A useful first answer
Use finding, evidence, uncertainty, and decision
Your listener should be able to repeat four points after you finish: what changed, what evidence supports that statement, what limits the conclusion, and what decision the result can inform. Put the finding in the first sentence. Use one comparison or quantity that matters, then translate the main limitation into its practical effect. Finish with a recommendation that stays inside the evidence.
Give the finding
Name the direction and the group or period. Avoid starting with the query, chart, or method.
Customers who saw the reminder completed more renewals during the test.
Show the evidence
Choose the comparison that supports the finding and define the population or time window.
The comparison covers completed renewals among eligible customers during four test weeks.
Translate the uncertainty
Say why the result could change and which conclusion is unsafe. Do not hide behind the single word significant.
Weekend traffic was limited, so we cannot assume the same effect on weekends.
Connect it to a decision
Recommend a next step that matches the strength and scope of the evidence.
Keep the weekday test running and collect a larger weekend sample before expanding it.
Worked conversation one
Explain what the result supports and what it does not
A retention experiment update+
A product lead wants to know whether a reminder should be shown to every customer.
The reminder group completed more renewals than the comparison group during the four-week test. The difference was concentrated among customers renewing for a second year.
Does that mean we should release it to everyone?
The result supports a wider test for customers renewing for a second year. It does not answer the question for customers renewing for the first time because that subgroup was small and behaved differently across weeks.
What would reduce that uncertainty?
We can extend the test until the first-renewal group is large enough for a stable comparison, while checking whether the weekly pattern continues.
Let’s expand only for customers renewing for a second year and keep the first-renewal test open.
The analyst gives the useful result first, then narrows the recommendation when the stakeholder asks for a larger decision.
Worked conversation two
Respond when a stakeholder wants a stronger conclusion
A regional performance question+
A director sees a lower support time in one region and asks whether the new process caused it.
The region using the new process is faster. Can we say the process caused the improvement?
We can say the lower time occurred in the same period. We cannot isolate the process as the cause from this comparison alone. That region also had fewer complex cases.
So is the result useful or not?
It is useful as a signal. It tells us where to investigate. The next step is to compare similar case types and check whether the gap remains after accounting for case mix.
What should I say in today’s meeting?
Say that the region reported lower handling time, but the cause is not established. We are testing whether the difference remains for comparable cases.
The analyst does not call the result useless and does not convert association into causation. The answer offers a decision-relevant signal and a concrete way to learn more.
The data proves the new process works.
The result is consistent with an improvement, but this comparison does not isolate the process as the cause.
The sample is too small, so we know nothing.
The small subgroup makes the estimate unstable, so use it as an early signal rather than a rollout decision.
Adjust the level of detail
Use the same logic for three audiences
- For an executive: give the finding, decision implication, and largest uncertainty in three sentences; offer the method only if asked.
- For an operational team: add the affected group, time window, and the condition they should monitor.
- For another analyst: name the definition, comparison, assumptions, and checks before discussing the recommendation.
- Choose a fictional finding and write four headings: finding, evidence, uncertainty, decision.
- Give a 45-second update for a non-technical listener without describing every chart.
- Answer the challenge “Can we roll this out everywhere?” with a bounded recommendation.
- Repeat the update in 25 seconds while keeping the most important limitation.
Choose how to continue
Use independent voice practice for repetition, or choose teacher guidance for a broader learning plan.
Practise the result aloud
Practise spoken English with Flalingo: AI Speaking Coach and get feedback after you speak. Build the confidence to deliver a short explanation and respond to a follow-up question.
Explore the AI speaking coachDevelop professional explanations with a teacher
For planned live English lessons and teacher support across broader workplace communication, review the Flalingo method.
Explore the Flalingo methodSources
- Communicating quality, uncertainty and change
UK Government Analysis Function · Reviewed 10 Sept 2026
Used for transparent explanation of strengths, limitations, uncertainty, and supported conclusions. Its formal audience is producers of official statistics; this page adapts only the general communication principle.
- The Aqua Book: guidance on producing quality analysis for government
HM Treasury / UK Government Analysis Function · Reviewed 10 Sept 2026
Used for the principle that uncertainty is inherent, its reasons should be described, and a best estimate alone may be insufficient. It does not validate any fictional result in this lesson.