Celestial Mode: Day

The auditor your IT department wishes it had.

Every spec is a claim. Every SLA is a claim. Every datasheet is a claim. Most of them are not measured against what is actually delivered. We do the measurement.

(Not software-defined networking. Different sense of the phrase.)

Fidelity, defined

Fidelity is the correspondence between a claim and its referent. A claim is what something says it is. A referent is what it actually is. Fidelity is high when the two are tightly coupled and degrades as they diverge.

A buyer’s exposure to a wrong decision equals the consequence of the decision multiplied by one minus the fidelity of the claim it was made against. The market over-invests in reducing consequence and under-invests in measuring fidelity, because there is no widely-adopted instrument for it.

MFid (Measured Fidelity) is the instrument.

Four dimensions, four questions

A claim can fail to match its referent in four distinct ways. Each is the answer to a different empirical question the buyer can ask.

Determinism

Does it behave consistently with the claim, across repeated observation?

Efficiency

Does it perform at the quantitative level the claim specifies?

Observability

Can the gap, if any, be seen by the buyer?

Intentionality

Does the design serve the purpose the claim states?

MFid = (D × E × O × I)1/4

Geometric mean of four normalized dimension scores in [0,1]. A geometric mean punishes weak links — a single broken dimension drags the whole score, because that dimension is a hole through which the buyer can be deceived regardless of the others.

Read the full theory →

The Porsche standard

Porsche routinely underrates its 911 horsepower. Independent dyno measurements aggregated at 992forum.com put the 992 Carrera S at roughly 487 hp against a claimed 443 hp — a ~44 hp delta in the buyer’s favor.

Source: https://www.992forum.com/threads/is-the-992-carrera-is-producing-more-power-than-advertised.169/ (retrieved 2026-05-28). A forum-aggregated figure is not a single instrumented run; the prose reflects that.

Under-promising is rare and instructive. It is the standard against which every claim should be measured: not “did the thing match what they said,” but “did the thing exceed what they said.” Most don’t.

We score ourselves first

Our own response times, our own resolution rates, our own promises — measured by the same framework, published to the same client. If our MFid slips, you see it before we explain it.

Live status →

Where to go from here

  • Theory — the framework: fidelity, the four dimensions, the math, the limits.
  • Methodology — the operational spec: evidence tiers, scoring procedure, version history.
  • Walkthrough — how MFid scores a real public SLA, demonstrated on Cloudflare.
  • Manifesto — the ethic: we don’t assert, we calculate.
  • Status — our own MFid, live.

Bring us a claim. We will tell you the number.

Vendor renewal on your desk, infrastructure refresh, platform decision? Score it before you sign.

Request an Investigation
Celestial Mode: Day