Goal Systems: OKRs and KPIs in Plain Terms

Why Process and Bureaucracy Exist showed how organizations encode the way work gets done so it doesn’t depend on any one person. But encoding the method isn’t enough. A thousand people can each be following perfect process and still be rowing in a thousand slightly different directions. Process answers “how do we do the work reliably?” It does not answer “are we all doing the same work, aimed at the same outcome?”

That second question is what goal systems exist to answer. This page is about the two tools almost every large organization now uses to aim itself — KPIs and OKRs — stripped of their jargon. By the end you’ll know exactly what each one is, how goals cascade from the top of the company down to your own desk, why that alignment is worth the trouble, and the single most important failure mode to watch for: what happens when a number stops describing reality and starts distorting it.

The problem: everyone busy, no one aligned

Recall the core problem of this whole part — coordinating people you’ll never meet. Structure decides who reports to whom and who owns what. Process decides how the work flows. But there’s a gap even those don’t close.

Picture a company of a thousand people where every single person is working hard, doing sensible things, following process. Is the company succeeding? You can’t tell. Hard work aimed in different directions doesn’t add up — it cancels out. The support team is optimizing for closing tickets fast; the product team is shipping features that generate more tickets; the sales team is promising things neither of them planned to build. Everyone is busy. Everyone is “productive.” And the company is going nowhere in particular, because no shared answer exists to the question: what are we actually trying to achieve this year, and how will we know if we got there?

A goal system is the mechanism that supplies that shared answer and pushes it into every corner of the organization. It’s how structure — which only tells people where they sit — gets a shared direction.

KPIs: the numbers you always watch

Start with the simpler of the two. A KPI — Key Performance Indicator — is an ongoing number you watch to know whether things are healthy. That’s the whole idea. Not a goal, not a target: a gauge.

Think of the dashboard in a car. The speedometer, the fuel gauge, the temperature warning — these run all the time, whether or not you’re on a trip. They don’t tell you where to go. They tell you whether the machine is running well enough to get anywhere at all. A KPI is a business gauge in exactly that sense: a vital sign you keep an eye on continuously.

Setting            A few of its KPIs (ongoing gauges)
-----------------  -----------------------------------------------
A cafe             Daily revenue, average wait time, waste %,
                   repeat-customer rate
A hospital ward    Bed occupancy, average length of stay,
                   readmission rate, patient-fall count
A warehouse        Orders shipped per hour, picking-error rate,
                   on-time dispatch %, inventory accuracy
A software team    Uptime, response time, active users,
                   support-ticket backlog

The defining feature of a KPI is that it’s continuous and directional, not finished. You never “complete” your uptime. You watch it. It lives at a healthy level or it drifts, and when it drifts you investigate. A good set of KPIs is chosen so that if all of them look fine, the operation is genuinely fine — they’re the small handful of numbers that actually indicate health, which is why the word is key. Track fifty numbers and you have a spreadsheet; track the five that matter and you have a dashboard.

In practice — a KPI that’s just noise

A warehouse manager proudly tracks “total items picked per day.” It climbs every month, and everyone feels good. But orders keep arriving late and customers keep complaining. The number was going up because staff were picking the easy, high-volume items first and leaving awkward orders to rot. “Items picked” was a real number, but it wasn’t a key indicator of the thing that mattered — orders fulfilled correctly and on time. The fix wasn’t a better dashboard; it was watching the right gauge. Half the skill of KPIs is choosing which few numbers genuinely reflect health, and ignoring the flattering ones that don’t.

OKRs: the ambitious goals for a period

An OKR — Objective and Key Results — is a different animal. Where a KPI is an ongoing gauge, an OKR is a goal you set for a specific stretch of time — usually a quarter or a year — and then close out. It has two parts, and the names are unusually literal:

• The Objective is a short, plain, slightly inspiring sentence about what you want to achieve. It’s qualitative and memorable. “Become the ward other hospitals visit to learn from.” “Make first-time buyers feel confident enough to come back.”
• The Key Results are the two-to-four measurable outcomes that would prove you hit the Objective. They’re numbers with targets, and crucially they measure the result, not the effort. “Cut average discharge wait from 40 minutes to 20.” “Raise first-month repeat rate from 15% to 25%.”

