Process Boundaries and Value · Lesson 4

Selecting a process for improvement without mistaking a score for evidence

4 min read

Lesson objectives

  • Rank improvement candidates with stated impact, evidence and sensitivity limits.
Reading tools

A loud complaint alone does not select an improvement effort, and a high rate without a denominator or period does not either. First state the outcome the decision should protect: service arrival time, output accuracy, rework effort or beneficiary experience. Gather comparable data for candidates or label the comparison preliminary. A low-volume process may have severe impact per case, while a high-volume process can look worse simply because its count is larger. Record unit of analysis, period, log source and missing data before giving a score. Ask who can change the process and bear the pilot's effects. ‘Improvement is possible’ does not just mean the technical change looks easy.

A priority matrix exposes assumptions; its numerical result is not automatically objective. Name each criterion and the direction of its scale. If a larger risk score means more danger, do not add it to criteria where larger means better without reversing it or relabelling it as ease of risk control. Weights must total one and reflect the organisation's purpose. Calculate a weighted total by multiplying each score by its weight and adding the products. Do not combine rates with different denominators as though they were raw event counts. Where scores tie or nearly tie, test sensitivity: would a small increase in beneficiary-impact weight change the order? If so, seek distinguishing evidence or a small trial before committing widely.

A selection decision should name a candidate, testable reason, deferred alternative, missing fact and condition for revision. One matrix must never trade away a safety or privacy constraint: respond to a non-negotiable issue first, even when that candidate has a low routine score. Easy-to-measure processes should not automatically defeat high-impact processes with poor records; the first action for the latter may be to improve measurement. Separate priority to investigate from priority to deploy a known solution. A high return rate may justify studying its causes; it does not establish in advance that automation or additional staff is the remedy.

Lesson summary

Selecting a process for improvement without mistaking a score for evidence: apply the method to case evidence and check the result and its limits.