RSearch Research Pilot
This pilot is designed for anyone evaluating RSearch as a complete quantitative research platform, including doctoral researchers, master's students, faculty, dissertation chairs, methods instructors, research staff, analysts, and research teams. Test the workflow as you would actually use, teach, supervise, review, or evaluate it, not only isolated statistical procedures.
Open the sample dataset or a de-identified file, verify the detected columns and file summary, then leave and reopen the analysis workflow.
Check: CSV/TSV/Excel/SPSS/Stata/SAS import, multi-sheet handling when applicable, 100 MB guardrails, and recovery behavior.
Use Overview, Variables, Missingness, Preparation Log, Coding, Distributions, Outliers, and Constructs. Apply, retain, or plan at least one preparation decision.
Check: clarity of warnings, researcher control, construct grouping, preparation traceability, and whether any guidance feels misleading.
Choose at least two methods from different families. Use the recommendation and method-eligibility guidance, configure the analysis, and run it.
Check: method discovery, parameter clarity, assumption handling, error messages, execution speed, and whether you understand why the method is eligible.
Inspect tables, interpretation, assumptions, effect evidence, provenance, and reproducible R code. Copy a result table, create a Research Report, and review reporting or citation-format choices where available.
Check: dissertation- or manuscript-ready table width, appropriate reporting structure, no unnecessary columns, copy/paste quality, numerical consistency, and whether the prose overstates the evidence.
Try Built-In Support first, then optionally choose AI-assisted Stats Assist. Ask for help with method choice, assumptions, or interpretation. Deliberately ask one ambiguous question and one question that should trigger a caution or limitation.
Check: clarity of the Built-In Support versus AI choice, consent and privacy messaging, usefulness, restraint, transparency, and whether the answer stays grounded in current RSearch evidence instead of inventing results.
Open SEM Workspace, reuse saved constructs, build relationships manually and with the text builder, estimate a qualified model, then use Review My Model to work through the fitted evidence.
Check: ease compared with SmartPLS, construct reuse, mediation/moderation workflow, higher-order modeling, PLS/PLSc and qualified CB-SEM choices, diagram values, reliability and validity evidence, repeated Prediction, cSEM CVPAT, MICOM/group path comparison when groups are relevant, Action Priorities when theoretical scale ranges are known, exports, and fitted-result questions.
Complete a browser-local analysis, refresh the app, open History, reopen the saved result, and download its retained evidence/report. Then confirm that new computation correctly asks you to reopen the source file because raw rows are not persisted.
Check: recovered tables, interpretation, provenance, reproducible code and report export; clear evidence-only labeling; local deletion; and a clear distinction between retained result evidence and memory-only raw rows.
First confirm that the core research workflow remains usable without a Google account. Then sign in with Google for Collaboration, create a collaboration project, invite another tester as a reviewer and as an editor, add an anchored review comment, record a revision-specific approval or changes-requested decision, inspect version history, and deliberately create a conflicting edit.
Check: guest access to the research workflow, identity flow when Collaboration is chosen, invitation handling, reviewer/editor role clarity, anchored comments, version history, reviewer approval, conflict recovery, and whether the workspace supports supervisor, dissertation-chair, methods-review, peer-review, or research-team workflows.
Install RSearch when available, close and reopen it, revisit a previously opened workflow, and test the app after a network interruption.
Check: installation, cached/offline assets, version refresh after reload or relaunch when an update is available, clear messaging when data must be reopened, and mobile/tablet usability.
Optional Alpha test
Use a real but non-confidential business question, or load one of the built-in examples. Frame the decision, outcome, design, comparison, segments, candidate measured factors, owner, and materiality criterion. Then review the generated Executive Brief, Evidence Plan, data requirements, open questions, claim boundaries, decision risks, and next actions.
Check: whether the blueprint would save meaningful time before an analyst touches the data; whether it asks the right executive questions; whether the evidence families are useful without pretending to choose a method; whether the claim boundaries prevent overreach; and whether the downloaded blueprint is good enough to hand to an analyst, research team, or executive reviewer.
Use the Send feedback control available throughout RSearch. It asks what you were trying to do, what you expected, what happened, and whether the feedback concerns correctness, usability, accessibility, a missing feature, or a suggestion.
RSearch attaches only visible technical context such as the release channel, page, safe method identifier when available, browser family, workflow step, SEM beta status, and an anonymous feedback-session identifier. It does not attach raw dataset rows, file contents, research notes, report contents, or arbitrary URL parameters. Do not include raw participant data or confidential research records.
Review known limitationsFor general support, you can also email contact@rsearch.io.
Before or during the pilot, reviewers can inspect how RSearch qualifies analyses, the boundaries that remain, the accessibility program, and institutional-readiness preparation.
