An original deterministic toy model

An argument that earns its conclusion

“Measuring something helps improve it.” Sometimes. Give a fictional workshop a counter, a cost, and a temptation. Then make the claim survive its own arithmetic.

Set the workshop

One shift contains 100 work units. The true objective is useful pieces. Without measurement, half the units produce useful pieces: 50 useful pieces, also reported as 50.

With measurement, recording consumes some units. Of those left, some go to decorative marks that count on the dashboard but make no useful pieces. A signal directs the remaining work toward useful pieces.

Preset: Useful signal. Change any slider to test the assumptions.

0% routes half the productive work usefully; 100% routes all of it usefully. This is an assumed routing effect, not a measured accuracy or probability.
Units consumed by counting and attention, out of the original 100.
Share of remaining work diverted to dashboard marks. Each diverted unit earns one reported credit and zero useful pieces.

Fractional pieces are accounting units in this model. No random draws, fitted data, or hidden coefficients.

What actually improves?

Useful piecesProxy-only credits
True useful pieces
Reported credits
True change from 50

The arithmetic is the argument

Let q and g be fractions from 0 to 1, and c a cost from 0 to 100. These equations define this imaginary workshop; they are chosen assumptions, not scientific findings.

Remaining capacity A = 100 − c
Productive capacity P = A × (1 − g)
Useful routing share s = (1 + q) / 2
True outcome T = P × s
Proxy-only credits D = A × g
Reported outcome R = T + D

The no-measurement strategy keeps all 100 units and routes half usefully. Its true and reported outcomes are both 50. The measured strategy helps the true objective exactly when (100 − c) × (1 − g) × (1 + q) > 100.

If productive capacity P is positive, the equality boundary is q* = 100 / P − 1. Alignment above q* gives a true gain; below it gives a loss. If q* exceeds 100%, no available signal can compensate. If P is zero, the true outcome is zero and division is unnecessary.

The conclusion it can afford

Within this model, measurement helps when useful routing gains exceed recording costs and diversion to the proxy. “The dashboard rose” alone does not establish improvement: the dashboard includes credits that the objective rejects.

Real measurement might change learning, coordination, trust, or behavior in ways absent here. Real systems also require evidence about costs, routing and incentives. This page proves a conditional arithmetic claim and supplies counterexamples to an unconditional slogan. It does not estimate any real workshop, person, organization, patient, or financial outcome.

Source: original fictional model and direct algebra. All constants and scenario choices are visible above. No external sources or resources.