OBJECTIVE  (the ambition, in words)
   → Make first-time buyers confident enough to return.

KEY RESULTS  (how we'll know we got there, in numbers)
   → Raise 30-day repeat-purchase rate from 15% to 25%
   → Cut checkout abandonment from 30% to 18%
   → Reach a support-satisfaction score of 4.5 / 5

Two features make OKRs distinctive. First, the Key Results are outcomes, not activities. “Run three marketing campaigns” is not a Key Result — it’s a to-do list, and you could do all three and change nothing. “Raise repeat rate to 25%” is a Key Result, because it describes the world being different. This forces a team to be honest about whether their effort actually moved anything.

Second, OKRs are usually meant to be ambitious, even uncomfortable. Many organizations set them so that hitting roughly 70% is considered a good result — the target is a stretch you’re not sure you can reach, so that consistently scoring 100% is a sign you aimed too low, not that you’re excellent. (This isn’t universal, and it’s a frequent source of confusion, which we’ll come back to.)

The cleanest way to hold the difference in your head:

KPI                              OKR
-------------------------------  --------------------------------
Ongoing gauge you always watch   Time-boxed goal you set & close
Kept at a healthy level          Aimed at, then scored
"Are we still healthy?"          "Where are we trying to get?"
Uptime, wait time, error rate    "Cut wait time in half this quarter"

Notice that the two work together: an OKR often takes a KPI that’s sitting at an unhealthy level and sets a time-boxed push to move it. The gauge tells you something’s wrong; the goal is the deliberate effort to fix it.

How goals cascade

A goal system’s real power isn’t in any single goal — it’s in how goals connect from the top of the company down to one person’s week. This connecting is called cascading, and it’s how a strategy that lives in a few executives’ heads becomes something a warehouse picker or a nurse can act on.

It works top-down, level by level. The company sets a small number of high goals. Each division then sets its own goals whose job is to contribute to a company goal. Each team does the same beneath its division. And each person’s goals ladder up to their team’s. Done well, you can stand at any desk in the building, point at what that person is doing this week, and trace an unbroken line all the way up to what the whole company is trying to become.

COMPANY      "Become the most trusted brand in our market"
   │
   ▼
DIVISION     "Cut product returns by a third this year"
   │
   ▼
TEAM         "Reduce sizing-related returns on our top 20 items"
   │
   ▼
PERSON       "Rewrite the size guides and add fit photos
              for those 20 items this quarter"

The person rewriting size guides may never meet the executive who set the trust goal. But their work is connected to it, and — just as important — they can see the connection. That visibility is the point. It answers the most demoralizing question in any large organization: “does what I do all day actually matter to anything?” A working cascade turns “I rewrite product pages” into “I’m part of how we earn customer trust,” and that line is not spin — it’s literally traceable.

The evidence — line of sight to the goal

Research on employee engagement and goal-setting generally finds that people work with more focus and motivation when they can see how their own work connects to a larger purpose — often called “line of sight.” The precise size of the effect varies by study and setting, so treat any specific figure with caution, but the direction is consistent and unsurprising: work that visibly connects to something bigger is easier to sustain than work that feels like a disconnected chore. A cascade, at its best, is a line-of-sight machine.

The alignment payoff

Step back and see what the cascade buys. Structure, on its own, is static — it tells people where they sit, not where to go. A goal system is what makes structure point somewhere. It’s the difference between a machine that’s fully assembled and a machine that’s actually running toward something.

Concretely, a working goal system delivers three things a large organization can’t otherwise get:

• Shared direction. Everyone can name the same small set of things that matter this period, so effort adds up instead of cancelling out.
• Local autonomy without drift. Because the goal is explicit, teams can be trusted to figure out how to hit it themselves — you get the speed of decentralization without everyone wandering off. (This is the alignment side of the trade you met in Centralization vs. Decentralization: clear goals are what make pushing decisions to the edges safe.)
• A shared definition of “done well.” At the end of the period there’s an honest, agreed way to say whether it worked — not vibes, not the loudest voice, but the results you named in advance.

That’s the promise. Now the catch — and it’s a big one, because it’s the failure mode you’ll actually encounter.

