Clarity test

What it is

The clarity test asks whether a statement is defined well enough that a person with perfect information could settle it. Applied to a business decision, it forces four things into one sentence: who chooses, what they do, by when, and which observable result would count as success. A question that passes can be answered and later scored. A topic that fails can absorb research for months without ever being closed.

Where it comes from

Ronald A. Howard, a Stanford professor of engineering-economic systems, named decision analysis as a discipline in 1964 and set out the clarity test in "Decision Analysis: Practice and Promise", published in Management Science in June 1988. His device is the clairvoyant, whom he described as a person who knew the future and had access to all future newspapers and readings of physical devices. In his later teaching, the clairvoyant knows every event and measurable quantity, past, present and future, and holds no judgment at all. An element of a decision passes the clarity test when the clairvoyant could state its value, or say whether it happened, without asking a single clarifying question. Howard applied the test to every part of a decision model: the alternatives, the uncertainties, the outcomes and the measures of value. He pressed the same discipline further in "Speaking of Decisions: Precise Decision Language" (2004), which argues for terms exact enough to be checked. Running the test on the framing sentence itself, as this playbook does, is practitioner custom rather than a claim in the 1988 paper.

What it corrects

The failure begins with a question that sounds important and names no choice. "Should we enter Europe?" invites a competent team to do competent work: market sizing, competitor scans, a pricing study, a legal review. Because nobody has said what will be decided, no piece of evidence can be shown to be irrelevant, so the work has no stopping rule and the meeting ends with a summary instead of a commitment. Ordinary care makes this worse rather than better. The more thorough the analysis, the more solid the topic feels, and the later anyone notices that the decision was never stated. Two people can also agree on the answer to a vague question while meaning different actions, and nobody can grade the call afterwards, so the team learns nothing from having made it.

How it works

1. Write the choice as one sentence: which actor takes which action, by which date.
2. Name the boundary. Say what this decision does not cover and what stays fixed while it is taken.
3. Define success as a number someone could read off a named report on a named date.
4. Hand the sentence to the clairvoyant. Could a person with every fact and no judgment say whether that result occurred? Repair whichever words they could not settle.
5. Check that at least two genuinely different actions still fit the sentence. If only one fits, you have written a plan and not a decision.

Worked example

"Are hedge funds worth their fees?" is an argument that can run forever. On 19 December 2007 Warren Buffett turned it into a proposition registered at Long Bets: the S&P 500 would outperform a portfolio of funds of hedge funds over ten years, with performance measured net of fees, costs and expenses. Every term settles. The index is named. The counterparty, Protégé Partners, picked five funds-of-funds holding interests in more than 200 hedge funds. The window ran from 1 January 2008 to 31 December 2017. The fee treatment, which is where such arguments usually escape, sat in the sentence itself.

The scorecard in Berkshire Hathaway's 2017 shareholder letter reports the index fund up 125.8% over the decade, an average of 8.5% a year, against cumulative gains of 21.7%, 42.3%, 87.7%, 2.8% and 27.0% for the five funds-of-funds, one of which was liquidated in 2017. Girls Inc. of Omaha received $2,222,279. What the clarity test bought was not the win but the settlement: because the measure, the period and the fee basis were fixed in advance, the result could not be reinterpreted by either side.

In a Business Case Weekly case

The first fork in the prediction markets case sits in Iowa City in 1988, the year three Iowa economists launched the Iowa Political Stock Market, whose 192 traders called George H. W. Bush's vote share to within 0.2 percentage points. It asks whether you would put $10,000 behind prediction markets as a business. "Prediction markets as a business" is a topic. The clarity test converts it into a decision you could be held to: which entity you fund, under which regulator, which licensing or enforcement event you would need to observe and by when, and what return over what horizon would justify the money. The case is a fair place to practise the move, because an exchange's own contracts have to pass the same test, with a settlement source and a deadline, before anyone can trade them.

In your answer

• "The decision is whether [actor] will [action] by [date]; it excludes …"
• "Success means [number] measured on [date] in [named source]."
• "A person with perfect knowledge could settle this by checking …"
• "Two genuinely different options fit this question: … and …"

Common misuse

The convincing counterfeit is precision aimed at the wrong quantity. "Will we reach 40,000 downloads by March?" would satisfy any clairvoyant and still tells you nothing about whether the business works. The one-line test: name what you would do differently under each answer. If both answers lead to the same action, the sentence is exact and idle, and the framing work has not been done.

References

• Ronald A. Howard, "Decision Analysis: Practice and Promise", Management Science 34(6), 679–695, 1988. The paper that sets out the clarity test, and short enough to finish in an evening.
• "Clarity test", Wikipedia. A one-page summary with Howard's definition of the clairvoyant and its sources.
• Warren Buffett, "2017 shareholder letter", Berkshire Hathaway, pages 11-12. The full terms and the final scorecard of the ten-year bet.
• Ronald A. Howard, "Speaking of Decisions: Precise Decision Language", Decision Analysis 1(2), 71–78, 2004. Citation record and abstract; the full text is behind the publisher's paywall.