What "This Proves" Actually Proves
Your operations manager wants this paragraph in a report to the executive team:
"Our new 30-day onboarding checklist has clearly improved retention. Since we rolled it out in January, turnover in the first 90 days has dropped, and employees seem happier. Managers have told us the checklist is helpful. This proves the program is working and should be expanded to all four sites."
You know the following: first-90-day turnover at the pilot site fell from 22% in Q4 to 15% in Q2; the checklist launched in January at that one site; the same site raised its starting wage by $1.50 an hour in February; 9 of 14 managers answered an informal poll; the other three sites run different shift patterns.
1. Separate what the paragraph claims from what the evidence supports, then rewrite it in three or four defensible sentences.
Read the worked answer
Claim: the checklist caused a retention gain and should go to all four sites. What the evidence supports: turnover fell at one site during a period that also contained a wage increase.
Gaps, one at a time:
- 22% to 15% is a real, seven-point drop - but it compares Q4 to Q2, different seasons, with no earlier baseline to show whether 22% was normal.
- The $1.50 raise landed one month in. Two changes, one window: the data cannot separate them.
- "Employees seem happier" and "managers have told us" are impressions. 9 of 14 is a 64% response rate on an informal poll - usable as color, not proof.
- "This proves" is the giveaway. Correlational evidence supports "consistent with," never "proves."
Rewrite: "First-90-day turnover at the pilot site fell from 22% in Q4 to 15% in Q2, the first two quarters after the 30-day onboarding checklist launched in January. A $1.50-an-hour starting-wage increase began in February, so the drop cannot yet be attributed to the checklist alone. Manager feedback has been positive (9 of 14 respondents to an informal poll). We recommend piloting the checklist at a second site with no wage change to isolate the effect."
2. The operations manager says, "Hedging weakens it. Just say it worked." How do you answer?
Read the worked answer
"Hedged" and "accurate" look similar on the page but do different work. Hedging is softening a claim you could support - "results may vary somewhat." What the rewrite does is state the finding plainly, at full size (a seven-point drop), and then name the one limit on it.
Three reasons the limit belongs in the paragraph:
- Credibility is a single balance. If the executive team learns about the wage increase from someone else, they will also discount the seven-point number, and every number in the next report.
- Decision-makers act on effects they trust. An overstated claim gets discounted harder than an accurate one gets praised. The manager is optimizing the sentence; the report has to survive the follow-up question.
- The limit points to the next action. "Pilot at a second site with no wage change" is a stronger recommendation than "expand to all four sites now," because it tells the executives exactly what their approval buys.
The practical answer: if the operations manager wants a firmer sentence, the fix is more evidence, not a firmer adjective.
3. Suppose you can obtain exactly one additional piece of data before the report goes out. What should it be, and why that one rather than a manager survey or a longer write-up of the pilot?
Read the worked answer
The wage history at the other three sites. Specifically: did any of them raise starting pay between Q4 and Q2?
Why this one:
- It attacks the single strongest rival explanation. If the other sites held pay flat and their first-90-day turnover stayed near 22%, the checklist becomes the leading explanation at the pilot site. If their turnover fell too, the pilot result is probably labor-market driven.
- It costs nothing. HR already holds the numbers, and it takes an hour.
- It strengthens the recommendation without needing a second pilot, which buys the operations manager most of what they wanted.
Why not the alternatives:
- A manager survey measures opinion about the checklist, not whether behavior changed; it cannot separate the checklist from the raise.
- A longer write-up of the pilot adds detail about a site that is already described. Volume is not evidence.
Runner-up: four quarters of pre-rollout baseline at the pilot site, to show whether 22% was a spike or the norm.