Each project template gives a new project its first folders. This page explains every
one of them: what it is for, what to put in it, and how it is used as the research goes
on, from the first reading to the thesis.
The names are suggestions, not rules. Any folder can be renamed, moved or deleted, and
you can add your own at any depth. To choose a template first, see
which template to start from.
The workbench is still being built. Parts marked
Built exist in the local workbench, which is not yet
released; Being built is under way now, and
Planned is designed and scheduled but not yet written.
Reading and the literature
Where sources arrive, and what you make of them.
Literature Review
The sources your research stands on: papers, books, reports and grey literature, with the notes you take on them.
What goes in it
Imported sources, in subfolders by theme or question if that helps, with reading notes beside the sources they are about. The Empirical study and Qualitative study templates call it Literature, because there the reading is background to the data rather than a chapter of its own.
In practice
In a doctorate this is the folder every paper and the thesis draw on, and the reason the doctorate is one project: a source is imported once, annotated once, and cited from the transfer report, a paper in the second year and a thesis chapter years later. Resist a literature folder per paper.
In the workbench
Importing a file that is already in the project records where the new copy came from instead of creating a duplicate. Being built
Everything in scope for the review, as you find it. Sources you later exclude can move to a subfolder rather than being deleted, so the review can still say what it left out and why.
In practice
If the review is part of a doctorate, build this structure inside the doctorate project instead of starting a project for it: its sources are then the doctorate's sources, and the notes you make on them stay with the rest of your research.
What you think while reading: summaries, quotations you may use, questions and disagreements.
What goes in it
A note per source or per reading session, whichever you will keep up. Record the page of anything you might quote.
In practice
Reading notes are the raw material of the synthesis. A note that says where in the source it came from can be checked years later; one that does not has to be read for again.
In the workbench
Notes and annotations attached to the passage of the source they are about. Planned
What the sources say together: themes, comparisons, agreements, contradictions and the gaps your research fills.
What goes in it
Theme maps, comparison tables, concept lists and drafts of the argument that ties them together. In a systematic review, the analysis across the included studies: narrative synthesis, meta-analysis outputs, forest plots and certainty assessments.
In practice
The synthesis is where a review earns its conclusions, so keep each theme traceable to the sources behind it. In a doctorate, the gap it identifies is what the methodology then sets out to fill.
The folders that make a review reproducible, in the order the work happens.
Protocol
The plan for the review, written before the searching starts: the question, the eligibility criteria, where to search, and how studies will be screened, extracted and synthesised.
What goes in it
The protocol, its registration record (PROSPERO, OSF or similar), and each amended version with the reason for the amendment.
In practice
Examiners and journal reviewers compare the finished review with its protocol. Keep the registered version unchanged and file amendments beside it, dated, rather than editing it in place.
Exactly how the literature was searched, recorded well enough for someone else to repeat it.
What goes in it
The databases and registers searched, the full search string for each, the date each search ran and how many records it returned, exports of the results, and any hand-searching or citation chasing.
In practice
Reporting guidelines such as PRISMA ask for every search in full. A review that runs across several years of a part-time doctorate will be searched again before submission: save each run with its date instead of overwriting the last one.
The decision on every record, at each stage: in or out, and why.
What goes in it
Title-and-abstract and full-text decisions, reasons for exclusion, disagreements between screeners and how they were resolved, and the counts for the flow diagram.
In practice
The flow diagram is built from these numbers, and the reasons for exclusion are what a reviewer asks for. Recording them as you go takes little effort; rebuilding them at the end rarely gives numbers that add up.
The studies that passed full-text screening: the evidence the review is built on.
What goes in it
The full text of each included study, one document per study, with its supplementary material.
In practice
Keep this set equal to the final count in the flow diagram. If a study is added or removed after screening, record the decision in Screening too, so the two never disagree.
What was taken from each included study, in the same form for every study.
What goes in it
The extraction form, the completed extraction for each study, risk-of-bias or quality assessments, and correspondence with authors about missing data.
In practice
Pilot the form on a few studies before extracting from all of them, and keep the piloted version. Where an AI helps with extraction, record which model read which study and which of its values you accepted.
In the workbench
A record of which model extracted what from which study, and which of its suggestions you accepted. Planned
How the research is done, what it collects, and the protection participants are owed.
Methodology
How the research is designed, and why: the approach, the methods, and the argument for choosing them.
What goes in it
Research design documents, sampling plans, instruments and procedures such as questionnaires, interview guides and experimental protocols, and your reading on method.
In practice
In a doctorate the methodology grows from the transfer report into a thesis chapter. Keep superseded versions of instruments: data collected with the first version of a questionnaire is interpreted with the first version.
Permission to do the research, and the conditions that came with it.
What goes in it
Ethics applications and approval letters, amendments, participant information sheets, blank consent forms, data management plans, and approvals from the organisations where the research takes place.
In practice
Research that runs for years usually needs amendments as it changes; file each with its approval. Signed consent forms name participants, so if you keep them here, mark this folder as holding human-subject data.
In the workbench
Any folder can be marked as holding human-subject data. Built
What the research collects or generates, with the documentation that makes it usable.
What goes in it
Raw data, kept unchanged; cleaned and derived datasets; codebooks and data dictionaries; and a note of where each dataset came from and how it was processed.
In practice
Raw data is never edited: cleaning produces a new file. In a doctorate the same datasets feed several papers and the thesis, so one Data folder serves them all. If any of it identifies people, mark the folder as holding human-subject data.
In the workbench
Marking a folder as holding human-subject data gives everything in it the strictest level of protection, and moving it somewhere else never lowers that protection. Built
Material that comes from, or could identify, the people who took part in the research.
What goes in it
Recordings, transcripts, field notes about participants, completed consent forms and consent records, and anything else that could identify a participant.
