Drawing Conclusions From Your Results
Four rules for the conclusions of a research report — limited to what the results support, equal emphasis on positive and negative findings, and based on fact and logic rather than supposition. Three of the four turn on terms the source never defines.
The four rules
Read together, all four are versions of one instruction: do not claim more than you established. That is the entire discipline of a conclusions section, and it is harder than it sounds because by the time you write it you have lived with the material long enough to believe things the data never showed.
Rule 1 — limit conclusions to what the results support
The most-broken rule in research writing, and the breach is rarely deliberate. It happens through accumulation: a hedged reading in the discussion becomes a firmer statement in the conclusion, and a firmer statement in the conclusion becomes a flat claim in the abstract. Nobody decided to overclaim. Each step was a small tightening.
Rule 2 — equal emphasis on positive and negative findings
This is the rule that most improves a report and is most often quietly ignored. A study that reports only what it found reads as though the world cooperated. A study that reports what it looked for and did not find reads as though someone was actually testing something.
What gets quietly dropped, and why it matters
The relationship that was not there
You expected experience to predict something. It did not. That is a finding about the world, and omitting it means the next person runs your study again.
The theme that did not emerge
Your literature review said this would dominate the interviews. Nobody raised it. Reporting the absence is reporting evidence.
The case that contradicted the pattern
Five interviews agreed and one did not. The sixth is usually where the interesting question is, and it is the first thing to disappear under a tidy theme.
Rule 3 — secondary and post hoc analyses
The reason behind it, which the source does not give, is that an analysis chosen after you have seen the data is not testing anything: the data suggested the analysis, so the analysis cannot independently confirm what the data suggested. Presenting such a result as explanatory rather than as evidence is the honest treatment. It is a hypothesis for someone else to test, not a conclusion you established.
Rule 4 — fact and logic, not supposition
The fourth rule is the one that sounds like boilerplate and is not. It is worth separating into its two halves, because they fail differently.
Failures of fact
- A conclusion stating something your data did not measure
- A figure that drifted between chapters
- A claim attributed to a source that does not contain it
- Caught by the traceability pass
Failures of logic
- A correlation reported as a cause
- A conclusion true of your sample stated of the population
- An alternative explanation never considered
- Not caught by the traceability pass — every sentence traces, and the inference is still wrong
The step the material never explains
A closing sequence
Writing the conclusions
List what you actually established
One line per finding, with the result that supports it beside it. This is the traceability pass and it takes half an hour.
Add what you looked for and did not find
Given the same weight. If it changes what someone would do, it is a conclusion.
Move anything post hoc out
Into the discussion, framed as explanatory and requiring separate testing. It does not belong here and it must not reach the abstract.
Test each remaining sentence for logic
Association or mechanism? Sample or population? Is there an alternative explanation you have not addressed?
Bound each claim explicitly
Say what it applies to. A bounded conclusion is stronger than an unbounded one because it can survive being checked.
Reconcile against the abstract last
The abstract is written early, revised least and read most.
What to carry forward
- All four rules are one instruction: do not claim more than you established. The breach happens by accumulation, not by decision.
- Trace every conclusion to a specific result. Anything untraceable is an inference or an overclaim.
- Report what you looked for and did not find, with equal weight. It is the single change that most improves a research report.
- An analysis chosen after seeing the data cannot confirm what the data suggested. Present it as explanatory, in the discussion, and keep it out of the abstract.
- The material names Recommendations as a closing step and never says what one is. That boundary — conclusion states what is, recommendation states what to do — is yours to hold.
Frequently asked questions
What is the difference between a conclusion and a discussion point?
A conclusion is something a reader would reach from your results chapter alone. A discussion point is something they would reach only with an argument you supplied. Both belong in the report, in different places, and conflating them is how overclaiming starts.
What is a negative finding in a qualitative study?
The supplied source never says — the terms are hypothesis-testing categories and neither is defined anywhere in six weeks. A workable reading is: anything you looked for and did not find, and anything your data ran against. That reading is this library's, not the source's.
What is a post hoc analysis and why does it get different treatment?
The source uses the term once and never defines it. The distinction it draws is between analyses your design planned and analyses you ran after seeing the data. The second cannot independently confirm what the data suggested, which is why it is presented as explanatory rather than as evidence.
How do I write recommendations?
Not from this material. The week's own analysis sequence names Recommendations as a closing step and never mentions it again — no definition, no scope, no audience. The boundary the other rules imply is that a conclusion states what you found and a recommendation states what to do about it, inheriting every limit on the conclusion beneath it.
Why check the abstract against the conclusions?
Because the abstract is written early, revised least and read most, and it is where a conclusion that was properly qualified in the body turns up with its qualifiers stripped. Do that reconciliation as the last action before submission.
References and source attribution
- The supplied teaching source: weekly study notes, slide decks and assessment activities for a master's-level research methods subject in project management. Author, institution and year not stated in the supplied files.
- The conclusions rules in the supplied slides carry an attribution to a research foundation's web address; the address as printed is malformed and does not resolve, so the source of the rules cannot be verified.
Suggested questions for Ask KEVOS
- How do I tell whether a conclusion is supported by my results?
- How should I report findings I expected and did not get?
- What is the difference between a conclusion and a recommendation?
- Why should post hoc analyses be presented differently?
- How do I stop my abstract from overclaiming?