Photo by Mason on Unsplash

One of the best programs I ever built had no defensible methodology.

It was a recognition application developed for a sales and customer service team in a former role. Every manager received a monthly budget of points and the discretion to award them: for staying late, for covering someone else’s shift, for an unusually good comment from a customer, or an improvement idea worth implementing. Agents accumulated balances and spent them on things that were not money; usually swag, small appliances, or other gadgets. There was no normalization. A generous manager inflated the currency, while a stingy or less enthusiastic one starved it. Nothing in the design would have survived a serious audit of its assumptions.

It worked, if unmeasurably. That is most of the point.

The other program I built around the same time was a scorecard. Fully standardized metrics, consistent assumptions, careful construction and validated KPI weights. It was intended to drive coaching and iterative improvement, and it was reviewed by everyone who should have reviewed it. It also shipped on time and to specification.

What it became was something else. Because it could not hold nuance, it concentrated attention on the bottom of the distribution, where improvement was often most expensive and least voluntary. It offered nothing to the people at the top, who learned nothing from an instrument built to surface gaps they did not have. Coaching conversations turned into negotiations about the number rather than the work behind it. And somewhere in the second year, the scorecard stopped being a tool for development and became a description of the person.

The project closed exactly as planned.

Both programs served the same population, under the same leadership, in the same years. The rigorous one made the following year harder; the sloppy one made it easier.

And nothing in either project review, whether the scope, the schedule, or the acceptance criteria, was capable of telling you which was which.

This week, I want to explain why that happened, and then why the review could not have caught it.

Rotation Is Free

Picture a point on the rim of a flywheel. Trace it through one full turn. In the plane of the wheel, it ends precisely where it began.

Add displacement along a third axis and the circle becomes a helix. In organizational life, time supplies the rotation, while the chosen outcome defines the vertical axis.

Time is the one input nobody has to requisition. The month closes on its own, as do the quarter and the year. The planning cycle reopens without anyone deciding that it should. Every business runs on revolutions of this kind: monthly close, quarterly review, annual budget, release cadence, pipeline review, or performance cycle.

The wheel turns whether or not anything was accomplished on it.

This is the source of a specific and common illusion. An organization can feel as though it has momentum simply because events keep arriving on schedule. Meetings occur. Reports generate. Deliverables land. The apparatus is plainly in motion, but motion is easy to mistake for direction.

The part we sometimes miss is that repetition is guaranteed and accumulation is not. A completed cycle is not evidence of progress. It is only evidence that time passed.

What One Revolution Contains

Before anything can be called upward movement, something has to define which way is up, and there is no context-free positive slope. Reducing delivery time can be upward on speed and downward on quality. Increasing customization improves customer fit but often damages scale. Cutting headcount lowers cost and lowers capacity. Every claim of progress is a claim relative to a chosen outcome, and an organization that has not chosen one cannot evaluate direction at all. It can only evaluate whether things happened.

Assume the outcome is chosen. What determines how much of a cycle’s activity becomes movement toward it?

For today, we’ll point to six things, worth separating because organizations habitually collapse all six into a single word: effort.

Effort deployed – not effort available. Capacity that exists and is not committed produces no motion, and counting it is the most common form of planning fiction.

Efficiency – the fraction of deployed effort that survives rework, waiting, handoffs, and duplication.

Applicability – the fraction of surviving effort that bears on the chosen outcome at all. Work can be executed flawlessly and be about something else.

Alignment – the direction of that effort relative to the outcome. Toward it, orthogonal to it, or against it.

Against this baseline, we must also track two subtractions. Coordination cost, which can grow faster than the operation does, because the number of relationships requiring management grows faster than the number of people. And drag – arriving through interruption and unplanned obligations, the standing tax that does not scale down when the team does.

Progress = Effort × Efficiency × Applicability × Alignment − Coordination − Drag

Two properties of that expression matter more than the expression itself.

The first is that the relationship is multiplicative rather than additive. Any term near zero collapses the product regardless of the others.

Efficiency cannot repair irrelevance; a team executing perfectly against work with no bearing on the objective produces a very clean nothing. This is why operational metrics so often reassure without informing. Closing tickets faster, shipping more features, running more reports – each establishes that the motion was economical, but alone none establishes that it mattered.

The second is that the terms are not independent, and the dependency runs in an inconvenient direction. Misaligned work frequently shows inflated efficiency. Nothing is contested, so nothing is reworked. Nobody sent it back, because nobody was waiting on it.

The smoothest-running workstream in a portfolio is sometimes the one that has stopped mattering to anyone.

There is one further consequence, and it is the bridge to everything else here. Because coordination cost can accelerate faster than productive output, there is a point beyond which additional effort reduces net progress. Further, that point is not fixed; it moves with alignment.