Goodhart’s Law: when the measure becomes the target

Here is the most important idea on this page. It’s usually stated as Goodhart’s Law, after the economist Charles Goodhart, and the plain-English version is:

When a measure becomes a target, it stops being a good measure.

Why? Because the instant a number controls something people care about — their bonus, their standing, whether their team looks good — they stop trying to improve the reality the number was meant to describe and start trying to improve the number itself. Those two things sound identical but are not, and the gap between them is where organizations quietly rot.

A number is only ever a proxy — a convenient stand-in for something real you can’t measure directly. “Customer trust” is real but unmeasurable, so you track return rate as a proxy. As long as nobody’s optimizing the proxy directly, it tracks the real thing nicely. But make it a target with teeth, and people find the cheapest way to move the proxy — which is almost never the same as improving the reality underneath.

What you actually want   →   The proxy you measure   →   What people optimize
                                                          once it's a hard target
------------------------     ---------------------        ---------------------
Healthy patients             Short average stay           Discharge people faster
                                                          (even too soon → readmits)
Good customer service        Calls closed per hour        End calls fast, don't
                                                          actually solve the problem
Quality software             Bugs closed per week         Split one bug into five;
                                                          close trivial ones
Safe workplace               Reported-incidents = 0       Stop reporting incidents

Look hard at that last row, because it’s the nastiest. Push “incidents must be zero” as a target and you can get to zero two ways: make the workplace safer (what you wanted) or make people afraid to report (what you’ll actually get). The number improves while reality gets worse, and the metric now actively hides the danger it was installed to reveal. That’s Goodhart’s Law at its most dangerous — the measure doesn’t just stop working, it inverts.

The common trap — treating the metric as the goal

The everyday version of this trap is subtle. A team is told “get support-satisfaction scores up,” so they start begging customers for top ratings, or quietly closing tickets they couldn’t solve so they don’t count. Satisfaction scores climb; actual satisfaction doesn’t. The team isn’t being cynical — they’re doing exactly what the system rewarded. The fault is in confusing the scoreboard with the game. Whenever you set a metric as a target, ask immediately: “what’s the laziest way someone could move this number without doing the real thing?” If a cheap gaming path exists, someone under pressure will eventually find it — usually without meaning to.

Case study — the stretch-goal trap at Wells Fargo

Through the early 2010s, Wells Fargo pushed aggressive sales targets for the number of accounts and products sold per customer — a metric meant to stand in for “deepening customer relationships.” When the scandal broke publicly in 2016, it emerged that employees, under intense pressure to hit the numbers, had opened a very large number of accounts customers never asked for. The metric (accounts per customer) rose beautifully; the reality it was supposed to represent (customers being well served) was being actively harmed. The lesson isn’t “targets are bad” — it’s that a target with strong pressure and a cheap gaming path is a near-guaranteed way to get the number without the outcome. Treat any specific figures reported at the time as approximate; the mechanism is the durable lesson.

How to read your own company’s goals critically

You now have enough to do something most people never do: read your organization’s goal system with clear eyes, instead of just absorbing it. Four questions do most of the work.

1. What is actually being optimized — and what real thing is it a proxy for? For each headline metric, name the reality it stands in for. “We optimize average handle time.” What real thing is that a proxy for? Fast, good service. Does driving handle time down actually produce good service, or just fast service? The gap between the number and the thing is where trouble hides.

2. What’s being ignored because it isn’t measured? Every goal system spotlights a few numbers and leaves everything else in the dark. What matters but isn’t on the dashboard? Often it’s the slow, hard-to-count things: code quality, staff morale, long-term customer trust, mentoring. Work that isn’t measured tends to quietly starve, because the measured work always looks more urgent. Ask “what would degrade for a year before anyone noticed, because no metric watches it?”

3. What’s the cheapest way to hit these numbers without doing the real work? Run the Goodhart test on your own goals. For each target, imagine a tired, pressured person trying to hit it with the least effort. If there’s a gaming path — closing tickets without solving them, discharging patients too early, shipping features nobody asked for to hit a “features shipped” count — assume it’s being walked, at least sometimes.

