ACADEMIC WRITING GUIDE
How to Write Dissertation Recommendations Based on Your Findings (With Worked Examples)
You have finished analysing your dissertation data and explained what the results mean. Now your supervisor asks: “What should happen because of these findings?” This is the purpose of the recommendations section. It turns carefully interpreted evidence into justified suggestions for future research, professional practice, or policy—without pretending that one study can prove more than it actually does.
Recommendations are often the weakest part of an otherwise solid dissertation. Students may write “universities should improve support services” or “more research should be conducted” without identifying what needs to change, who should act, or why the evidence supports that action. A useful recommendation is specific, proportionate to the evidence, and honest about uncertainty.
This guide explains how to move from findings to actionable recommendations, distinguish recommendations from implications and conclusions, and write recommendations for quantitative, qualitative, and mixed-methods dissertations. The complete examples are hypothetical teaching illustrations, not results from real studies. Adapt the process to your institution's required dissertation structure and the evidence you actually collected.
What are recommendations in a dissertation?
Dissertation recommendations are evidence-informed proposals about appropriate next steps following the study. Depending on your field and assignment, they may concern:
- Practice: actions that professionals, organisations, educators, clinicians or other practitioners could consider.
- Policy: changes that decision-makers could examine, consult on, pilot or evaluate.
- Future research: unanswered questions, populations, measures or designs worth investigating next.
A recommendation is not a new finding. It is a carefully reasoned response to something the study found—or could not establish. For example, finding that students report difficulty understanding written feedback can support a recommendation to pilot clearer feedback explanations. It does not automatically prove that a particular intervention will improve grades.
Many universities place recommendations near the end of the discussion or conclusion; others require a separate section or chapter. The University of Southampton notes that recommendations for further research may appear in the discussion or conclusion, depending on departmental conventions. University of Southampton dissertation guidance is useful background, but your own dissertation handbook takes precedence.
Recommendations, implications, conclusions and limitations: what is the difference?
These terms are related but perform different jobs.
| Element | Main question | Hypothetical example |
|---|---|---|
| Finding | What did the evidence show? | Interviewees reported difficulty interpreting brief feedback comments. |
| Interpretation | What might explain or contextualise it? | Students may lack the criteria or dialogue needed to act on those comments. |
| Implication | Why does it matter? | Providing feedback is not necessarily the same as making it usable. |
| Recommendation | What should someone consider doing? | Pilot short feedback-explanation sessions and evaluate whether students find later comments clearer. |
| Limitation | What is still uncertain? | Interviews with a selected group cannot show how common the problem is across all students. |
The recommendation needs the earlier steps, but it should not simply repeat them. Our guide to writing a research discussion chapter explains how to establish those interpretations; our research conclusion chapter guide explains how to bring the argument to a close. This article focuses on the further step of proposing defensible action.
A seven-step method for writing strong recommendations
1. Start with an actual finding, not a preferred solution
Create a brief inventory of your most relevant results, including findings that conflict with your expectations or show no clear relationship. Trace each candidate recommendation to a specific table, theme, estimate or documented observation. If you cannot identify the evidence, do not present the idea as a recommendation from this study.
Suppose a survey finds that students who report greater feedback clarity also report higher academic confidence. It is tempting to recommend a university-wide compulsory feedback programme. But a cross-sectional association does not show that clearer feedback caused confidence to rise. The supported starting point is narrower: the observed relationship suggests that feedback clarity deserves closer examination.
2. Decide what type of recommendation is appropriate
Not every dissertation needs a practice or policy recommendation. A basic-science, theoretical or exploratory study may legitimately end with recommendations for better measurement, testing competing explanations or examining other cases. Conversely, an applied project might identify a modest service improvement that warrants a pilot.
Ask who can reasonably use the evidence and whether the proposed action lies within their responsibility. This prevents vague statements such as “the government should solve this problem” when the study concerned a small, local dataset.
3. Match the strength of your wording to the evidence
Recommendation verbs matter. “Must implement” is stronger than “could trial,” “should consider,” or “would benefit from evaluation.” Use stronger claims only when the design, evidence base, and broader literature genuinely warrant them.
A randomised intervention study with reliable outcomes may support a firmer action than interviews about perceptions. Even a statistically significant result does not automatically establish practical benefit, value for money, feasibility or generalisability. Treat a non-significant result as uncertainty about the tested association or effect—not proof that an intervention has no effect.