In practice
This folder starts marked as holding human-subject data, so everything filed in it gets the strictest level of protection, and moving a file out does not lower it. Treat a transcript as anonymous only once it truly is.
In the workbench
The folder is created already marked as holding human-subject data. Built
Linking material to each participant's consent, and restricting which AI providers may process it. Planned
Analysis scripts and notebooks, statistical outputs, figures and tables, and notes interpreting them.
In practice
Keep each analysis runnable from the data it used: the script and its output together, not a table copied from somewhere. When a paper comes back from review a year later, you will need to run it again.
The analysis of qualitative material: codes, categories and the themes built from them.
What goes in it
Codebooks and their versions, coded extracts, memos written while coding, theme maps, and notes on agreement between coders.
In practice
A codebook changes as the analysis develops; keep its versions, because a theme means what the codes it was built from meant at the time. Coded extracts still come from participants: keep identifying detail out of them, or mark this folder as holding human-subject data too.
Where the reading and the data become papers, reviews and a thesis.
Submissions & Reviews
What comes back from sending a paper out: the versions submitted, and what reviewers said about them.
What goes in it
A folder per paper, holding the file sent to each journal or conference, reviewers' comments, your responses, and the accepted version. In a systematic review, the review manuscript as it was submitted; in a master's thesis project, a paper written from the thesis.
In practice
The paper itself is written in the workbench rather than kept here — this folder is for the record of sending it out. Never overwrite a submitted version: years later you should still be able to show which text a reviewer read, and that is what these files are for. A doctorate's papers are written from the project's shared literature and data, which is why they belong in one project rather than in projects of their own.
In the workbench
Papers are written in the workbench, a section at a time, and assembled with a generated contents list. Built
A version can be sealed: the workbench keeps a copy of every section exactly as it was, with a fingerprint of the text, so you can always show which version a supervisor read. Built
A sealed version can be saved as a Word document, with its headings, contents list and figures. Built
Sharing a sealed version with a supervisor who has no workbench, and importing their comments back anchored to the text. Planned
The doctoral thesis, from the first chapter plan to the version examined.
What goes in it
Chapter drafts, the outline, front matter and appendices, the submitted and examined versions, and corrections after the viva.
In practice
The thesis comes at the end but draws on everything before it, and chapters often grow from published papers. Keep the examined version unchanged and make corrections in a copy, so what the examiners read stays on record.
In the workbench
Comparing a chapter with the paper it grew from, and checking each claim against the sources it cites. Planned
The chapters of a master's thesis or dissertation, as they are written.
What goes in it
A document per chapter, the outline, and the versions shared with your supervisor and submitted.
In practice
Separate chapters let you send one to your supervisor without the rest. If a master's by research becomes a doctorate, carry on in the same project: these chapters are the start of its thesis.
Supervision, progress, and ideas that have no home yet.
Supervision
Your work with your supervisors, on the record.
What goes in it
Notes from each meeting, feedback on drafts, agreed actions and deadlines, and supervision agreements or plans.
In practice
Write up the agreed actions after every meeting and file them the same day. Over a part-time doctorate of six or seven years supervisors can change: this folder is how a new one catches up, and what you point to at a progress review.
In the workbench
Sharing a project or a single paper version with a supervisor, with their comments coming back anchored to the passage they are about. Planned
The formal points of a doctorate, and the evidence that you passed them.
What goes in it
Annual progress reviews, the MPhil-to-PhD transfer or confirmation report and its outcome, training records, conference presentations, and the notice of intention to submit.
In practice
The transfer report is often a doctorate's first long piece of writing, and it draws on literature already annotated in the same project. Each progress review asks what has changed since the last one; the last one is here to answer that.
Ideas, loose notes, things to follow up, possible future papers, and material not yet filed.
In practice
An inbox, not an archive: empty it now and then by filing things where they belong. An idea that keeps coming back may be the next paper, and deserves to be started as one.
The folders of the AI-supervision study template, for research whose subject is AI itself.
Experiments
The experiments the study runs across models, prompts and conditions.
What goes in it
What each experiment sets out to test, the conditions compared, their settings, and the results of each run.
In practice
Define an experiment before running it, and leave the definition unchanged once runs exist: a change makes a new experiment. Results mean little without the exact conditions that produced them.
In the workbench
Experiments with numbered versions, and every run recorded against the version it used. Planned
Models of how AI agents behave, built from the records their runs leave.
What goes in it
Analyses of recorded traces, taxonomies of agent behaviour and failure, and the measures and models derived from them.
In practice
Every request the workbench sends to an AI model is meant to leave a record, called a trace: the instructions, the model and its settings, the sources and the result. This folder is for what the research makes of those traces.
In the workbench
A trace for every request sent to an AI model. Planned
The standards generated code is held to. Software code, not the qualitative coding of Coding & themes.
What goes in it
Coding standards and style guides, the rules and checklists used to judge generated code, and their versions.
In practice
In research on AI coding tools the standard is part of the method: saying that code is acceptable only means something alongside the version of the standard used at the time.
Architecture, sequence and workflow diagrams, kept as text (Mermaid, PlantUML or similar) wherever possible, with exported images alongside.
In practice
A diagram kept as text can be compared between versions like any other writing, so you can always tell which version of the system a diagram in a paper describes.
Add a folder when you will keep coming back to what is in it. A piece of work that
belongs to bigger research gets its folders inside that project: a systematic review
done during a doctorate is a Systematic review folder in the doctorate project,
holding its own Protocol, Search strategy, Screening and the rest,
so its sources stay the doctorate's sources.
Wherever material that identifies people ends up, mark its folder as holding
human-subject data: everything in it gets the strictest level of protection, and
moving it somewhere else never lowers that protection. Built
