Milestone map
Milestone map
3 milestones
Design collection protocol or select open dataset
1–2 weeks
Either design a field or laboratory data collection protocol with clear variables and sampling strategy, or select and justify a public biodiversity or biological dataset. Accessible alternative: public repositories (iNaturalist, GBIF, NCBI) contain peer-quality data that fully satisfies this outcome — no field access required. The proof standard is identical whichever route you take.
Proof required
Submit a written protocol or dataset justification (2–3 pages) covering: your biological question, your variables and sampling strategy (field/lab) or dataset source and accession, your rationale for the collection design or dataset selection, and your planned statistical analysis approach.
What gets checked
- Biological question is specific and testable — 'Does species diversity correlate with habitat edge distance in urban parks?' not 'What lives in parks?'
- Protocol or dataset selection includes a clear justification — why this design or dataset can answer the question, not just that it exists
- Statistical approach is named at proposal stage — t-test, ANOVA, regression, species diversity index — confirms the submitter has thought through the analysis before collecting
Common mistakes
- Designing a protocol without considering what statistical test the data will support — data collected without a planned analysis often cannot answer the intended question
- Selecting a public dataset because it was the first result without checking whether it contains the variables needed to address the question
Resources
Foundationstart here
Depthgo deeper
Masteryfor the dedicated
What a verifier looks for
- Ask the submitter to state their null hypothesis in one sentence — confirms the question is genuinely testable.
- Ask why they chose this sampling strategy or dataset over alternatives — a considered choice indicates genuine understanding of the options.
- Ask what statistical test they plan to use and why it is appropriate for their data type — a correct justification (e.g., 'I'll use Mann-Whitney because my count data is not normally distributed') indicates methodological literacy.
Collect or extract data and perform initial analysis
2–4 weeks depending on collection method
Execute your collection protocol or extract and clean your chosen dataset, then perform an initial statistical analysis and produce at least one results figure. The initial analysis should be honest about what the data does and does not show — over-interpretation of preliminary results is a common error.
Proof required
Submit your raw data table or extracted dataset file, your R or Python code for the initial analysis (or a spreadsheet with clearly labelled calculations), and at least one results figure with a caption explaining what it shows.
What gets checked
- Raw data is submitted alongside the analysis code — the analysis should be reproducible from the data provided
- At least one figure is provided with an accurate caption that describes what the figure shows, not what the submitter hopes it shows
- Analysis code or spreadsheet is readable and structured — not a single unlabelled script with no comments indicating what each step does
Common mistakes
- Collecting data without a structured recording method and then trying to analyse an inconsistently formatted spreadsheet — data entry format decisions must be made before collection, not after
- Reporting preliminary results as conclusions before running the planned statistical test — initial figures are exploratory, not confirmatory
Resources
Foundationstart here
Depthgo deeper
What a verifier looks for
- Ask the submitter to explain what the figure shows and what they would conclude from it provisionally — a genuine analyst gives a hedged interpretation, not a confident conclusion from preliminary data.
- Ask what the biggest data quality issue they encountered was and how they handled it — realistic data collection always surfaces quality problems.
- Ask whether the preliminary results are in the direction they expected — and if so, ask how they would guard against confirmation bias in the final analysis.
Complete analysis report with methodology Q&A
2–3 weeks to write up and schedule review
Present your completed analysis to a biologist who will challenge your methodological choices. The Q&A session is the proof standard that distinguishes genuine scientific reasoning from AI-generated analysis — an informed reviewer will raise scenarios or alternative interpretations you cannot have scripted in advance.
Proof required
Submit your final analysis report (4–6 pages: background question, methods, results, discussion, limitations) and a written record of your Q&A session: the specific questions the reviewer raised about your methodology or interpretation and your responses. The reviewer must be named and their biological background stated.
What gets checked
- Report includes a limitations section that honestly acknowledges what the analysis cannot conclude — not just what it found
- Q&A record shows at least three specific methodological challenges from the reviewer and substantive responses from the submitter
- Reviewer is a biologist, ecologist, or field scientist — not a general academic or a friend without domain knowledge
Common mistakes
- Writing a discussion section that only confirms the hypothesis without engaging with alternative explanations — a genuine scientific discussion considers what might have produced the same data by a different mechanism
- Using a reviewer who agrees with everything without challenging any methodological choice — the Q&A must surface at least one real challenge
Resources
Foundationstart here
Depthgo deeper
What a verifier looks for
- Ask the submitter what the reviewer's strongest challenge was and how they responded — someone who went through a genuine Q&A can recall this clearly.
- Ask whether the reviewer identified any limitation the submitter had not included in their report — an honest 'yes' indicates a genuine exchange.
- Verify the reviewer is a biologist with specific field or laboratory experience relevant to the data type — general academics without biological research experience are not sufficient for a methodology Q&A on ecological or molecular data.