Goals Were Never Meant to Be Tasks

A planning meeting at a software company. The manager opens: "Your goal this quarter is to ship the new onboarding flow by end of September."

Three weeks in, the squad's research surfaces something the goal didn't anticipate. New users aren't bouncing during onboarding. They're bouncing two weeks in, after their first failed task in the product. The squad lead raises this with the manager.

The reply is calm and reasonable: "Let's not lose focus on the goal. We can do another customer round next quarter."

The squad ships the onboarding flow on time. New users continue to bounce two weeks in. Adoption doesn't move. The squad spends the next quarter wondering what went wrong with their work.

What went wrong with their work was that their work was never going to address the actual problem. The direction was wrong. The manager wouldn't let them change it. The squad had a target, no authority to revise it, and an evaluation waiting at quarter-end that would only check whether the feature had shipped.

That's not a goal. That's a task, evaluated as a task, with the word "goal" attached as decoration.

What a goal actually requires

A goal has three structural parts. All three are mandatory. Drop any one and you no longer have a goal.

A trend. The desired direction of movement. Without it, you cannot tell whether what happened was desired. "Be at 8%" tells you nothing if you do not know whether less is better or worse. The trend is the direction the work is supposed to move outcomes in.

A target. The wished magnitude that gives the trend a measurable shape. Without it, you cannot tell whether one point of movement was sufficient or trivial. "Reduce churn" alone is an aspiration, not a goal: there is no number to evaluate against. A real goal has to commit to a magnitude, even if the magnitude is the worker's current best estimate of what good would look like.

An evaluation that respects both. At the moment outcomes are assessed, both the trend direction and the target magnitude have to be checked, alongside the learning that produced whatever happened. Did we move the right way? Did we approach the target? What did we discover about whether the target itself was right? Drop any of these questions and the trend has been stripped from the work.

A task, by contrast, has only a specified state to reach. Reach it or fail. Evaluation is binary attainment.

The decisive distinction is structural, and it is determined at evaluation. Setting can look the same in either case. The difference shows up when the work is assessed.

Drucker knew this

Peter Drucker introduced Management by Objectives in The Practice of Management (1954). The chapter title was "Management by Objectives and Self-Control." Self-control was not stylistic. It named the structural condition under which goal-evaluation actually works: the worker measures their own performance against the target. The three-part decomposition above (trend, target, evaluation that respects both) is my reading of what self-control structurally requires; Drucker named the principle, not the decomposition. The principle is his. The reading is mine.

Without self-control, the worker is being evaluated by someone else against a single number, which is the shape of task evaluation regardless of what the work is called.

As management writer David Hurst documents in his analysis of MBO's failure modes, the self-control aspect was "lost in translation from a German cultural context (shared by both Drucker and Follett) to an American one," where MBO morphed into a six-step, top-down process. Drucker's emphasis on self-discipline got swamped by his injunction to "set clear objectives," and the objectives won. Seventy years later, what remains in many organizations is target-attainment evaluated from above, with the trend dimension stripped from the assessment.

The corruption, in other words, is not linguistic. It happens at evaluation.

How evaluation got stripped

Goal-evaluation did not survive contact with the modern firm. It eroded through four stages. Each was reasonable in its own context. Together they hollowed the assessment out.

Self-control evaluation. In Drucker's original frame, the worker measured their own performance against the target. That measurement included the trend (are we moving the right way?), the magnitude (did we approach the number?), and the learning (was the target itself the right target?). The worker had context the cascade did not have. Self-control was not an empowerment slogan; it was an evaluation requirement.

Evaluation centralized. As organizations grew and the goal cascade extended, evaluation moved upward. The worker no longer evaluated themselves; managers did. Managers a level or two removed had less context for the trend dimension. They could see whether the number was hit. They could not easily see whether the direction was right, or whether the target had been the right target, or what the worker had learned.

Evaluation simplified to attainment. Performance review, bonus calculation, and board reporting needed legible signals across many positions. The most legible signal was binary target attainment: hit or miss. The trend dimension required interpretation, conversation, and worker context, none of which traveled well through the cascade. The number did. Over time, evaluation reduced to the one signal that scaled.

Lock-in. Compensation, performance review, and OKR cascade infrastructure all built around binary attainment. The trend dimension is now structurally absent from how outcomes are assessed in many modern organizations. The Objective declares a trend; the Key Result specifies a number; the evaluation, at quarter-end, asks one question: did you hit it? The trend lives in the words. The target lives in the evaluation. Whether the original goal was set well or set as decoration no longer matters, because the evaluation has the same shape either way.

The evaluation cannot go backward. Each stage created stakeholders who benefit from the simplified assessment: the manager whose dashboard reports green when numbers are hit, the executive who can roll legible signals up to the board, the OKR vendor whose product depends on attainment math, the HR system that depends on hit-or-miss thresholds. Restoring trend evaluation threatens the simplicity of the cascade; it requires interpretation that the infrastructure does not support and the calendar does not provide.

This is why the speech-act dressing (managers saying "your goal" instead of "your task") is a downstream symptom rather than the cause. The vocabulary follows the evaluation, not the other way around. Once the evaluation is binary attainment, every prescription will be presented in whichever language the firm currently rewards. In modern firms, that language is goal-grammar. The word does not change the shape of the work; the evaluation does.

