In short
A good research report states what you asked, what you did, what happened and what it means, in that order, with numbers and limits. Overclaiming usually happens in small steps: a group becomes "consumers", a difference becomes "better", and a correlation becomes a cause. This lesson shows how to prevent those slips. It'll also protect you, because someone may later use your sentences as a claim.
What you will learn
- Structure a report so it can be checked
- Report results with numbers and limits
- Rewrite overclaims into accurate statements
Before this lesson: Recording and Sharing Data
A report structure
| Section | Answers |
|---|---|
| Abstract | What did you ask, do and find? What's the main limit? |
| Introduction | Why this question? What did you find already published? |
| Method | Materials, samples, participants, design, measures, analysis plan, ethics |
| Results | What happened, in numbers, in the planned order, then exploratory |
| Discussion | What does it mean? What else could explain it? What are the limits? |
| Conclusion | One or two sentences your evidence supports |
| Appendix | Data dictionary, deviations, consent text, code or formulas |
Results describe, and discussion interprets. Keep them apart.
Overclaims and rewrites
| Overclaim | What's wrong | Accurate version |
|---|---|---|
| Consumers prefer A | Generalizes from a small group | 11 of 16 adults recruited from our workplace chose A |
| A lasts twice as long | Hides medium and measure | Median last-detected time on strips was 8 hours for A and 4 for B, in six assessors |
| Proves that Hedione adds freshness | One test proves little | Freshness ratings rose by about 1 point on our 0 to 10 scale; we haven't repeated it |
| No difference between old and new | Confuses null with equivalence | 4 of 12 correct; this doesn't show a difference, and the test had low power for moderate differences |
| Safe, as nobody reacted | Wrong kind of evidence | Strips only; no safety conclusion can be drawn |
| Natural materials improve wellbeing | Not tested; a possibly medicinal claim | Delete, unless a suitable study exists |
Numbers to report
- The count. Say how many people, and how many judgments.
- The group. Who were they, how were they recruited?
- Size. The size of the difference matters more than whether it passed a threshold. Give the raw counts, or means with a measure of spread.
- Uncertainty. Use a confidence interval or an exact p-value, where you can. Don't write "p < 0.05" when you know the value.
- Everything planned. Report all your planned measures, not only the ones that look interesting.
Discussing alternatives
For any finding, list what else might explain it. Could order, batch, the room, or something about your group be responsible? Could the assessors have guessed? Could the musk have outlasted the citrus you meant to measure? Say what you did to rule these out, and what you didn't do. If an obvious rival explanation remains, say that clearly, as it's what an informed reader would be thinking.
Words to watch
| Avoid | Use |
|---|---|
| proves, confirms | suggests, was consistent with |
| causes | was associated with, unless your design supports it |
| significantly (with no test) | a stated number |
| people, consumers | the specific group tested |
| always, never | in this test |
After writing
Ask three people to read it: someone who knows the method, a perfumer and someone new to all this. Ask each one what they think you showed. If they say more than you found, that's an overclaim in your writing, even if each sentence is accurate. Check that a marketing team couldn't easily lift a sentence that overstates it. If your work is for commercial use, have a suitable person review any claim based on it, as the claims guide advises.
Practical exercise: Rewrite and report
- Rewrite these claims, inventing plausible numbers and labeling them as invented: (a) This perfume lasts all day. (b) Our testers loved it. (c) Clinically tested. (d) Scientifically proven to relax you. (e) The new formula is identical. (f) Nobody can tell the difference. (g) Natural means gentler. (h) Women prefer version A. (i) Better than the leading brand. (j) Scientists agree.
- Mark which ones you would simply delete, and why.
- Using your dry-run data, marked as test data, write a one-page results section with counts, uncertainty and a list of planned measures.
- Have someone read it, and tell you what they think it showed.
You have done this well if no sentence reaches further than your data, and you were willing to delete at least some of these claims rather than rescue them.
Common mistakes
- Reporting only what's interesting. Report the plan.
- Mixing results with discussion. Keep them separate.
- Hiding the group's description. It limits every claim.
Knowledge check
Written and reviewed by the RUŌOD Lab team. This is general education, not legal, regulatory, medical or safety advice. Example dilutions and exercises are for learning, are not validated products, and have not been safety-assessed. Completing a lesson confers no qualification. How we check these lessons.