Private by architecture
Import, computation, and model training run on your Mac. Your data never leaves the device — built for IRB, FERPA, and grant-compliance realities.
Native macOS · On-device
Serious research analysis. Entirely on your Mac.
RSCHR is a native research & evidence workbench for institutions, evaluators, and analysts. Import data, run rigorous statistics and predictive models, and publish grant-ready evidence — without a byte ever leaving your machine.
Import
Analyze
Predict
Publish
Notarized & signed by Apple
100% on-device
No account to analyze
Validated vs. Python pipeline
The tension
Cloud tools demand you upload sensitive data. Spreadsheets can't run real inference. BI dashboards look polished but prove nothing.
The shift
RSCHR collapses that gap.
It brings the full arc of evidence work — import, rigorous statistics, predictive modeling, and publication — onto one native Mac app, where the computation happens locally and the data simply never leaves.
The result is analysis you can defend in a grant review or an IRB packet, produced on hardware you already trust.
Why RSCHR
Four reasons it earns a place in serious research.
Real statistics
Correlations, hypothesis tests, effect sizes, AUC-ROC, and confidence intervals — validated to 100% parity against a reference Python pipeline.
Predictive modeling, no code
Train classifiers and regressors on your own data with live threshold tuning, feature importance, and model comparison — powered by Apple's on-device ML.
From data to white paper
Projects, reports, surveys, and a paper renderer take you from a raw CSV all the way to grant-ready evidence and polished PDFs.
How it works
From spreadsheet to evidence in four moves.
Import
Any CSV or Excel. Types are inferred, quality scored, and PII automatically flagged.
Explore & analyze
Descriptive stats, correlations, distributions, and group comparisons — instantly.
Predict
Train and evaluate models, tune thresholds live, and rank features by importance.
Publish
Export reports, white papers, and grant-ready evidence narratives.
Methods engine
Transparent enough for the analyst. Clear enough for the stakeholder.
RSCHR treats methodology as a first-class product surface. Every chart, model, and export carries the assumptions, thresholds, and validation context needed to defend the result later.
Assumption checks
Missingness, type inference, outliers, group balance, and distribution shape are surfaced before conclusions are drawn.
Effect sizes beside p-values
Statistical significance is paired with practical significance, confidence intervals, and reviewer-ready language.
Model decisions recorded
Thresholds, feature sets, algorithm comparisons, and run history stay attached to the project for reproducibility.
0%
reference parity
0
cloud uploads
macOS 0
minimum version
Inside the workbench
The moments that prove it's a real tool.
Correlation matrix you can defend.
Pearson, Spearman, and Kendall side by side, each cell carrying its Fisher-z confidence interval — so a reviewer sees not just the coefficient but how certain you are of it.
- Five coefficient families, including point-biserial and phi.
- Significance flags and effect-size shading built in.
- Export the full matrix straight into a report.
Correlation · Pearson r
Tune the threshold. Watch the trade-off.
Four algorithms overlaid on one ROC curve. Drag the decision threshold and precision, recall, and accuracy recompute live — the same trade-off conversation you'd have in a model review, made tangible.
- Logistic regression, random forest, boosted trees, and SVM.
- Confusion matrix and F1 update with every move.
- Saved run history so you can compare experiments.
Feature simulation · ROC
Precision
0.88
Recall
0.81
F1
0.84
AutoResearch finds the signal for you.
An evolutionary engine searches your dataset for the strongest predictive relationships, tracking every experiment across generations so you can see exactly how a finding emerged.
- Automated feature and interaction search.
- Full experiment lineage, kept on-device.
- Promote any candidate into a saved model.
AutoResearch · experiment log
What you ship
Analysis becomes a package people can read, review, and fund.
The app is designed around the handoff: not just producing numbers, but turning them into evidence artifacts that survive scrutiny.
Executive narrative
Findings, methodology, limitations, and implications arranged for a grant reviewer or board packet.
Run evidence
Metrics, thresholds, feature importance, and validation notes exported with the analysis instead of recreated later.
Defensible record
Project decisions stay together, so the route from dataset to conclusion remains visible.
- Import quality checked
- Assumptions reviewed
- Model threshold saved
Privacy & security
Your data stays on your desk.
There is no cloud step to opt out of, because there is no cloud step. Every calculation and every model is computed locally — the strongest privacy guarantee is the one that needs no policy.
On-device by default
Import, statistics, and CreateML training all run locally on Apple Silicon.
No cloud processor
No third-party data processor sits between you and your analysis. Ever.
No account required
Download, open, and start analyzing. No sign-up, no telemetry to opt out of.
Notarized & signed
Distributed as a notarized, signed build — safe to install and approved by Gatekeeper.
Who it's for
Built for people who have to prove it.
Institutional research
Defensible, reproducible analysis for grants and board reports at colleges and agencies.
Program evaluation
Impact studies for afterschool, workforce, and public-health programs — with real statistics behind the outcomes.
Clinical & social science
Research where data privacy and IRB compliance are non-negotiable from day one.
Grants & development
Evidence narratives and white papers, fast — turning analysis into fundable stories.
Questions
Everything you'd ask before installing.
Yes. Import, every statistical computation, and all model training happen locally on your Mac. There is no cloud upload requirement and no third-party data processor — your data never leaves the device.
No. RSCHR guides you from import through analysis to publication with sensible defaults, plain-language readouts, and research-method templates — while still exposing the rigorous methods a statistician would expect.
CSV and Excel files. On import, column types are inferred, data quality is scored, and potential PII is flagged automatically.
No. RSCHR is a native Mac app that runs fully offline. Once downloaded, no connection is required to import, analyze, model, or publish.
An Apple Silicon Mac running macOS 14 or later. There's no Python runtime or other dependency to install — it's a single native app.
Yes. The build is signed and notarized by Apple, so macOS Gatekeeper recognizes it as a trusted developer application.
RSCHR is free to download and use.