4. Make the action specific enough to evaluate
A practical recommendation normally identifies an actor, a proposed action, the reason, and a way to assess it. It may also identify scope, timing or resource constraints where known.
Too vague: “Universities should enhance student support.”
Stronger: “The participating department could pilot a short, optional feedback-interpretation workshop for first-year students, then ask participants whether the workshop helped them identify concrete revisions and compare responses with existing support arrangements.”
The stronger version does not promise that the workshop improves grades. It defines a plausible next step and a modest evaluation plan.
5. Acknowledge feasibility, trade-offs and unintended effects
A recommendation that sounds beneficial may be expensive, burdensome, inequitable or impractical. Consider staff time, access, privacy, organisational constraints, and who might not benefit. If feasibility has not been studied, state that implementation would need consultation or piloting rather than treating feasibility as established.
For example, mandatory additional tutorials might disadvantage students who work or commute. Optional online and in-person alternatives may deserve consideration—but those alternatives also require testing.
6. Make future research recommendations answer a real uncertainty
“Future researchers should use a larger sample” is often too generic. Specify why a different sample, time frame, measurement or design matters.
For a cross-sectional survey, a longitudinal design could examine whether changes in one variable precede changes in another, although longitudinal data alone still may not prove causation. For interviews in a single department, a comparative study across different departments could test whether the accounts depend on context. Future research should address an identified question, not merely increase the number of participants.
The limitations of a study guide helps distinguish genuine evidential gaps from the boundaries you deliberately chose.
7. Write the recommendation as an evidence-to-action chain
A useful drafting pattern is:
Finding → interpretation → proposed action → intended purpose → uncertainty or evaluation.
For example:
“In this hypothetical interview study, several participants described uncertainty about how to apply comments on earlier assignments. This suggests that access to feedback alone may not ensure that students understand what to revise. The department could therefore pilot brief, voluntary sessions in which students practise interpreting anonymised feedback against assessment criteria. Student evaluations and subsequent assignment-review processes could help assess whether the approach is useful before broader adoption.”
Notice what the passage does not say: that the sessions are already proven effective, that all students need them, or that grades will improve. The recommendation is actionable precisely because it keeps the strength of the claim appropriate to the evidence.
Worked example 1: Quantitative research recommendations
Hypothetical study: A cross-sectional survey at one university examines the relationship between self-reported weekly independent study hours and examination performance among 180 consenting first-year students. The made-up data show a modest positive correlation, r = .24. This illustrative number is not a published result. No intervention was tested, and the survey did not establish whether more study hours preceded higher scores.
Weak recommendations
“Students should study longer to get better grades. Every university should require ten hours of independent study weekly.”
These claims assume causality, invent a universal threshold, and generalise beyond the sampled institution. The study measured an association; it did not test compulsory study schedules or identify an optimal number of hours.
Improved practice recommendation
“Academic advisers at the participating university could consider offering optional study-planning resources to first-year students, particularly those who report difficulty organising independent work. Because the observed relationship between reported study hours and examination scores was modest and observational, the resources should be framed as support rather than a proven route to higher marks. A small pilot could measure uptake, students’ perceived usefulness, workload and any appropriately collected learning outcomes before expansion.”
Improved future research recommendation
“A longitudinal study could record study habits repeatedly across a semester, including study quality, competing work commitments and prior academic preparation. Such evidence could clarify the timing and consistency of the association found in the present survey. Where ethical and feasible, a properly evaluated study-support intervention would provide stronger evidence about whether a particular form of support changes outcomes.”
Why these recommendations work
They preserve the original study's limits, avoid an unsupported ten-hour rule, identify a plausible university actor, and separate immediate support from the research needed to estimate effects. They also show how to propose useful action without claiming the evidence proved it effective.
Worked example 2: Qualitative research recommendations
Hypothetical study: Fourteen semi-structured interviews explore how first-year students at one institution understand written lecturer feedback. In this invented example, some interviewees describe actionable comments, while others report uncertainty about what brief remarks such as “analyse more critically” require. Participants differ in whether they feel comfortable asking for clarification.
Weak recommendations
“All lecturers must provide longer feedback because students do not understand their comments.”
Interview accounts cannot establish the prevalence of misunderstanding among all students; longer feedback is not necessarily clearer; and participants also reported positive experiences.
Improved practice recommendation
“The department could work with lecturers and students to develop a small bank of annotated feedback examples showing how common comments relate to assessment criteria and possible revisions. An optional trial across selected modules could gather feedback from students and staff on clarity, workload and accessibility. This would respond to the uncertainty described by some interviewees without assuming that one format fits all disciplines.”
