DTM.8:5 - Archetypal Grounding

DTM.8:5.1 - Fewer contacts and fewer errors are different changes

Consider a hypothetical network in which 20 teams already use procedure B. Each performs 10 runs per week. Under the stated conditions, a run has probability 0.25 of producing an erroneous result. The expected number of new erroneous results is therefore 50 per week.

The following actions address different targets:

Reducing recommendations may reduce new adopting teams. It does not by itself change the 200 weekly runs already performed.
Stopping half of the established runs, with all other assumptions unchanged, reduces expected generated errors to 25 per week. The needed work those runs served must still be considered.
An available independent review that catches 80% of generated errors before delivery reduces expected erroneous deliveries to 10 per week. It leaves 50 generated errors, consumes review effort and does not by itself change transmission of the procedure.
Correcting earlier erroneous results reduces the unresolved stock. It does not by itself change the current rate of new errors.

Combining the halving of established runs with review of every remaining result leaves 100 runs, 25 expected generated errors and 5 expected erroneous deliveries per week. This requires capacity to review all 100 results at the stated performance without displacing other needed work. If capacity permits only 50 uniformly selected reviews, expected detections are 50×0.25×0.8=10, leaving 15 erroneous deliveries rather than 5. Use OPS to resolve that capacity question.

The error probability and review performance are assumptions for illustration, not empirical claims. If review changes future learning or recommendations, that additional feedback belongs in the model.

DTM.8:5.2 - The same unresolved stock can hide a different source

Let L be the expected number of unresolved erroneous results, e the expected arrival rate and r the resolution rate per unresolved result. Under the illustrative approximation L’=e−rL, the stationary stock is e/r.

At e=50 per week and r=1 per week, the stock is 50. Halving e or doubling r gives a stationary stock of 25, but through different actions. In the latter case 50 new erroneous results still arrive each week.

The approximation assumes the resolution capacity supports a rate proportional to L. If a fixed-capacity correction team becomes saturated, that assumption fails; an OPS model is needed. The equal stationary stocks do not establish equal cost, delay, external consequence or further spread.

DTM.8:5.3 - Protecting exchange without rejecting a useful new method

A group receives work represented in a new format. Restricting admission of incompatible records can preserve the current exchange function while also obstructing a useful method.

Possible changes include a verified adapter, a more informative compatibility check or a bounded migration of the receiving work. Their subject engineering methods establish whether exchange succeeds. DTM.2/.6 examine how compatibility affects adoption; DTM.9 becomes relevant if the checking response itself changes useful participation.

Calling the new format an infection would conceal the actual problem. Continuing the old arrangement, converting at the boundary and changing the arrangement are alternatives with different costs and consequences.