Industry Playbooks
Research workflows for scientists are the structured processes for designing experiments, reviewing literature, analyzing data, and building the reproducible evidence base that moves a research program forward — from initial question through publication.
A third-year PhD student is struggling to write her dissertation proposal. She has run 47 experiments over two years. Some worked; most didn't. She has read — or at least downloaded — over 300 papers. She knows a lot. But when she sits down to write, she can't quickly answer the question: "What, exactly, have you established? What's the evidence?" The experiments are in lab notebooks in inconsistent formats. The papers are in a Zotero library she hasn't organized since her first year. The story of her research exists in her head; the documentation to support it is scattered.
A research workflow for scientists is the structured process for moving from research question through experiment design, data collection, literature integration, and analysis to conclusions that can be documented, communicated, and built upon. Without it, even productive research produces knowledge that lives in the researcher's memory and fragmented notebooks rather than in usable, reproducible form.
Reproducibility: Every significant result should be reproducible — by you a year from now, by a colleague, by a reviewer. The research workflow must capture enough procedural detail to make reproduction possible.
Cumulative progress: Each experiment should build on prior knowledge — what's been established, what's been ruled out, what the current best hypothesis is. A research workflow that tracks this progress enables strategic experiment design, not random exploration.
Literature integration: New experiments should be informed by existing literature. A literature review that updates with the field — incorporating new preprints, new methods, recent contradictory findings — keeps the research relevant.
Documentation for publication: The methods section of every paper is a distillation of the research workflow. A workflow that captures the decisions, conditions, and rationale for experimental choices produces methods sections that can actually be written.
Before designing an experiment, define what you're actually testing:
From curiosity to question:
Literature integration at hypothesis stage:
Document the hypothesis before testing: Write a 1-2 paragraph hypothesis statement before designing the experiment. This prevents HARKing (Hypothesizing After Results Known) — the problematic practice of reverse-engineering hypotheses after seeing data.
A well-designed experiment answers the question efficiently with appropriate controls:
Controls:
Power calculation: For studies comparing groups, a power calculation before beginning determines how many samples/replicates are needed to detect a meaningful effect with adequate statistical power. Underpowered studies produce unreliable results; the work is wasted either way.
Pre-registration (for hypothesis-testing studies): Pre-registering your hypothesis and analysis plan on OSF or AsPredicted before data collection creates a time-stamped record of what you predicted and how you planned to analyze it — a key protection against unconscious bias in analysis.
Protocol finalization: Before beginning, ensure the protocol is written, versioned, and complete. Experimental modifications after starting should be documented as deviations, not silently incorporated.
A sustained literature review workflow keeps you current and connected to the field:
Systematic database searches:
Alert systems to stay current:
Efficient reading workflow: Apply a triage rather than reading everything fully:
Capture as you read: Reading notes captured at point of reading are far more useful than returning to a paper to re-extract the key finding. For each paper you read:
The data collection phase is where research workflow discipline pays off most:
Every experiment documented with:
Same-day documentation: Document while the experiment is running or immediately after. Notes written from memory the next day miss details that may matter for interpretation or troubleshooting.
Failure documentation: Document failed experiments as carefully as successful ones. A properly documented failure is informative; an undocumented failure is a forgotten effort that may be repeated unnecessarily.
Analysis before and after data collection: Your analysis plan should be written before data collection (see pre-registration, Stage 2). Exploratory analyses after seeing data are legitimate — but they should be labeled as exploratory, not presented as confirmatory.
Analysis documentation: If using code for analysis (R, Python, MATLAB), the analysis code is part of the research workflow and should be:
Statistical approach: Know your statistical approach before running analysis. Choosing statistics after seeing results (p-hacking) inflates false positive rates. Pre-specify the primary analysis; additional analyses are exploratory.
Synthesis document after each significant result: After a significant experimental finding, write a 1-page synthesis:
Communication planning:
| Tool | Use | Notes |
|---|---|---|
| Zotero | Literature management and citation | Free; browser extension; Word/Google Docs integration |
| PubMed + Google Scholar | Literature search | Complementary; bioRxiv for preprints |
| OSF | Pre-registration and data sharing | Also a collaboration platform |
| Benchling / Labarchives | Electronic lab notebook | ELN with protocol management |
| GitHub | Code version control for analysis | Also enables reproducibility checking |
| R / Python | Statistical analysis | With documented, version-controlled code |
| WebSnips | Preprint and lab intelligence tracking | Web-clipping for scientific intelligence |
WebSnips for scientific research workflows: Scientific knowledge moves through multiple publication stages — preprints, conference posters, talks, and methods blog posts — before formal journal publication. WebSnips captures specific web pages with date and source URL, organized by project or topic. A preprint you cited as "submitted" can be clipped when it becomes a published paper; a methods protocol published on a lab's website can be clipped for reference alongside the protocol you're using. For fields with fast-moving preprint ecosystems (computational biology, machine learning applied to science), the ability to clip and date-stamp current preprints is essential for tracking who published what and when.