Improved research recommendation
“Future qualitative work could compare feedback interpretation among commuting, part-time and residential students, while protecting confidentiality and avoiding assumptions that any group necessarily experiences greater difficulty. A complementary questionnaire may help estimate how common particular interpretation challenges are within a clearly defined population; it should not be treated as a substitute for the contextual accounts captured through interviews.”
Why these recommendations work
They take participants' differing accounts seriously and do not turn one qualitative theme into a population-wide statistic. The action is a limited co-designed trial, and the future research suggestion identifies exactly what additional evidence would contribute.
Worked example 3: Mixed-methods research recommendations
Hypothetical study: A sequential mixed-methods dissertation surveys students about satisfaction with digital feedback, then interviews a purposive subsample with relatively high and low satisfaction ratings. Invented survey results suggest varied satisfaction, while interviews suggest that access to a digital portal matters less to some students than knowing what to do with the comments. The two strands are integrated to explain patterns, not to claim cause and effect.
Integration before recommendation
A mixed-methods recommendation should not be written solely from the survey result or a memorable interview quotation. A simple integrated evidence table can help:
| Quantitative pattern (invented) | Qualitative explanation (invented) | Cautious implication |
|---|---|---|
| Satisfaction ratings differ between respondents | Some students describe clear, specific next steps; others describe generic comments | Satisfaction may depend partly on how actionable feedback feels, not just delivery technology |
| Portal access is widely reported | Some interviewees still struggle to decide what to revise | Access and usability should be evaluated separately |
| Responses vary across courses | Interview accounts describe different feedback practices | A universal solution may overlook local course needs |
These are invented teaching observations; they are not findings about a real university.
Improved institutional recommendation
“The university could pilot a feedback-usability review in a small range of modules, asking teaching teams and students to evaluate whether digital comments identify priorities and realistic next steps. The pilot should include accessibility and staff-workload considerations, as well as a mechanism for students to request clarification. The university could assess perceived clarity before deciding whether wider guidance is justified. Because this dissertation did not experimentally test a feedback intervention, improvement in attainment cannot be assumed.”
Improved future research recommendation
“A subsequent multi-site study could compare feedback practices and outcomes across different disciplines, while distinguishing satisfaction, comprehension, actual use of feedback and academic performance. If an intervention is developed, its evaluation should use appropriate comparison methods and pre-specified outcomes to investigate effectiveness rather than relying on satisfaction scores alone.”
Why this example works
The recommendation is linked to the integration of survey patterns and interview accounts. It distinguishes a feasible pilot from a proven solution and gives future researchers a specific design problem to investigate.
A recommendation matrix you can build from your own dissertation
Before drafting prose, make a table like this for your actual findings:
| Finding or uncertainty | What the evidence supports | Suitable recommendation | Important qualification |
|---|---|---|---|
| Interview participants struggle to interpret some comments | A documented experience among the interviewed group | Pilot clarified feedback examples | Do not assume all students share it |
| Study time correlates with scores | An association in an observed sample | Explore study-planning support and evaluate it | Do not infer causation |
| An intervention's effect estimate is imprecise | The result is uncertain at the observed precision | Conduct a properly powered replication or synthesis where warranted | Do not call an inconclusive result proof of no effect |
| A recruitment method omitted a relevant group | A coverage limitation | Include that group using appropriate access and safeguards in a later study | Feasibility and ethics still matter |
Only use rows that reflect your actual evidence. The table is a planning tool, not necessarily a required published dissertation table. If you need help showing numerical or thematic results accurately, see our guide to presenting research data in tables and figures.
Where should dissertation recommendations appear?
There is no universal chapter number or mandatory heading. A short recommendations subsection may follow the discussion or appear near the end of a conclusion chapter. Some professional doctorates, applied projects and institutional templates call for a separate “Recommendations” chapter or an implementation section.
Use the structure in your programme handbook. If recommendations appear inside the discussion, avoid repeating the same paragraph verbatim in the conclusion. If your conclusion summarises the main recommendations, make it brief and direct. For longer lists, group them by practice, policy, and future research, but include only categories your evidence warrants.
How many recommendations should you write?
There is no evidence-based universal number. Three well-supported, clearly differentiated recommendations can be stronger than ten vague suggestions. Prioritise by relevance to the research questions, evidential support, feasibility, and the stakes of implementation. If a recommendation concerns high-risk professional decisions, it may require far stronger evidence and specialist review than a low-risk exploratory pilot.
