Distinguish traceability, reproducibility and replication
Traceability first means that a reader can inspect the route from question and material to result and conclusion. Depending on the field, reproducibility may mean rerunning an analysis with existing data, recalculating from documented inputs or rebuilding a workflow. Replication often investigates the question with new data, participants or cases.
Research communities do not use these terms identically. The assessment brief should state which performance is expected and provide a field-appropriate example. Markers apply the meaning announced for the assessment rather than selecting a new definition after submission.
A historical source interpretation, qualitative interview study, laboratory experiment, software project and mathematical proof require different evidence. This grid describes assessable functions and remains intentionally discipline-open. It does not replace programme rules or create a universal institutional standard.
Make selection and provenance of materials inspectable
A bibliography alone does not explain how sources, cases, measurements or datasets entered an analysis. Relevant information includes the search or selection route, period, inclusion and exclusion criteria, and consequential preprocessing. For newly collected data, document the instrument, recruitment or sampling logic and deviations from the plan.
In humanities work, archive references, edition details and a reasoned corpus boundary can make evidence reconstructable. In creative practice, design iterations, material decisions and test conditions can establish the inquiry trail. A discipline may specify suitable forms without treating only open quantitative data as scholarly.
Separate access from documentation quality. A public repository can be poorly explained, while a protected dataset can have an excellent method record. A student must not breach consent, copyright or confidentiality merely to satisfy a transparency criterion.
Document consequential steps and the working environment
A methods section should do more than name a technique. It records sequence, meaningful parameters, selection or stopping rules and treatment of exceptions. Software work may require versions, dependencies, a start instruction and random seeds. Interview research may require the topic guide, analytical stages and reflection on the researcher's interpretive position.
Not every click warrants a log. Focus on information whose change could plausibly affect the output or its interpretation. A knowledgeable reader should be able to distinguish the application of a standard method from a consequential local adaptation.
Automated systems, AI tools and external services are described through their function, workflow stage and retained contribution. A brand name is not a method. Where a service may later disappear or change, an export, parameter description or justified alternative route improves continuity.
Treat protected and non-repeatable objects fairly
Personal data, confidentiality agreements, hazardous methods, copyrighted material or vulnerable locations should not be disclosed merely to enable public repetition. Assessment can instead use a data description, collection logic, code demonstrated with synthetic material, access conditions or a controlled review route.
A singular event cannot be recreated identically. Good practice then documents the source situation, date, position of observation and interpretive steps so that another person can inspect the argument and reach a reasoned view using the same accessible materials.
A missing permission and poor planning are nevertheless different. If protection needs were foreseeable, a management plan should have provided suitable evidence. If a restriction arose unexpectedly, assess the quality of the justified adaptation rather than pretending the constraint does not exist.
Calibrate criteria and anchor judgements in evidence
Before marking, the discipline assigns weights and any minimum requirements to the announced criteria. Examples can show strong documentation across more than one methodological tradition. Markers identify specific passages or artefacts before awarding a level, leaving an inspectable reason for the judgement.
A successful independent rerun can be powerful evidence, but a dissertation assessment may not require or permit one. Conversely, executing a script does not establish that the question, material selection and interpretation are academically sound. The grid assesses the documented inquiry, not technical execution alone.
When something is unclear, distinguish an absent explanation, a contradiction and an inaccessible artefact. Any clarification follows the published assessment process. Criteria must not be tightened because a surprising or inconvenient result was found.
Finish with a bounded handover test
A concise handover test asks a knowledgeable reader to use the dissertation and permitted appendices to identify three routes: locate the material, follow the method and check a central result. The reader records missing information without attempting the entire study again. This can expose documentation gaps, but it belongs in marking only when announced beforehand.
Use the AI disclosure rubric for tool records and the group contribution matrix for collaborative projects. Find further resources in the higher education practice hub. The single foundation route, how verifiable checking works, gives the wider evidence context.
After assessment, review recurring documentation gaps across anonymised work. Improve the next brief, examples or teaching where the same ambiguity recurs. Do not use cohort patterns to retroactively adjust current criteria or to impose a preferred research method. A transparent revision history distinguishes educational improvement from changing the rules after performance has been observed.