Skip to content

Turnover and Selectivity Assay

Measurement assay — instantiates Catalytic Pathway Enablement

Measures how many good cycles each facilitator unit actually delivers and how cleanly it hits the target versus off-target outputs — against a no-facilitator baseline.

A facilitator that looks like it is working is not the same as one that is earning its keep. Turnover and Selectivity Assay produces the evidence that tells them apart: it measures how many successful cycles each facilitator unit actually delivers per unit time and over its life (turnover), and what fraction of its output hits the target versus off-target or rejected results (selectivity) — always compared against a counterfactual baseline of what would happen without it. Its defining discipline is that counterfactual: without a with-versus-without comparison, ordinary process improvement, more resources, or a favorable case mix can all masquerade as catalytic effect. The assay's job is to characterize the facilitator honestly, so the numbers everyone else acts on are measured rather than assumed.

Example

An engineering org runs a shared internal data-pipeline service — the facilitator that turns raw event logs into analytics-ready tables for dozens of teams. Everyone believes it is a huge win, but nobody has measured it. The assay does. Turnover: each pipeline engineer's service handles ≈40 dataset onboardings a week. Selectivity: of those, roughly the large majority land correct-first-time, while a measurable tail need rework or land subtly wrong. Degradation: throughput sags as schema complexity climbs, so the per-unit rate is not constant over the service's life.

The decisive step is the baseline. The assay compares the pipeline against teams still doing the transformation by hand, and finds the real speedup is roughly sixfold — attributable to the service itself, not merely to those teams happening to have stronger analysts. That comparison is what stops the org from crediting the pipeline for a gain that better staffing would have delivered anyway, and it is what surfaces the rework tail before it grows into a trust problem.

How it works

  • Count successful cycles per unit. Turnover is measured as good conversions per facilitator unit per unit time, and tracked over the unit's life so degradation shows up rather than hiding behind a fresh snapshot.
  • Split target from off-target. Selectivity is measured across the whole output distribution — correct, reworked, and wrong — not just the successes, so the non-target tail is visible.
  • Compare against a counterfactual. Every figure is set against a without-the-facilitator (or alternative-pathway) baseline, so only the increment genuinely due to the facilitator is credited to it.

Tuning parameters

  • Measurement window — a short snapshot versus a full-life run. Snapshots are cheap but flatter a facilitator that degrades; long windows capture decay at the cost of time.
  • Baseline choice — no-facilitator versus a credible alternative pathway. What you compare against defines what "catalytic effect" even means here.
  • Selectivity granularity — target/off-target only versus the full output distribution. Coarse granularity is cheaper but hides where the non-target mass actually goes.
  • Sampling rate — census versus sample. Sampling scales to high volume but widens the uncertainty on the tail you most care about.
  • Degradation horizon — how far over the facilitator's life turnover is tracked before the estimate is refreshed.

When it helps, and when it misleads

Its strength is separating real catalytic leverage from impostors — selection bias, added headcount, or ordinary improvement — and surfacing selectivity decay before it becomes a public failure. It is the mechanism that lets an organization claim a facilitator works and actually mean it.

It misleads mostly through how it is scoped. A short window flatters a facilitator that degrades; a weak or absent baseline over-credits it; and selectivity measured only on the wins hides the off-target tail entirely. The classic misuse is running the assay after a facilitator has been chosen and measuring only its successes — turning a diagnostic into a justification. The discipline is to fix the baseline and window before measuring and to characterize the non-target distribution, not just the hits.[1]

How it implements the components

Turnover and Selectivity Assay fills the measurement-and-evidence side of the archetype — producing the numbers other mechanisms consume — not the rules or displays built on them:

  • turnover_capacity_model — it measures successful cycles per facilitator unit per unit time and over life, the operational bridge from a small facilitator quantity to real substrate throughput.
  • baseline_and_counterfactual_measure — it compares rate, quality, and side effects with versus without the facilitator (or against an alternative pathway), so catalytic effect is not confused with ordinary improvement.

It measures selectivity but does not enforce the selectivity rule that acts on the numbers (that is Fast Track with Eligibility Rules and the byproduct guardrail), display live saturation and queue state (that is Active-Site Capacity Dashboard), or detect contaminants and poisons (that is Inhibitor and Poison Screen).

Notes

The assay measures selectivity; it does not set the rule that acts on the result. That division matters: its numbers feed both the live Active-Site Capacity Dashboard and the deactivation decision, and keeping measurement separate from enforcement is what lets a team improve its assay — a better baseline, a longer window — without renegotiating the policies that consume it.

References

[1] Turnover number and turnover frequency — in catalysis, the total and per-time count of substrate conversions a single active site delivers before it deactivates; selectivity is the fraction of conversions that go to the desired product rather than to byproducts. Together they are the standard way a catalyst's real productivity is characterized rather than assumed. ↩