Finding the best explanation

The first plausible cause is a trap

Once a problem is defined, a cause suggests itself — usually the one the client already believes. Stage 2 asks for something stronger: rival causes, each tested against every fact in the case, and a choice made by stated criteria. Van Cleave calls the pattern inference to the best explanation: the cause to accept is the one that best explains all the evidence, judged on a set of explanatory virtues:

• Explanatoriness — does it explain all the facts, or only some?
• Depth — does it raise fewer new questions than it answers?
• Power — does it apply more widely than this one case?
• Falsifiability — could evidence show it wrong?
• Modesty — does it claim no more than the facts support?
• Simplicity — does it need fewer separate assumptions?
• Conservativeness — does it fit what is already known?

The first and the last two do most of the work in a business case.

Worked example — three rival causes for Lakeshore's late jobs

The facts to be explained: 14 breakdowns last year; repairs $9,000\to$38,000 in three years; on-time delivery 96 % → 84 %; complaints from the accounts began when the breakdowns became frequent; large-format work is where the lateness concentrates.

1. The press is failing. Explains every fact: the breakdown log, the rising repair curve, the timing of complaints, and the concentration in large-format work (that is the work the press does). Raises one new question — why now? — answered by its age. Simple: one cause. Conservative: old machines fail.
2. An operator shortage. Explains lateness and, perhaps, the complaints — but not the breakdown log or the repair curve, and no operator has left. It raises a new question (why would staffing make the press break?) with no answer. Explains some of the facts.
3. Poor scheduling. Explains lateness in general and could explain the concentration in large jobs. It does not explain fourteen breakdowns or the repair bill, and it requires a further assumption (scheduling got worse this year) with nothing in the case to support it.

Verdict: the press, on explanatoriness (all the facts), depth (fewer new questions) and simplicity. Scheduling is the runner-up — rejected not because it is impossible but because it explains a subset of the facts and needs an assumption the case does not supply. Write that down: the runner-up acknowledged and why rejected is a mark-bearing line.

Score two rival causes for the clinic

Facts: about 12 % no-shows; two double-bookings last month; eleven years on paper; two receptionists share one book and one phone line. Rival cause A — patients forget, because no reminder reaches them: explains the no-shows; does not explain the double-bookings. Rival cause B — one paper book, two people, one phone: explains the double-bookings (two hands in one book) and part of the no-shows (no reminder can be sent from a paper book) — every fact, one cause. B is the better explanation; A is a component of it. Notice that choosing B does not yet choose a fix — that remains open for stage 3.

Name the virtues cold

Cover the list and write the four that decide most cases: explains all the facts; raises fewer new questions; simpler; conservative. Then the three that guard against over-reach: falsifiable, modest, wider-applying. A cause that explains only the facts you happen to like — the vendor's explanation, the client's explanation — loses on the first virtue, and that is the one to check first.