What Else Could Explain This?

The problem with the first explanation

When something surprising happens, an explanation usually arrives before you have asked for one. It arrives fast, it fits the available facts, and it feels less like a hypothesis than like a perception.

The difficulty is that “fits the available facts” is a weak test. Several explanations usually fit. The first one to arrive has an unearned advantage: everything you learn afterwards gets evaluated against it, and evidence that fits it also fits the alternatives you never generated.

The remedy is not to doubt your first explanation. It is to write down the others before you start weighing evidence.

A four-step procedure

This is a compressed, everyday version of the structured technique known in analytic tradecraft as analysis of competing hypotheses. The full method is designed for teams working on high-stakes questions over weeks. The compressed version takes about fifteen minutes and captures most of the benefit for ordinary decisions.

Step 1: List the evidence, separately from any explanation

Write down what you actually observed, one item per line. Keep interpretation out. “The email arrived at 2 am” is evidence. “The email arrived suspiciously late” is already an explanation wearing evidence’s clothes.

Include evidence that seems unimportant. Items that seem irrelevant under your leading explanation are often the ones that discriminate between hypotheses.

Step 2: Generate at least four explanations

Four is a deliberate number. Two produces a binary, and binaries invite you to argue for a side. Three tends to be two plus a token. Four forces you past the obvious.

Useful prompts for generating them:

• The benign explanation. What would this look like if nobody intended anything?
• The incompetence explanation. What if someone was careless, rushed, or misinformed?
• The different-goal explanation. What if someone acted deliberately but for a purpose other than the one you assumed?
• The measurement explanation. What if the thing you observed is an artefact of how you observed it?

Include your leading explanation as one of the four. It should compete, not preside.

Step 3: Score each piece of evidence against each explanation

For each item of evidence, ask a specific question: how consistent is this with each explanation? Mark it consistent, inconsistent, or neutral.

The critical move here is counterintuitive and worth stating directly. You are looking for evidence that is inconsistent with an explanation, not evidence that supports it. Support is cheap, because most evidence is consistent with most explanations. Inconsistency is what eliminates.

An explanation with no inconsistent evidence against it stays in play. An explanation with several clear inconsistencies is in trouble, regardless of how much supporting evidence it also has.

Step 4: Ask what would discriminate

The explanations still standing after step three are the ones your current evidence cannot separate. So the useful question is not “which do I believe?” It is “what observation would be consistent with one of these and inconsistent with the others?”

That question converts an argument into a research task, and it is the single most productive habit in this article.

Worked example

A community organization’s donation total drops sharply in one month.

Evidence: total donations down 40 percent versus the prior month; number of donors down 15 percent; average gift down 30 percent; the online donation page was redesigned on the 8th; a large annual donor gave in the prior month rather than this one; a competing local appeal ran in the same period; the email newsletter went out four days later than usual.

Explanations:

1. The redesigned donation page is losing people partway through.
2. The timing of one large annual gift accounts for most of the change.
3. Attention shifted to the competing appeal.
4. The late newsletter reduced the number of people prompted to give.

Scoring: the drop in average gift size is strongly consistent with explanation two and only weakly consistent with the others, since a broken checkout would be expected to reduce the number of completed donations more than their size. The 15 percent decline in donor count is consistent with explanations one, three, and four. The redesign date of the 8th is a discriminator: if explanation one is right, donations before the 8th should be roughly normal.

What would discriminate: the daily donation series split at the 8th, the completion rate on the donation page before and after, and the total excluding the single large annual donor. All three are obtainable in an afternoon.

Notice that no explanation was rejected on plausibility. Each was reduced or sustained by specific evidence, and the output is a short list of things to check.

Two failure modes

Generating alternatives you do not take seriously. If three of your four explanations are strawmen, the procedure produces the answer you started with while feeling rigorous. A useful test: could you argue each explanation convincingly for two minutes?

Treating the exercise as a verdict. Coming out of it with “explanation two is probably right, and the daily series would settle it” is the correct kind of conclusion. Coming out with certainty is a sign the scoring step was performed on preferences rather than evidence.

When to use it

Not for everything. It is worth the fifteen minutes when the question meets two conditions: the decision has real consequences, and your first explanation arrived quickly and comfortably.

That second condition is the important one. Explanations that arrive with a feeling of obviousness are precisely the ones that have not been compared with anything.

Summary

• Write the evidence down before writing any explanation.
• Generate at least four explanations, including a benign one and a measurement one.
• Score for inconsistency rather than support.
• Identify the observation that would discriminate between the survivors, and go get it.

No external references were required for this page. The procedure is a simplified adaptation of a long-established structured analytic technique, and the example is constructed for illustration.

Next step

Practice with the assumption audit worksheet

Topics covered: Critical ThinkingInformation Resilience