Did you know that the last sentence of a school report can sound like a discovery when it is actually advice? Students collect responses about a library’s opening hours and then write, “The library must open earlier.” The survey might offer a good reason to consider a change, but the change itself is not a measured result.
A research finding describes what the evidence showed within the investigation’s defined scope. A recommendation proposes what someone should do in response, often bringing in priorities, costs, feasibility and uncertainty that the investigation did not measure directly. Strong reports show the bridge between them rather than presenting a preferred action as a proven fact.
Quick distinction: what happened, what it means, what to do
- Finding: “Among our 80 respondents, 48 selected the morning option.” This reports a recorded result.
- Interpretation: “Morning was the most frequently selected option within this group.” This explains the observed pattern.
- Recommendation: “Consider a trial morning session to test actual attendance.” This proposes an action.
- Limitation: “These 80 voluntary respondents cannot automatically represent every student.” This limits the claim.
- Decision: the organisation weighs evidence, staffing, safety, cost and other needs before adopting or rejecting the recommendation.
Why the distinction changes the quality of writing
Research answers a question about evidence; recommendations address a question about action. The two often belong together, but the second does not logically follow without further judgement. A class might find that some students prefer longer library hours, while the library still needs to check supervision, availability of space and actual demand before adjusting a schedule.
Separating these roles also protects the student from overstating a result. “Sixty percent of all students want an earlier session” is a different claim from “Sixty percent of our survey respondents selected an earlier session.” The first requires evidence about the whole population. The second can be checked against the actual sample.
A complete worked example with the numbers shown
Consider an invented school library study. A voluntary form is completed by 80 students, who each select exactly one option about an additional study session: morning, afternoon or no preference. The recorded answers are: 48 morning, 20 afternoon and 12 no preference. These numbers are illustrative teaching data, not observations about a real eduKate learner or school.
Check the counts before choosing words
Total responses = 48 + 20 + 12 = 80. Morning share = 48 ÷ 80 × 100% = 60%. Afternoon share = 20 ÷ 80 × 100% = 25%. No preference share = 12 ÷ 80 × 100% = 15%. Verification: 60% + 25% + 15% = 100%. If the figures do not total 100% in a single-choice survey, investigate before writing the narrative; rounding alone might sometimes explain a small mismatch, but it does not here.
Write the finding without adding imaginary certainty
A suitable finding is: “In this voluntary group of 80 respondents, morning was the most common preference, selected by 48 students (60%).” The word respondents is deliberate. We have not shown how many non-respondents wanted a different option, whether respondents had equal opportunity to reply, or whether attendance would match the stated preference.
Explain the interpretation and limits
One cautious interpretation is: “The responses suggest that a morning session may be worth testing.” That is narrower than saying, “The school needs a morning session.” This study measures a stated preference among self-selected respondents. It does not establish staffing costs, timetable feasibility, actual attendance or the preferences of all students.
Make the recommendation testable, not grand
A proportionate recommendation is: “Subject to staffing, access and school approval, pilot one morning study session for four weeks and compare attendance with available places.” The four-week pilot is a proposed action, not a scientific necessity. It can be evaluated, adjusted or rejected by the responsible school staff. An unjustified recommendation would be “Replace every afternoon session with morning sessions immediately.” The questionnaire did not test that policy.
The bridge from result to action
A helpful report uses four linked sentences. First, give the question: “Which additional session time do our respondents prefer?” Second, give the finding with a denominator. Third, identify a limitation: “Participation was voluntary, so views of non-respondents remain unknown.” Fourth, state a conditional recommendation and what evidence would improve it: “Trial a morning slot if staffing permits, then track attendance.” Each sentence performs a distinct job.
Notice that the recommended trial can fail without invalidating the survey result. Forty-eight respondents might genuinely have preferred mornings at the time of the survey, while few later attend because of transport, family routines or other commitments. The finding is about responses; the pilot is about behaviour. Distinguishing the two prevents the report from becoming self-contradictory.
Diagnostic: spot the sentence doing too many jobs
Give a pupil the statement, “Our survey proves that extending library hours will improve exam scores.” Ask for a three-colour audit. What was actually measured? Which mechanism was proposed? Which outcome was never tested? Unless exam-score changes were validly studied under an appropriate design, the sentence leaps from preferences to academic outcomes. Even if students say they would use the library more, that is not a guaranteed score improvement.
A better version might say: “Some respondents expressed interest in longer hours. The report did not measure study outcomes; a supervised trial could examine actual use.” This keeps the useful observation, removes the unsupported causal claim and proposes a next investigation. The goal is not to make every sentence timid; it is to match its strength to the evidence.