What the research actually says

The research on autonomy in knowledge work has been consistent for decades. Self-Determination Theory (Deci and Ryan, fifty years of work across organizations, education, and healthcare) treats autonomy as one of three fundamental psychological needs behind intrinsic motivation. Autonomy here means experienced volition over the work, not permission to choose how to execute someone else's prescription. Daniel Pink's Drive splits autonomy into four dimensions that matter for complex work: task (what), time (when), technique (how), and team (with whom). Modern work-design typically grants the third and withholds the first.

In plainer language: the team that can change the direction when the direction is wrong outperforms the team that can only execute faster on a wrong direction. Hackman and Oldham's job characteristics model treats autonomy as one of five core dimensions that drive motivation and performance.

The infrastructure inverts what the research recommends. The OKR cascade, the goal-rollup, the performance review, the bonus structure all assume top-down target-attainment as the unit of evaluation, with the trend dimension stripped before the assessment runs. The cascade reports green: the numbers were hit. The trend question, the magnitude judgment, and the learning the worker accumulated never enter the reporting. The structural fact that direction-evaluation lives nowhere in the assessment goes unreported because the assessment has no slot for it.

What this produces is reporting without truth-telling. The worker reports goal-progress while privately knowing the goal is wrong. Multiply this across an organization of ten thousand people, and the dashboard at the top shows a detailed picture of work that is not addressing the problems the work was set up to address.

How to tell

The cleanest test is to look at how a "goal" gets evaluated, not how it was set or presented.

Three questions to ask of any "goal" in your organization's current cycle:

Direction check. Does the evaluation ask whether outcomes moved the right way, or only whether the number was hit?

Magnitude check. Is the target treated as a current best estimate of where the trend should land, or as a binary attainment threshold whose miss equals failure regardless of context?

Revision check. Can the target be revised mid-cycle when learning suggests the target was wrong?

If all three checks happen, the work has goal-shape. If only the magnitude check happens (binary attainment, no direction inquiry, no revision authority), the work has task-shape, regardless of what verb was used to set it or what word preceded it.

Equivalent shortcut: do you have authority to revise the target mid-cycle when learning says it was wrong? Yes = goal. No = task.

A mid-cycle observable

The same distinction shows up halfway through the quarter, in the question that gets asked at the check-in.

For a goal, the mid-cycle question is: what is helping us, what is blocking us? The check-in is open-ended. New information is input to revising direction. The path is unfolding; adjustments are part of the work.

For a task, the mid-cycle question is: where is the plan? The check-in is closed-ended. New information is deviation from schedule. The path was decided up front; the work is adherence.

If you cannot remember the last time your "goal" check-in asked what was helping or blocking the trend rather than where the plan stood, the work has been task-shape all along. The mid-cycle behavior reveals what the end-of-cycle evaluation will check. You do not need to wait for quarter-end to know.

A pre-check at the speech act

Some "goals" announce themselves as tasks before the evaluation question even comes up. State-verbs at the Objective layer (ship, migrate, win, hit, launch, build, become, achieve) have no trend in their grammar. They name states, not directions of movement. There is nothing to evaluate as a trend even in principle. State-verb at the Objective = task before evaluation enters the picture.

For example:

• "Your goal is to ship Feature X by end of Q3."
• "Your goal is to migrate the platform to Kubernetes by Q2."
• "Your goal is to win three enterprise deals this quarter."
• "Your goal is to build a healthy team."

The verbs (ship, migrate, win, build) name states. There is no trend in the grammar of any of these statements. The word "goal" attached at the front is decoration. Task at the speech act, before evaluation enters the picture.

Trend-verbs (reduce, improve, expand, strengthen, accelerate) keep the question open. The setting is goal-compatible. Whether the work is actually a goal depends entirely on how it gets evaluated, and that is where the three-question test above applies. "Reduce customer churn from 12% to 8% by Q3" can be a real goal or a task in trend-grammar. The number does not settle the question. The evaluation does.

Pick a goal and run the diagnostic

The pattern is not specific to software. Sales quotas evaluated as targets-hit. Marketing lead-volume targets evaluated as numbers-rolled-up. HR engagement scores evaluated as score-attainment. NPS targets, which I wrote about previously, are part of the same family: trends declared at the Objective layer, numbers specified at the KR layer, binary attainment at evaluation.

Pick the most prominent "goal" in your current cycle. The one whose attainment is on your performance review.

Ask the three questions:

1. Will the evaluation at quarter-end check whether the trend moved the right way, or only whether the number was hit?
2. Is the target a current best estimate of where things should land, or a hit-or-miss threshold whose miss equals failure?
3. Can the target be revised mid-cycle if learning says it was wrong?

If the answers all sit on the goal-side, you have a goal. If any answer sits on the task-side, you have a task, regardless of how the work was set or what word preceded it.

Equivalent shortcut: do you have authority to revise the target mid-cycle when learning says it was wrong? Yes = goal. No = task.

If you want real goals in your organization, do not fix the words. Fix the evaluation. The vocabulary follows the assessment.