Did errors fall after the change, or because of it?
Picture sixty warehouse staff moving from individual bonuses to team recognition. Picking errors fall from 4.8 to 3.1 percent, but overtime and the order mix change too. Correlation means two things moved together; causation means one produced the other. This trial shows only the first.
Report the movement with a time link, then hedge. To test cause, propose a reversible comparison: a change you can undo.
- “Errors fell after team recognition was introduced.”
- “The fall coincided with changes in overtime and order mix.”
- “The result is consistent with the new scheme, but the trial cannot establish causation.”
Giving an update before all the figures are in
Set the context, flag the risk, admit the limit and name the next step. Here, monthly check-ins replaced annual ratings in a pilot.
Listen: what does Maya admit, and what will she publish by Friday? (The pilot cannot establish causation; anonymized fairness ranges.)
Maya Where we are at the moment is a three-month sample of forty specialists.
Owen And I think you should be aware that twelve employees reported goal changes without written confirmation.
Maya That’s a good point. I have to admit that the pilot cannot establish causation.
Owen So, the main thing to remember is that a six-month extension will add a shared goal log.
Maya I’ll verify manager time and publish anonymized fairness ranges by Friday.
Phrases for limits and missing figures
These frames buy time without inventing data.
- Context: “What I can tell you is that documented feedback rose from 46 to 88 percent.”
- Limit: “I’d like to be able to give you the final cost, but unfortunately I can’t until Friday.”
- Approximate: “I can’t remember exactly, but off the top of my head it was about 35 minutes per employee.”
- Alternative: “I don’t have the exact figures, but what if I send them to you by Friday?”
Questions that test the evidence
Indirect questions keep statement word order: “Could you explain why the cost increased?”, not “why did the cost increase”. An opposite-polarity tag with falling intonation checks what you believe; with rising intonation it asks a real question. A positive tag on a positive statement reacts to news, often with surprise or doubt.
- Checking: “The trial covered sixty staff, didn’t it?”
- Reacting: “So you support the extension, do you?”
- Challenging: “And the lower error rate proves the scheme worked, does it?”
Refusing confidential questions without stonewalling
Some fair questions cannot be answered in full, such as which manager scored lowest or what one person earns. Refuse, give the reason, then offer something safe, such as anonymized ranges by shift.
- Formal: “I’m afraid I can’t disclose that information. It would identify individual managers.”
- No simple answer: “Did recognition cause the fall? Well, it’s not that straightforward.”
- Not your decision: “Will the trial be extended? It’s not for me to say.”
The board record in five moves
Write 170–200 words a director could approve without a meeting.
- Evidence: “Picking errors fell from 4.8 to 3.1 percent after team recognition replaced individual bonuses.”
- Boundary: “Overtime and the order mix also changed, so the trial cannot show that recognition caused the fall.”
- Safeguard and owner: “Individual pay data will stay confidential; HR will publish anonymized ranges by shift.”
- Decision gate: “We will extend the trial for eight weeks if errors stay below 3.5 percent with overtime unchanged, and stop it if they exceed 4.5 percent for two consecutive weeks.”
- Response route: “Please confirm the extension by Thursday at 17:00.”
Common mistakes
Avoid these:
- “The new scheme reduced errors.” Write: “Errors fell after the new scheme was introduced.”
- “Could you tell me why did the cost increase?” Use “why the cost increased”.
- “The trial covered sixty staff, isn’t it?” The tag follows the verb: “didn’t it?”
- “It’s nothing to do with me.” Too blunt; say “It’s not for me to say.”
- “Which employees …?” asks for names; for numbers, ask “How many …?”
A 15-minute daily practice routine
Use a real result from your work. Weekly, argue aloud: is recognition more motivating than pay?
- 3 minutes: describe it with “after”, “coincided with” and “cannot establish”.
- 4 minutes: give a 90-second update that admits one limit.
- 3 minutes: refuse three confidential questions, offering something safe.
- 5 minutes: draft the five-move record; check every causal verb.
What comes next
Next: a greenhouse pilot proposal with unverified savings, then an approval record limited to what the evidence covers.

