Reviewers read quickly, and they read in a predictable order. A clear structure lets them find your question, your method and your result without hunting. A muddled one can hide strong research.
Structure will not rescue weak work, but it lets good work be seen. This guide explains what each section of a research paper must do, how the structure varies between fields, and which structural mistakes cost authors the most. If you are earlier in the process, start with our beginner's guide to writing a research paper.
Key Takeaways
Most papers follow IMRaD: Introduction, Methods, Results and Discussion, plus a short Conclusion.
Each section has one job. If a paragraph does not serve that job, move or cut it.
Computing papers often add Related Work and Experiments, so follow the structure used by recent papers at your target venue.
The venue's author guidelines override any general advice, including this guide.
Outline the paper before you write, and check the finished draft with a reverse outline.
Why Structure Matters to Reviewers
An editor or reviewer forms a first impression from the organisation of a manuscript. A well-ordered paper signals care and familiarity with academic conventions, and it lets reviewers judge the science instead of decoding the layout.
Structure also supports reproducibility. When the Methods section is complete and separate from the Results, another researcher can check what you did and test whether they get the same outcome. And because publishers set their own rules, always read the venue's author guidelines before you finalise the layout.
The Standard Structure at a Glance
Section | Its job | The question it answers |
|---|---|---|
Title | Name the contribution | What is this paper about? |
Abstract | Summarise the whole paper | Should I read on? |
Keywords | Help readers find the paper | How is this indexed? |
Introduction | Set up the problem and the gap | Why does this matter, and what is new here? |
Related work | Place the paper in the literature | What has already been done? |
Methods | Describe what you did | Could someone repeat it? |
Results | Report what you found | What happened? |
Discussion | Interpret the findings | What does it mean, and how far can we trust it? |
Conclusion | Close the argument | What should I remember, and what comes next? |
Acknowledgements | Credit support | Who helped and who funded it? |
References | List every source | Where did the ideas come from? |
IMRaD and the Variants You Will Meet
IMRaD is the default in the natural sciences, medicine and many engineering fields. Authors often add a short Conclusion, and some venues fold it into the Discussion.
Computing and machine learning papers frequently use a variant: Introduction, Related Work, Approach, Experiments, Results, Discussion and Conclusion. Review papers, case studies and theoretical papers have their own patterns. The safest method is to open three or four recent papers from your target venue, copy their outline, and check it against the author guidelines.
Section by Section: What to Write
Title and Abstract
The title should state the contribution in plain words. The abstract should cover the problem, the method, the main result and why it matters, in a few sentences each. Check the venue's word limit, because limits vary. Write the abstract last, and make sure its numbers match the Results.
Introduction
A reliable pattern, often called the CARS model after the linguist John Swales, has three moves. First, establish the territory: say what area you work in and why it matters. Second, establish the niche: show the gap, the contradiction or the unanswered question. Third, occupy the niche: state what your paper does and how it is organised.
A weak introduction summarises the field for a page. A strong one reaches the gap within a few paragraphs.
Related Work or Literature Review
Group prior work by theme, not paper by paper. For each group, say what it established and where it falls short. End by pointing to the gap you address. Cover the work that is closest to yours, not everything that has been published on the topic.
Methods
Describe the design, the data, the tools and the analysis in enough detail for someone else to repeat the work. Justify your choices where a reader might ask "why that way?". Reporting guidelines for many study types are listed by the EQUATOR Network, and they can serve as a checklist.
Results
Report what you found, using tables and figures where they help. Present the results objectively and save the interpretation for the Discussion. Put the most important result first, and move bulky secondary tables to an appendix.
Discussion and Limitations
Explain what the results mean in light of your research question and the earlier literature. Say where they agree or disagree with previous studies, and offer reasons. Then state the limitations honestly, such as sample size, data quality or method constraints, and say how they affect the conclusions.
Conclusion and Future Work
Restate the main finding and the contribution in a few sentences, without adding new results. Then name realistic next steps. Readers often turn to the Conclusion early, so make it stand on its own.
Acknowledgements and References
Thank mentors, institutions and funders, and make sure anyone named has agreed. List every cited source in one consistent style, such as IEEE, APA, MLA or Harvard, and match each in-text citation to an entry.
