How to Write a Decision Sentence

What Is the Problem?

A sentence-less metric does damage in three costumes.

The number is watched and triggers nothing

The dashboard is full, everyone is “on top of it” — yet nobody can say at what value of the falling rate anyone acts. Thresholdless watching is checking the thermometer without knowing when to close the window: data present, decision absent. The 61-percent waste band in McKinsey’s decision research accumulates exactly in this gap — the number attends the meeting; the decision does not.

The decision is made, then the number is fetched

The inverse costume: first the decision, then a supporting number found for it. Here the number is not agent but witness — a biased witness at that, since it was located by search. Writing the sentence in advance promotes the number from witness to referee: the threshold was set before the decision and bends to nobody’s mood of the day.

The same number speaks differently on different days

A metric without a written threshold is interpreted by the reader’s state of mind: in a good week it’s “fluctuation”, in a bad one “alarm”. The sentence closes this arbitrariness — the number is no longer interpreted; it is read. The cheapest form of institutional memory is a written threshold.

Why Does It Happen?

The unwritten sentence has three roots.

Work starts from the metric, not the decision

Dashboards get built on “what can we measure”; the right question is “which decisions do we make”. A list drawn from the measurable swells and never connects; a list drawn from decisions stays short and works in every row. The first step of sentence-writing is therefore not choosing metrics but counting decisions.

A threshold feels like a promise

Writing a threshold is saying “when this happens, we will do that” — and a promise invites accountability. Thresholdlessness is less neglect than comfort: nobody is ever wrong, because nobody predicted anything. Sentence culture replaces the fear of being wrong with the habit of adjustment: a missed threshold is not shameful; it is updated.

Ownership floats

“Let’s keep an eye on it” has no subject; a number everyone watches, nobody carries. The name field in the decision sentence gives the metric a hand — and resolves, before anyone sits down, the costliest item of decision waste: the ambiguity of who decides.

How Is It Done?

Drafting is three steps; ten minutes per decision.

Step 1: walk backwards from the decision

First list the recurring decisions: price adjustment, stock order, campaign stop, content repair, budget shift. Beside each, one question: which number would signal this decision at the right moment? A metric enters the list only as a decision’s answer — the sentence always points from decision to number, never the reverse.

Step 2: fill the five fields in order

Metric: single and precise — which source, which cross-section. Threshold: two-banded where possible — an alarm threshold and an opportunity threshold; a one-way threshold wastes the good-news signal. Act: the name of the task a breach triggers — not “we’ll look into it” but “a shopfront round opens”. Owner: one name; never a committee. Clock: reading frequency and, where needed, the decision’s deadline. The first threshold is set roughly — last period’s average plus its natural swing band is a fine start; precision comes from living with it.

Step 3: test it, hang it where it works

The sentence is tested against history: applied to the last six months, how often would it have fired? A sentence that never fires is loose, one firing weekly is tight — the band gets adjusted. Then the sentence is written not onto the dashboard but onto the reading table’s page: decision sentences work not where they are displayed but at the moment they are read.

How Long, at What Cost?

The sentence economy is deliberately small.

First round: half a day, ten to fifteen sentences

Counting the recurring decisions and drafting the first sentences is one half-day table. The target is core, not coverage: the ten to fifteen decisions made most often at the highest stakes. The rest join in turn once the arrangement settles — whoever tries to write all of them on day one operates none of them.

Maintenance: one adjustment per period

Thresholds drift with seasons, work and scale; bands that missed get updated at the period reading. Adjustment is not embarrassment but the arrangement’s breathing — an unchanged threshold is either perfectly calibrated or never read, and the second is far more likely.

The yield: shorter meetings, faster decisions

A sentenced metric arrives at the meeting ready: number, threshold and task on one line; the debate drops from “what shall we do” to “is the calibration right”. The courage to cap decision time is born here too — the ready sentence is fast, accurate deciding’s infrastructure.

The Common Mistake

The sentence gets written; it breaks in three places.

Confusing threshold with target

A threshold is the line where action starts; a target is where you hope to arrive. “Our target is 5 percent” triggers nothing; “under 3 percent for two weeks opens a round” does. Targets go on the motivation board, thresholds on the decision page — load both onto one line and neither works.

Sentence inflation

Once the arrangement is loved, the urge arrives to write a sentence for every metric; a fifty-sentence page is no better than a sentence-less dashboard. The limit is reading capacity: as many sentences as the weekly table can genuinely read — the rest belongs to the retirement round.

Leaving the fired trigger unrecorded

The threshold broke, the task was done — but what was done and what came of it went unwritten. An unrecorded trigger severs the sentence’s learning loop: which thresholds are well placed can only be read from trigger-and-outcome records. One line suffices: date, trigger, act, result.

Frequently Asked Questions