The better an organization understands its own direction, the more effort it can usefully absorb; the worse it understands it, the less. In the limit, an organization that cannot say which way is up has an optimal deployment of approximately nothing.

Which means these are not efficiency knobs. They are not marginal adjustments to a fixed engine. They set the ceiling on how much effort can be productively spent at all.

The Second Ledger

Every piece of work produces two results.

The first is displacement: did the chosen outcome get closer during this cycle?

That is what a project delivers, what a status report tracks, and what a closeout review evaluates.

The second is a change to the terms above. Is next cycle’s efficiency different because this cycle happened? Its applicability, its alignment, its coordination burden, its drag? Did the work leave the organization more capable of moving, or less?

Usually, only the first is scored, because the unit of evaluation is the revolution, and the coefficient effects arrive after the revolution has closed. They may land next quarter, or in another team’s queue. Worse, they may land in a maintenance obligation that nobody attributes back to the decision that created it; by then the decision has a completion date and a green status, but no remaining owner.

The consequence runs in both directions, and it is the same error each time. Foundational work looks like overhead, because its displacement is small and its coefficient change is unscored. Debt-creating work looks like success, because its displacement is real and its coefficient damage is unscored. An organization does not need bad judgment to fund the second and defer the first. It only needs to be evaluating one cycle at a time.

This is also where “technical debt” stops being a metaphor and becomes a description. It is not borrowing, exactly; nobody agreed to terms. Instead, it is negative coefficient change on a cycle that has already been graded, and the reason it accumulates is not that engineers are careless. It is that the accounting closed before the bill arrived.

This was not primarily a Goodhart problem. The scorecard was never obviously gamed, and the measure continued to report what it had been designed to report, accurately, for years. Its effect on subsequent cycles simply ran opposite to its stated purpose, and no instrument in the review process was pointed in a direction that could have seen that. The larger failure was in the accounting, not the metric.

Delivery accounting is per-revolution. Strategy is cross-revolution.

An organization that grades only project by project can systematically fund the wrong half of its portfolio – often precisely because each individual decision is graded rigorously.

There is a budget asymmetry that follows directly. The recognition program was a discretionary spend line, defended annually, and easy to characterize as morale expense. The scorecard, however, was infrastructure.

In any cost exercise, the thing improving the coefficients is cut first and the thing damaging them is maintained. Both decisions are entirely defensible inside a single revolution.

The Effort Required to Stand Still

Extend this over several cycles and it stops being an accounting curiosity.

If each revolution degrades the terms slightly – efficiency down a little, drag up a little, coordination heavier by one more dependency – then holding output constant requires continually rising effort. The organization does not experience this as decay, but instead sees scope growth, a staffing shortage, or a vague sense that a team which used to be fast has become careful instead.

So the reframe is this: a team requesting additional headcount every year against flat output is not necessarily failing at execution. It may be paying interest on cycles that were scored before the bill arrived. More effort may temporarily restore the output, but without addressing what was previously built, it will not stop the compounding.

The positive direction is identical in both structure and invisibility. A definition agreed upon once (what a customer is, what counts as active, when a period ends) raises alignment for every project that follows it, permanently, without appearing in any of their business cases. Automated ingestion lowers drag every subsequent cycle after being built once. None of this generates a line item on the next round, but all of it changes the shape of every revolution afterward.

I should be clear that I did not observe this from a distance. I built both programs. My title at the time contained the word improvement, and I was measured, correctly and unremarkably, on delivery.

The scorecard was the more impressive of the two by every standard I had then. Harder problem, cleaner methodology, better documentation, more stakeholders satisfied at launch. It is also the one that made the following year harder for the people it was built to help.

The part I did not see until much later was that the scorecard was itself aimed at the second ledger.

It was built to drive coaching and iterative improvement; that is coefficient work by definition. It was scored entirely on the first ledger: delivered, on time, to specification, adopted. Nobody ever went back to ask whether coaching quality had improved.

It was a project aimed squarely at the coefficients, graded exclusively on displacement, and closed green.

What Is Up, and What Is Left Behind

The standard question before execution is what will this deliver. It is a reasonable question but it is not sufficient, because it can only ever interrogate one of the two ledgers.

The better question is: what will be true about the next cycle because this one happened?

Nearly everything worth asking in planning falls out of that question rather than having to be listed beside it. You cannot answer it without having defined the outcome specifically enough to distinguish closer from farther. You cannot answer it without knowing what effort is genuinely available rather than nominally assigned; the same specialist appearing at full value in four plans makes each plan locally plausible and the portfolio collectively impossible. You cannot answer it without the participants sharing a direction, because effort applied by people who understand the objective differently does not merely arrive slowly; some of it arrives pointed the wrong direction. And you cannot answer it without a traceable path from the activity to the change.

The wheel is going to turn either way. Time supplies the rotation regardless, and every cycle will produce a full complement of meetings, deliverables, spent budget, and completed work.

