Build a thesis writing plan around claims, evidence, chapters, and revision

Turn the research question into a claim map

Write the main answer you currently expect, then list the claims required to support it. Mark each as conceptual, empirical, methodological, or contextual. Draw dependencies between them. This map is provisional, but it reveals whether your planned chapters collectively answer the question.

Beside each claim, record the evidence you have, the evidence still needed, and what would disconfirm it. A thesis is stronger when the plan includes ways you may be wrong instead of treating every task as confirmation of the first proposal.

For each major claim, record the evidence, the reasoning that connects it, a credible objection, and the limit you accept. This is more revealing than a topic outline. A claim with evidence but no warrant may depend on an unstated assumption; a claim with no plausible objection may be descriptive rather than arguable. The map becomes a diagnostic tool for research and analysis, not a decorative diagram made after the chapters exist.

Give chapters contracts

Create a one-page contract for each chapter: purpose, central claim, contribution to the whole thesis, evidence, method, scope boundary, relationship to adjacent chapters, and definition of done. Review these contracts with your supervisor before prose becomes expensive.

A chapter is not done because it reaches a word count. It is done when its claim is supported, its evidence is traceable, its limits are explicit, and its handoff to the thesis argument works. Word ranges are constraints, not definitions of intellectual completion.

A chapter contract should name its question, provisional answer, method or material, contribution to the whole thesis, and exit condition. The exit condition matters: it might be a completed analysis of a defined corpus, not 'everything relevant has been read'. Add inputs required from other chapters and an explicit non-goal. That boundary prevents background sections from expanding until they absorb the work the chapter was meant to support.

Plan by dependencies, not chapter numbers

Some chapters can be drafted before data collection ends; others depend on analysis, ethics approval, archive access, or a settled framework. Put those dependencies on the timeline first. Start independent work while slow processes run.

Identify the critical path: the smallest chain of delayed tasks that would delay submission. Give those items earlier internal deadlines and contingency plans. A beautiful weekly schedule cannot rescue an interview programme that began two months late.

Build a dependency board with three labels: ready, waiting on evidence, and waiting on a decision. A methods clarification may unlock two analysis sections; another week polishing the introduction may unlock nothing. Work first on items that reduce uncertainty or release several downstream tasks. Chapter numbers still matter for the reader, but they are often a poor order for the writer's week.

Plan the evidence and decisions that unlock pages; the pages will not unlock the evidence for you.

Convert months into weekly deliverables

For the next six weeks, define outputs a supervisor could inspect: a coded sample, a comparison table, a section outline, 1,500 rough words answering one sub-question, or a revised figure. ‘Work on literature review’ is activity without an observable finish.

Keep one primary writing deliverable and one supporting research task per week. Leave capacity for teaching, administration, illness, and difficult findings. A plan using every available hour is a plan that becomes fiction after the first interruption.

Define weekly deliverables as inspectable objects: coded interviews 1 to 5, a 700-word analysis of theme B, a revised figure with caption, or a decision memo on excluding a dataset. 'Work on chapter four' cannot be evaluated and easily expands around other demands. Estimate each object, include reading only when it changes an object, and leave capacity for the administrative and emotional friction that research projects inevitably contain.

Use supervisor meetings as decision gates

Send a short pre-read stating what changed, what evidence is attached, and which decision you need. Record the decision, owner, and next consequence afterward. This prevents meetings from becoming general reassurance sessions whose advice cannot be translated into the draft.

When feedback conflicts across meetings, return to the chapter contract and ask which requirement changed. Do not try to satisfy two incompatible directions silently. Resolve the decision while the supervisor can see it.

Send a short pre-meeting note with progress, the evidence behind one decision, and no more than three questions. During the meeting, separate advice from agreed action. Afterwards, record what changed, who owns the next step, and the date of any decision that affects scope. This turns supervision into a series of traceable gates and reduces the risk of interpreting an exploratory comment as a permanent instruction months later.

Reserve the whole-document revision

Protect substantial time after all chapters exist. The thesis-level pass checks that terms stay stable, methods and results agree, chapter conclusions support the final answer, cross-references work, and the introduction accurately promises the document now written.

Set an internal complete-draft date well before submission and work backwards through structural revision, supervisor review, corrections, references, figures, formatting, and deposit requirements. The buffer is part of the research plan, not spare time for procrastination to consume.

Reserve separate passes for argument, chapter sequence, evidence, cross-references, terminology, and proof. Whole-document revision also needs a contradiction list: definitions, sample counts, dates, labels, and claims that recur across chapters. Freeze major structural changes before final formatting. Otherwise a late section move can break numbering, references, tables, and the carefully proofread language around them all at once.