4. Are stretch goals being read as stretch goals? If your OKRs are meant to be ambitious (70% is a win) but leadership treats missing 100% as failure, the system is quietly broken. People will respond by setting soft, easily-hit goals — “sandbagging” — and the whole point of the ambition is lost. Whether a goal system is healthy often comes down to this cultural detail: is missing a stretch goal a learning signal, or a punishment?

In practice — reading a dashboard critically

A nurse manager is handed the ward’s goal dashboard: length-of-stay down, bed occupancy up, readmissions flat. It looks like a healthy quarter. But running the four questions: length-of-stay is a proxy for efficient care — is it dropping because care got better, or because people are being sent home a day early (which readmissions, if honestly counted, would eventually reveal)? And what’s not on the board? Staff burnout. Patient anxiety. Nothing measures them, so nothing protects them. The dashboard isn’t lying; it’s just narrow. Reading it critically means holding both what it shows and what it can’t.

The goal isn’t to become cynical about metrics — the same warning the overview gave about the whole part. Goal systems are genuinely how large organizations turn a strategy into a thousand people’s aligned effort, and nothing else does that job as well. The skill is to use them while remembering, always, that the number is a finger pointing at the moon, not the moon. The moment anyone confuses the two, the system starts working against the thing it was built to serve.

How it connects

Goal systems are the aiming mechanism that sits on top of everything earlier in this part: they give structure a direction and make decentralization safe by aligning the edges. They also feed forward. Because hitting numbers affects status and reward, goals are one of the biggest engines of organizational politics — people fight over which metrics get chosen and how they’re read. And a goal system only works if it’s honestly communicated up and down the cascade, which is exactly what How Structure Shapes Communication examines next.

Try this

This week, get hold of your team’s or company’s actual goals — the OKRs, the KPI dashboard, whatever exists — and run the four critical questions on them. For each headline metric, write down (a) the real thing it’s a proxy for, and (b) the laziest way someone could move the number without improving that real thing. Then write down one important thing that isn’t measured anywhere. If you can’t find your company’s goals at all, that’s itself the finding: an organization with no visible goal system is running on hard work with no shared aim — and you’ve just located why so much effort there seems to cancel out.

Reflect

1. Can you name your company’s top few goals for this period and the handful of KPIs it watches continuously? If not, is the goal system unclear, or just not communicated down to you?
2. Pick one metric that affects how you or your team is judged. What real thing is it a proxy for, and how wide is the gap between moving the number and improving the reality?
3. Where have you seen Goodhart’s Law in action — a number that improved while the thing it measured got worse or stayed the same? What behaviour did the target quietly reward?
4. What matters in your work but isn’t measured by any goal or KPI? What’s slowly degrading because no metric is watching it?
5. Trace your own current work up the cascade. Can you draw an unbroken line from what you did this week to a company-level goal? Where does the line break, and what does the break tell you?

Show reflections

1. Many people find they can’t name either — which usually means the cascade broke somewhere above them and the goals never reached their level. That’s a real diagnosis: work with no visible aim is work that can’t add up. If you can name them, notice whether they’re KPIs (ongoing gauges) or OKRs (time-boxed ambitions); confusing the two is common.
2. The useful move is being honest about the gap. A small gap (the number closely tracks the real thing) means the metric is well chosen. A wide gap means you’re being judged on a proxy that can be moved without doing the real work — which is exactly where pressure produces gaming, often without anyone intending it.
3. Almost everyone has seen this once they look for it — sales targets met by pushing unwanted product, “tickets closed” hitting quota while customers stay unhappy, zero reported incidents hiding real risk. The insight is that the people gaming it usually aren’t villains; they’re responding rationally to what the system rewarded. Blame the target design, not the person.
4. This is the “what’s in the dark” question, and it’s often the most revealing. The unmeasured things — morale, quality, trust, mentoring — tend to erode silently precisely because no gauge sounds an alarm. Naming them is the first step to protecting them, even informally.
5. A clean, traceable line is the sign of a healthy cascade and a source of genuine motivation. A break in the line is diagnostic: either your work genuinely isn’t connected to the strategy (worth questioning), or the connection exists but was never made visible to you (a communication failure the next page addresses).