What you control is whether the cycle ends higher – and whether the next one begins with an advantage. A revolution that delivers everything asked but leaves the one that follows unchanged does not compound. At best, it is a very well-run circle.

Addendum: For People Who Like the Math

The argument above stands on its own. What follows is the model underneath it, for readers who find that sort of thing enjoyable and are willing to accept a few honest caveats about what it can and cannot do.

The five questions, in operational form. If you want the checklist rather than the essay:

1. Define the axis – what outcome counts as upward?
2. Establish usable effort – what capacity remains after competing obligations?
3. Align the participants – do the people involved understand the objective the same way?
4. Trace applicability – how does each major activity affect the outcome, or a coefficient that governs it?
5. Evaluate the second ledger – what will be easier or harder in the next cycle?

Author’s note:

This model began as a mental image, not an equation: a wheel completing one revolution for each fixed unit of time. I began asking what would have to change for that repeated motion to accumulate rather than return to its starting point, and how effort, alignment, coordination, and drag would alter the path.

The math came afterward. It does not generate or prove the argument; it tests the image for consistency, makes its assumptions visible, and shows what follows if those assumptions are accepted.

~Dom

One cycle. Let u be a unit vector pointing toward the chosen outcome. For each activity i, let dᵢ be a unit vector representing its direction, with effort Eᵢ, efficiency ηᵢ ∈ [0,1], and applicability qᵢ ∈ [0,1]:

aᵢ = u · dᵢ        (alignment, ranging −1 to 1)
Δz = k · Σᵢ (Eᵢ · ηᵢ · qᵢ · aᵢ) − C(E) − D

Δz > 0 means closer to the outcome. Δz < 0 means farther. Alignment is a projection: aᵢ = cos φᵢ, where φ is the angle between the activity and the objective. Treat that as an illustration of projection, not a claim that anyone can measure an organizational angle to two decimals.

Why an optimum exists. Coordination cost can become superlinear when expanding effort introduces dependencies faster than the organization can modularize them. The number of possible pairwise relationships grows as n(n−1)/2, though only some of those must actively be managed – hierarchy, standardization, and clean interfaces exist precisely to keep the managed set small. The convenient quadratic form C(E) = cE² illustrates the case where that containment fails. Note also that this justification is stated in people while E is defined as effort; the form holds cleanly only where effort is roughly proportional to the number of independently coordinated actors.

Δz = k·E·η·q·a − cE² − D
∂Δz/∂E = kηqa − 2cE
E* = kηqa / 2c

The optimal effort is proportional to alignment. As a → 0, E* → 0. And if a < 0, then ∂Δz/∂E < 0 – every additional unit of effort makes the result worse. Scale does not correct misalignment; it accelerates it.

Shared capacity. Effort available to the focal objective is what remains after everything else claims its share:

E_focal = E_total − E_competing − E_reserved − E_unplanned

Fragmentation then hits twice: capacity is allocated elsewhere, and the remainder loses efficiency to context switching, η_eff = η₀(1 − s). A person can be productive across four priorities and unavailable for all of them.

The recursion, which is the actual thesis.

z(n+1) = z(n) + Δz(n)
(E, η, q, a, C, D)(n+1) = F[ (E, η, q, a, C, D)(n), work performed in cycle n ]

A cycle compounds when it satisfies both Δz(n)>0 and E[Δz(n+1)]>E[Δz(n)]. A cycle may still be strategic when it sacrifices present displacement to improve the expected displacement of later cycles.

Foundational work maps cleanly onto specific coefficients: defining the objective and standardizing terminology raise a; establishing measurement methods raises q; automated ingestion and reusable tooling raise η and lower D; documented decisions and removed handoffs lower C; retiring obsolete processes returns E.

This also supplies a usable definition of applicability that does not degrade into “it felt foundational” over time. An activity is applicable if it can be shown to raise E, η, q, or a, or to lower C or D. If it does none of those, it is occupying the cycle.

Three non-decorative caveats.

k is unidentifiable. Progress toward an outcome has no natural, general unit, so the conversion constant cannot be pinned down. The model therefore produces comparisons within a fixed objective – across initiatives at one moment, or one initiative across cycles – but never comparisons across objectives.

The underlying terms are treated as cardinal in the formal model, but in organizational practice they are usually observable only ordinally. Nobody can defend $a=0.6$ rather than $a=0.8$, though most teams can agree on sign: toward, sideways, or against. The mathematics therefore describes the relationships among the terms; it does not make their values directly measurable.

Finally – do not build an index from this. The moment these become a composite score in a slide, the estimates become arbitrary decimals pretending to be measurement, and the model starts concealing the assumptions it was created to expose.

Its only legitimate (and intended) use is to make an argument possible: which term is actually limiting us, and what would change it?