed while changing nothing. "Raise activation within seven days from 34% to 50%" is a key result: it can only be achieved if something actually improved. If a key result can be marked complete by finishing work rather than by changing a number, rewrite it.

What are committed and aspirational OKRs?

Committed OKRs are commitments the organisation expects to deliver in full; missing one is a failure requiring explanation. Aspirational OKRs are deliberately set beyond what is known to be achievable, and are scored well if they reach around seventy per cent. Both are legitimate, and problems arise when the type is left unstated, because a team will treat an aspirational target as a commitment and then sandbag the next round. Label the type when the objective is set.

How are OKRs scored, and what score should you aim for?

Each key result is scored from 0.0 to 1.0 based on progress between the start and target values, and the objective takes the average. For aspirational OKRs, around 0.7 is the healthy range: consistent scores near 1.0 indicate targets that were set too low, and scores below 0.4 indicate targets disconnected from capacity. Committed OKRs are expected at 1.0. The score is a diagnostic input for the retrospective, not a grade.

What cadence should an OKR cycle run on?

Quarterly setting, weekly check-ins, a mid-quarter review and a closing retrospective. The weekly check-in is where the system either lives or dies: it should be short, focused on confidence levels and blockers rather than status narration, and it should trigger a decision when confidence drops. The mid-quarter review is the point at which an objective can be legitimately dropped, which matters because a dead OKR left on the list teaches everyone that the list is decorative.

How should an organisation introduce OKRs for the first time?

Run one cycle with a small number of teams rather than launching organisation-wide. First cycles are almost always poor, with key results that turn out to be initiatives and targets set without knowing the baseline, and it is better to absorb that in a pilot than across every department at once. Expect two or three cycles before the quality of the objectives is worth the meeting time. Organisations that introduce OKRs everywhere at once usually abandon them within a year and conclude the method does not work, when what failed was the rollout.