Read research reports more intelligently
When students use an article or a report, have them label each important statement as observed result, author interpretation, recommendation, or background information. A report’s ‘Conclusions’ section may contain more than one of those; a heading is not a guarantee of evidential status. Read the methods and results to find exactly what the source measured, then compare that with the final recommendation.
Watch for modal language. “May help” is not equivalent to “will help”; “could consider” is not “must implement”; “the sample showed” is not “the population is”. Do not blindly swap one modal verb for another. Quote and paraphrase the author’s degree of certainty faithfully, and add your own evaluation separately.
A guided-to-independent writing sequence
Level 1: copy the data accurately
Give the learner a small results table and ask for one factual sentence containing a count, a base population and the time or place. For the library example: 48 of 80 voluntary responses favoured morning. Checking arithmetic and the denominator is the first learning goal.
Level 2: add a bounded explanation
Next ask what the data suggest and what they cannot establish. The learner should be able to say both without making up additional survey participants. A sentence such as “This may indicate interest among respondents, but it does not measure attendance” correctly separates preference from future behaviour.
Level 3: propose one feasible decision test
Now design a reversible, low-risk next step. Define who would approve it, what would be measured and what would count as useful feedback. A trial might measure attendance per available seat, but the school—not the student author—would determine its practicality and permission.
Level 4: transfer to a different topic
Use a science investigation, a classroom reading survey or a project cost comparison. The student must identify a claim made by the evidence, a proposed explanation and a suggested action without being given the labels. This stage is more revealing than reproducing the library sentences from memory.
Independent practice and hand-checked answers
- Case A: In an invented study, 18 of 30 volunteers chose option A. Write the finding as a count and percentage, without assuming all students agree.
- Case B: A report says “the measured water temperature rose by 3 °C during the trial” and “repeat the experiment with a lid.” Which is the finding and which the recommendation?
- Case C: In a fictional school project, a morning room has 24 seats and 30 students say they might attend. Does that by itself justify opening a second room?
- Case D: An article’s authors write “future work could use a larger sample.” Is that a measured finding about the study’s population?
Answers: A: 18 ÷ 30 × 100% = 60%; “18 of 30 volunteers (60%) selected A.” B: The temperature change is the observed result, assuming the measurement is valid; repeating with a lid is proposed work. C: No—intended attendance, available capacity, staff approval and other constraints must be checked; the responses may motivate investigation. D: No; the statement is a methodological recommendation.
Common mistakes and why they happen
- Changing “our respondents” into “all students” without representative sampling evidence.
- Writing a recommendation as if the project had directly measured its effectiveness.
- Claiming that a correlation or preference demonstrates a causal improvement.
- Ignoring inconvenient results that do not support the preferred recommendation.
- Presenting an interpretation as a verbatim quotation from the original source.
- Adding the recommendation in the middle of the results table, making evidence and judgement hard to distinguish.
- Treating a recommendation as automatically wrong when it is tentative; careful recommendations are a legitimate part of a useful report.
What parents, teachers and students should do next
Parents can ask, “Which sentence came from the data, and which sentence tells someone to act?” Teachers can return a marked-up report with two comments rather than one overall grade: one for fidelity to evidence, another for whether the recommendation follows reasonably from stated assumptions. Students can keep a two-column draft: “Evidence I have” and “Additional decision information I still need.” This makes revision constructive rather than punitive.
To check progress, collect a fresh paragraph before teaching, then a different paragraph after guided practice, and finally an unfamiliar paragraph a few days later. Score whether the learner can state the sample, a justified finding, a stated limitation and a conditional next step. Do not equate success on the identical rehearsed paragraph with independent mastery. A practical aim is to make every claim inspectable.
Frequently asked questions
Can a school report have a recommendation without a separate experiment?
Yes, if it identifies the proposal as a recommendation, explains the supporting evidence and acknowledges untested assumptions. It should not claim the proposal has already been proven effective.
Is every finding a fact about everyone?
No. A finding can describe a small sample, a single trial or a particular time. It becomes misleading when its scope is expanded beyond the design.
Can recommendations include values?
Yes. Decisions often balance fairness, cost, accessibility, safety or convenience. The writer should make those considerations explicit rather than pretending the data selected a value judgement automatically.
How do I check a sentence copied from a source?
Return to its original context and method. Ask whether it appears as a result, interpretation or recommendation and preserve that distinction when paraphrasing.
Related reading
For precise source support, use How to Check Whether a Citation Supports the Exact Sentence You Wrote. For sample limits, read How to Report a Volunteer Class Survey Without Claiming It Represents Everyone. Continue at the eduKate Education Hub, or explore our diagnosis-first approach through the Secondary 1 Mathematics Clementi reference.
Education is not simply the ability to produce a tidy recommendation. It is the habit of knowing what we found, what we think it means, what we still do not know—and what action is reasonable next.