A Skeleton You Can Copy
Use this as a planning outline and fill each line with one sentence:
Problem: What is the issue, and who cares?
Gap: What do existing studies miss?
Contribution: What does this paper add?
Approach: What did you do, in one sentence?
Evidence: What is the main result, with its number?
Limits: Where might the result not hold?
Next step: What should happen after this paper?
If you cannot fill one line, you have found the weak spot in the paper before a reviewer does. Our guide on finding a research gap can help with line two.
Ready to put your structure to work?
SAI Conferences organises IntelliSys, Computing Conference and CVC, with proceedings published by Springer in Lecture Notes in Networks and Systems. Each Call for Papers page lists the tracks and the current rounds. Read the Computing Conference call for papers
Structural Mistakes and How to Fix Them
Mistake | Why it hurts | Fix |
|---|---|---|
Introduction turns into a long literature survey | The reader loses the point of the paper | Keep only what sets up the gap, and move the rest to Related Work |
Methods leave out key details | Nobody can check or repeat the work | Add data, parameters, tools and analysis steps |
Results mix in long interpretation | Findings and opinion blur together | Report in Results, interpret in Discussion |
Discussion repeats the Results | No new insight is offered | Explain what the results mean and compare them with prior work |
Limitations are missing | Reviewers find the weaknesses themselves | State them and explain their effect on the conclusions |
Figures and tables are unlabelled or never cited | Readers cannot follow the evidence | Caption every item and refer to it in the text |
Abstract and Conclusion disagree with the Results | It looks careless and weakens trust | Check that all numbers match |
Irrelevant material is included | It buries the main contribution | Cut anything that does not support the research question |
Check Your Structure Before You Submit
Reverse outline. Write one line per paragraph describing its job. If two paragraphs share a job, merge them. If one has no job, cut it.
Question test. Check that each section answers its question from the table above.
Signposting. The Introduction should end with a short roadmap, and each section should open with its purpose.
Consistency. Terms, abbreviations and numbers should be the same everywhere.
Guidelines. Compare the final layout with the venue's template and page limit.
Frequently Asked Questions
Why does the structure of a research paper matter?
It helps reviewers find your question, method and result quickly, and it makes your work easier to check and repeat. A clear structure shows professionalism, but it cannot replace sound research.
Is IMRaD required for every paper?
No. It is the most common pattern, but some fields and venues use other outlines. Follow the author guidelines and the structure of recent papers at your target venue.
How long should an abstract be?
It depends on the venue, so check the author guidelines. Many venues set a limit somewhere in the low hundreds of words, and the abstract must fit it while covering the problem, method and main result.
In what order should I write the sections?
Many authors write Methods and Results first, then the Discussion, Introduction and Conclusion, and the Abstract and Title last. Choose the order that works for you, but always revise the Introduction and Abstract once the results are final.
Do I need a separate Related Work section?
Some venues expect one, especially in computing. Others expect the literature review inside the Introduction. Check recent papers from your target venue to see which they use.
What is the biggest structural mistake?
Many papers fail to make the gap and the contribution clear early on. If a reviewer cannot say what is new after reading the Abstract and Introduction, the rest of the paper has to work much harder.
Ready to Submit Your Paper?
With a clear structure, your paper is easier to review and easier to trust. SAI Conferences organises three in-person events, and each Call for Papers page lists tracks, review details and current submission rounds.
SAI event | When and where | Submit |
|---|---|---|
13th edition | 2 to 3 September 2027 Mercure Amsterdam City Hotel, Amsterdam, Netherlands | |
15th edition | 8 to 9 July 2027 Clayton Hotel Chiswick, London, United Kingdom | |
Computer Vision Conference 2027 | 15 to 16 April 2027 Mercure Amsterdam City Hotel, Amsterdam, Netherlands |
Submit to IntelliSys 2027 | Submit to Computing Conference 2027 | Submit to CVC 2027
Related Guides
SAI organises IntelliSys, Computing Conference and CVC. SAI cannot guarantee that any paper will be published or indexed, and the reviewers and indexing databases make those decisions. Confirm all details on each event page before you submit, and see the Open for submissions panel on this page for live round status.