Root cause analysis for business problems: how to test an explanation
An explanation that wins agreement in a meeting is not yet an established cause. Write down competing explanations, then look for records that support each one and counterexamples that would weaken it.
Three questions that test an explanation
What would we expect?
The record predicted by the explanation
What would challenge it?
A counterexample that weakens it
What remains open?
Unavailable evidence and the next test
In this article
Late deliveries, customer losses and increasing rework can arise from different processes. Before selecting a remedy, establish where, when and under what conditions the problem occurs.
Describe an observable problem
“The team is too slow” is an interpretation. “The time between order approval and shipment has increased for this product group” can be examined using records. Add the period, population, measurement method and comparison group to the initial description.
Record where the problem does not occur as well. Are delays concentrated in particular products, order sizes or types of work? Those differences help narrow the explanations worth examining.
ASQ's root cause analysis framework brings together clear problem definition, data validity and evidence-based selection of likely causes. ASQ, Root Cause Analysis.
Keep competing explanations open
Avoid turning the first explanation into an implementation plan before testing it. Capacity constraints, approval queues, material interruptions and a change in order mix might all explain a delay. Each begins as a candidate for investigation.
Prepare a short test card for each:
- If the explanation is correct, what should appear in which record?
- What counterexample would weaken it?
- Can we access the record, and is its definition reliable?
- What other explanation could produce the same symptom?
- Who will perform the next check, and by when?
A technique such as the five whys can help develop questions. Look for a record or observation at each step before treating the chain as established causation. A business problem may have several contributing causes.
Examine sequence and counterexamples
Suppose managers believe delays began after a staffing change. The timeline then shows a change in product mix during the same period. The coincidence alone cannot establish which explanation accounts for the delays.
Compare similar products and order conditions. Separate time waiting for approval from production time and time awaiting dispatch. A fast processing step does not necessarily produce a short end-to-end lead time. If the records cannot support that distinction, narrow the conclusion and state which explanations remain unresolved.
Turn the investigation into a decision-relevant finding
Give each candidate explanation a clear status, such as “supported”, “weakened” or “not yet tested”. Add the relevant record and its limitations. These labels describe the investigation's progress; they are not statistical confidence levels.
When a finding points to a process step, describe which outcome a proposed change is expected to affect. Use the same definition when checking what happens afterwards. Changes involving safety or technical suitability need assessment by the relevant specialists.
Preserve the unanswered questions
A useful diagnosis explains both the conclusion and why the evidence supports it. It also preserves counterexamples, gaps and next tests. Management can then judge which action has adequate support and which action would depend on an unresolved assumption.
For an important business problem with competing explanations, Verdict focuses on a comprehensive diagnosis from evidence to conclusion. Describe the problem you observe and the decision you need to make.