A cell biology PhD student, Priya, is studying how mechanical forces affect stem cell differentiation. Her research workflow:
Hypothesis development: After reading recent literature on mechanosensing, she formulates: "YAP/TAZ nuclear translocation in response to substrate stiffness increases within 4 hours of cell plating and is necessary for osteogenic commitment." She registers this on OSF before beginning.
Experimental design: She designs the experiment with three substrate stiffness conditions (1 kPa, 10 kPa, 40 kPa), positive control (stiff substrate + osteogenic media), and negative control (soft substrate + osteogenic media). Power calculation: n=6 per condition for immunofluorescence quantification.
Literature tracking: She sets up PubMed alerts for "YAP TAZ mechanosensing" and a Google Scholar alert for "substrate stiffness stem cell differentiation." Each week she spends 30 minutes triaging new papers; significant ones get reading notes in her Zotero groups.
Experiment 1 — documentation: In Benchling: Date: Sept 5, 2026 | Protocol: Polyacrylamide gel preparation v4.2 + Cell plating v2.0 Conditions: PA gels at target stiffness 1, 10, 40 kPa. Human MSCs, passage 4, batch 2026-07-15 Data: /data/YAP_stiffness/2026-09-05/ Observations: Gels appear well-formed; cell adhesion at 4h looks good in all conditions Staining result summary: YAP nuclear-to-cytoplasmic ratio increased with stiffness as expected. Effect at 40kPa was smaller than predicted — investigate protocol variable? (see troubleshooting note)
Synthesis after experiment 1: YAP translocation is stiffness-dependent but magnitude at 40kPa lower than published data. Three possible explanations: cell passage, gel validation needed, antibody batch. Plan: gel mechanical testing to verify stiffness, and repeat with fresh antibody batch.
Data sharing requirements: NIH, NSF, and many other funders now require data management plans and data sharing. Increasingly, journals require data availability statements. A research workflow that documents and organizes data makes compliance straightforward rather than an end-of-project scramble.
Statistical rigor: The reproducibility crisis in science is substantially caused by underpowered studies, undisclosed multiple comparisons, and selective reporting. The workflow practices above — pre-registration, power calculation, documented analysis plans — address these directly.
Authorship and contribution tracking: Large collaborative projects need documentation of who contributed what. Lab notebooks and data collection records that capture operator identity are part of authorship documentation.
Conflict of interest: Research workflows should include documentation of funding sources, as these are required for publication disclosures and grant reporting.
Mistake 1: Hypothesis after results (HARKing). Designing the hypothesis to fit the data after seeing results inflates false positives and undermines reproducibility. Pre-register hypotheses and analysis plans before data collection.
Mistake 2: Literature reviews as one-time events. Reading all the papers when you start a project and then ignoring new publications is how you miss contradictions, method improvements, and scooping events. Continuous literature monitoring is a research workflow requirement.
Mistake 3: Undocumented protocol modifications. "We changed the incubation time but didn't write it down" is the single most common source of irreproducible results. Every protocol modification is a documentation event.
Mistake 4: Analysis code without documentation. Analysis code that only its author can run, using only the original dataset and the original environment, cannot be reproduced. Documented, version-controlled analysis code is part of the scientific record.
A research workflow for scientists is what makes a research program genuinely cumulative and reproducible. The discipline of pre-registering hypotheses, capturing experimental details same-day, tracking literature continuously, and documenting failures as carefully as successes — this is not bureaucratic overhead. It's the practice that transforms individual experiments into a coherent research program, makes your methods section writable when submission time arrives, and ensures that your labmates two years from now can build on your work rather than starting over. The science is hard enough; the workflow should make it harder to waste what you already know.
More WebSnips articles that pair well with this topic.
AI knowledge work for scientists is transforming literature review, hypothesis generation, data analysis, and scientific writing — while raising critical questions about accuracy, reproducibility, and the scientific judgment that remains irreplaceable in research.
Knowledge management for scientists is the practice of organizing literature, experimental data, protocols, and research findings in searchable systems — ensuring that each experiment builds on prior work, that citations are retrievable at submission time, and that lab knowledge survives personnel transitions.
A note-taking system for scientists must capture experimental observations, paper reading notes, seminar insights, and lab meeting discussions — in retrievable, reproducible form that supports the lab notebook record, the literature review, and the ultimate publication.