The research paper workflow: from idea to submission in 7 stages
A research paper moves through seven stages: concept, novelty check, method and code, results, manuscript, integrity checks, and journal submission with review. Each stage should end with something you can point to, such as a one-page concept, a comparison table, a results folder or a formatted file. When a stage has no output, the next one stalls.
Many papers stall because the work happens in the wrong order: writing before the results are firm, or choosing a journal after the paper is formatted for another one. This guide lays out the whole path in one place, so you can see what each stage needs and when to move on.
What are the stages of a research paper?
The details differ by field, but the sequence is stable. The table shows each stage, the output you should have before you leave it, and a question that tells you whether it is done.
| Stage | What you produce | Done when you can answer |
|---|---|---|
| 1. Concept | A one-page concept note | What is the problem, and what do I plan to do about it? |
| 2. Novelty check | A table of the closest papers | What does my work add that they do not? |
| 3. Method and code | A plan and working code | Could someone else run this and get the same output? |
| 4. Results | Tables, figures and a run log | Do the results support the claim, including where they are weak? |
| 5. Manuscript | A full draft with references | Does every claim point to a result or a source? |
| 6. Integrity checks | A similarity report and a disclosure statement | Is every borrowed idea cited and every AI or tool use declared? |
| 7. Journal and review | A formatted submission and, later, a response letter | Does it follow this journal’s guidelines, and have I answered every comment? |
Stages 1 and 2: Concept and novelty check
Start by writing the idea on one page: the problem, the gap, the aim and the planned approach. A page forces choices that a notebook lets you avoid. Then check whether the idea is already published, using several databases and a comparison table. The full method is in how to check if your research idea is novel.
Tip Do not skip stage 2 because the idea “feels new”. A week of reading here is cheaper than a month of work on something already done.
Show the concept and the table to your supervisor before you build anything. Agreement at this point saves the most rework later.
Stages 3 and 4: Method, code and results
Turn the concept into a method you can run. For computational work this means code. Keep the code in separate files by task (data loading, the proposed method, baselines, metrics, plots), record the software versions, and fix the random seeds. Run it, save every figure and table with a file path, and keep the log.
- Run the baselines under the same conditions as your method, using the settings their authors report.
- Keep weak or negative results. They belong in the paper, and a reviewer will ask about them.
- Never adjust results to match a target. If the method does not beat the baseline, that is the finding, and you decide what to do next with your supervisor.
A later post covers how to share research code so others can reproduce it.
Stage 5: The manuscript
Write from the results outward. Draft the results and methods first, then the discussion, then the introduction and abstract. This order keeps the claims tied to what you found. The usual structure is Introduction, Methods, Results and Discussion (IMRaD), adapted to your field. See how to write a first draft from your results.
A first draft is allowed to be rough. What it must have is a number or a source behind every claim.
Stage 6: Integrity checks
Before you choose a journal, check the draft for text that matches other sources and fix it by rewriting and citing properly, using the checker your journal or institution requires. Also write down any use of AI tools, because many journals and universities now ask for a statement. A further post covers this: using AI in research papers.
Stage 7: Journal, formatting, submission and review
Choose a journal by scope and audience, check that it is trustworthy, and read its author guidelines before you format. After submission, the reviewers’ comments are part of the process, not a verdict. Split them into single points and answer each one. See how to choose the right journal, journal formatting requirements and how to respond to reviewer comments.
Where do researchers lose the most time?
- Moving to code or writing before the idea is checked against published work.
- Re-running experiments because settings, seeds or data versions were not recorded.
- Writing the introduction first and then rewriting it when the results change.
- Choosing the journal last, then reformatting the whole paper.
- Leaving the similarity and disclosure checks until the day of submission.
- Replying to reviewer comments in one block of text instead of point by point.
Each of these is a stage done out of order or skipped. The fix is the same each time: finish the output of a stage before you start the next.
How do I plan a paper by stage?
Give each stage an owner, an output and a date. A simple planning table is enough, and it makes delays visible early. The dates are yours, because the time each stage takes depends on your field, your data and your supervisor’s availability.
| Stage | Output | Who signs off | Date |
|---|---|---|---|
| Concept | One-page note | Supervisor | Your date |
| Novelty check | Comparison table | Supervisor | Your date |
| Method and code | Working, documented code | You and a co-author | Your date |
| Results | Tables, figures and logs | Supervisor | Your date |
| Manuscript | Complete draft | All authors | Your date |
| Integrity checks | Similarity report and disclosure | You | Your date |
| Submission | Formatted files and cover letter | All authors | Your date |
Which stages need a second pair of eyes?
Three points repay a review more than any others. After the novelty check, because a wrong decision here wastes the most time. After the results, because a flaw in the analysis changes everything that follows. After the full draft, because co-authors must approve what goes out under their names.
- Ask your supervisor to read the comparison table, not only the idea.
- Ask a co-author to run the code from the README on another computer.
- Ask someone outside the project to read the abstract and say what the contribution is.
A small example of one project
Suppose a PhD student wants to predict crop yield. The concept note names the problem and a gap (soil moisture is rarely used). The novelty check finds three close papers and one that already uses soil moisture in a different region, so the student narrows the claim to a new region and season. Code and results come next, with a weather-only baseline. The manuscript is written from the results table. The similarity report flags the methods paragraph, which is rewritten and cited. A journal with a regional focus is chosen, and after submission one reviewer asks for another baseline, which becomes a new analysis and a new paragraph.
This is an invented example to show the order of the work, not a case study. Notice how each stage changed what came after it.
Where a tool can help
EasySeer follows this sequence on six steps: input, novelty check, code generation, code execution, manuscript generation, and formatting and delivery. It writes the novelty report, code files, results and a first draft for you to read and approve at each step. You stay the author. You check every result and every reference, and nothing moves forward until you approve it.
The same stages work without a tool. A checklist, a folder per stage and a supervisor’s sign-off at stages 2, 4 and 5 will already remove most of the delays above.
See all eight stagesRead: how to make one project
Questions
How long does it take to write a research paper?
It depends on the field, the data and how many rounds of review follow, so any single number would mislead. Plan by stage instead, and set a date for each stage’s output.
Should I choose the journal before or after writing?
Shortlist it early, after the novelty check, so you know the audience, length and reference style. Confirm the final choice before formatting.
Do I need code for every paper?
No. The stages change by field. For computational and data work, code and results are stages 3 and 4. For other work the same stages hold a different kind of output, such as interviews, a model or a proof.
Where does peer review fit?
In stage 7. Treat the comments as a list of tasks, and reply to every one, including those you disagree with.
Keep reading
How to make one project, step by step
A project takes a handwritten concept to a first manuscript draft in six steps: Input, Novelty check…
Guide · 2 min readHow to prepare your inputs
Only the handwritten concept is required. Add a proposal, base papers and datasets if you have them:…
Guide · 2 min readEthics and AI disclosure
EasySeer creates only a first draft from your concept. You stay the author: verify every result and …
FeaturesAll eight stages
Every stage of the workflow, with real screens.
Sources
Checked on 2026-10-24. Policies change, so read the current version before you submit.
Published 2026-10-24. Last updated 2026-10-24.