You can rank recommendations as immediate candidate for evaluation, requires consultation or feasibility work, or requires new research first. Do not assign certainty or urgency unsupported by your findings.
Common mistakes to avoid
Making recommendations unrelated to the research questions. A dissertation about student feedback should not suddenly propose broad national education reforms without supporting evidence.
Treating correlation as proof of causation. A relationship between variables does not establish that changing one will change the other.
Overgeneralising from a narrow sample. Specify the studied population and setting; make transfer or wider implementation conditional on appropriate evidence.
Confusing suggestions with demonstrated outcomes. “Pilot a workshop” is a proposal; “the workshop improved grades” would require actual study evidence.
Writing “more research is needed” without direction. Explain what should be examined, with whom, by what appropriate design, and what uncertainty it addresses.
Ignoring negative, mixed or inconclusive findings. Such findings may support a refined question or better measurement rather than a stronger intervention claim.
Inventing cost, feasibility or stakeholder approval. If these were not studied, acknowledge that implementation would require practical assessment.
Repeating the full discussion. Cite or briefly recall the decisive finding, then explain the next step and its boundaries.
Dissertation recommendations checklist
Before submitting, ask:
- Can I point to a real result, interpretation or explicit limitation behind each recommendation?
- Does each suggested action identify a realistic actor or a specific future research question?
- Is the language appropriately cautious for my design and the quality of evidence?
- Have I distinguished practice, policy and further-research proposals where relevant?
- Have I avoided inventing causal effects, cost savings, implementation success or population-wide conclusions?
- Would a reader understand what could be piloted, examined or evaluated next?
- Have I considered ethical, accessibility, financial or operational constraints where relevant?
- Does the section appear in the location required by my institution, without unnecessary repetition?
Frequently asked questions
Can I make recommendations if my results are not statistically significant?
Yes, but they should address what remains uncertain rather than claim that an effect definitely exists or does not exist. Depending on the design, useful recommendations may concern improved measurement, additional observations, evidence synthesis, or a different research question. Report the effect estimate and uncertainty where appropriate, not only the p-value.
Are recommendations compulsory in every dissertation?
Not always as a standalone section. Expectations vary across disciplines, degree levels and institutions. A theoretical dissertation may emphasise conceptual questions, whereas an applied study may require practice or policy proposals. Follow your handbook and supervisor's instructions.
Can recommendations introduce new sources?
Yes, selectively, when additional high-quality literature helps explain the feasibility or relevance of a proposed step. However, a late citation cannot rescue an unsupported claim about what your study found. Recommendations should remain grounded in your evidence, appropriately interpreted in the context of other research.
Should recommendations use “should,” “must,” or “could”?
Use wording proportional to the certainty, importance and practical context. “Could pilot” or “should consider evaluating” is often suitable for an exploratory student study. A firm “must” requires a compelling justification that usually goes beyond one limited dissertation.
What is the difference between recommendations for practice and future research?
Practice recommendations propose something that a stakeholder could try or assess in the real world. Future-research recommendations identify a question or methodological improvement needed to produce stronger or different evidence. A dissertation may need one, both, or neither in a particular form, depending on its purpose.
Final takeaway
Strong dissertation recommendations complete an evidence-to-action argument. They begin with what your study actually established, propose a proportionate next step, identify who might act or what future researchers should investigate, and remain clear about uncertainty. The goal is not to sound forceful. It is to make a responsible suggestion that readers can understand and, where appropriate, evaluate.
If your recommendations feel disconnected from the data, work backwards: Which exact finding justifies this proposal, and which part remains untested? That question often turns a generic final section into a valuable contribution.
If you would like academic support with organising your dissertation, checking whether recommendations follow logically from your own analysis, or improving clarity and structure, explore Acadelyra's academic-support services. Your research evidence, interpretations and academic responsibilities remain your own.
References and further reading
- University of Southampton Library — Results and Discussion: Writing the Dissertation. Explains the distinction between reporting results and interpreting their implications and notes that recommendations may sit in discussion or conclusion.
- University of Southern California Libraries — Organizing Your Social Sciences Research Paper: The Discussion. Explains interpretation, implications, limitations and future research recommendations.
Editorial integrity note: All study scenarios, counts, quotations, patterns and numerical results in this guide are hypothetical illustrations. They are not empirical findings, references to real participant data, or evidence that an intervention works.