Library / Method Engineering Principles Framework
Jump to passage
In this reading

Link to current text

Published source confirmed at last check

Source changed 2026-10-03 05:29:54 UTC · snapshot created 2026-10-03 05:30:57 UTC · last check 2026-10-03 07:20:10 UTC

ME.25:5 - Archetypal Grounding

These constructed cases demonstrate changes to reusable procedures. Their numerical assumptions are supplied for the calculation; using the candidates in a project requires the corresponding working facts.

ME.25:5.1 - Share preparation while preserving the receiving results

Two engineering analyses use the same fixed set of measurements. Each first converts the source into a common unit and coordinate convention, then derives its own result. Let f be the deterministic conversion, g the first analysis and h the second. The required pair is:

old(x) = (g(f(x)), h(f(x))).

The transformation obtains y=f(x) once and distributes that unchanged value:

new(x) = let y=f(x) in (g(y), h(y)).

The required results agree for every admitted x when the same conversion and input apply and neither consumer changes y. The proposed working rule is to prepare one shared converted dataset for this pair of analyses, identify the input and conversion used, and let both consumers read that result.

Suppose conversion takes 12 minutes, the two analyses take 3 and 5 minutes, and preparing access to the shared result takes 1 minute. Summed effort changes from 12+3+12+5=32 minutes to 12+1+3+5=21 minutes under those assumptions. This is an effort calculation; it does not determine calendar completion when people and tools can work concurrently.

Now change the situation. The second analysis receives corrected measurements. Reusing y from the old input no longer computes h(f(x_new)). The candidate’s reuse rule must identify an unchanged input and conversion, or recompute the affected preparation. The factorization remains correct; its former working precondition has failed.

In another use, the two original preparations were independent measurements rather than repeated deterministic conversion. Sharing one measured value removes that independence. If the measurement errors are independent and each has variance s^2, averaging the two measurements has error variance s^2/2. Copying one measurement twice and averaging it retains variance s^2. The apparent duplication cannot be eliminated under a requirement for the former error variance. MMP.18 recovers the shared dependence; the mathematical function account must be changed before the work is redesigned.

ME.25:5.2 - Change the order of checks without changing the required decision

An administrative procedure accepts a case only if checks A and B both pass. The checks do not change the case or each other’s outcomes. Either may be performed first, and the current requirement permits stopping after the first failure. A takes 6 minutes and passes 90% of cases; B takes 2 minutes and passes 50%. These are supplied proportions for the same incoming population, unaffected by check order. Times are fixed.

With A first, expected effort is 6+0.92=7.8 minutes. With B first it is 2+0.56=5 minutes. Both return the same acceptance decision for every case. The candidate rule is to do B first and run A only after B passes.

For two such checks i and j with fixed costs c_i,c_j and pass probabilities p_i,p_j, placing i first has no greater expected cost when:

c_i + p_i*c_j <= c_j + p_j*c_i,
equivalently c_i*(1-p_j) <= c_j*(1-p_i).

For this two-check comparison, independence of their outcomes is unnecessary: the second check is incurred exactly when the first passes. Extending one fixed ordering rule to many checks needs the conditional probabilities among cases reaching each position; marginal ratios alone can fail when those probabilities change.

Now require the procedure to report every failed condition so that the applicant can correct the case in one return. Stopping after the first failure no longer supplies the required result. Both checks must then be completed, taking 8 minutes of summed effort under the same assumptions. A changed order may still affect the time of an early indication, but it no longer produces the claimed effort saving.

If the requirement instead remains the first-failure decision but A supplies information needed to perform B, the proposed order is unavailable. Add the preparation that would make B independently executable and recompute its cost, or retain A first. A lower algebraic value is not a usable method while its required input is unavailable.

The changed reusable screening rule belongs to method design. Scheduling a particular person’s A check on Tuesday under the unchanged rule would be a Work-planning result.