Mahalo Workbench

Analyze study data without moving it.

Workbench is a secure analysis environment built into Mahalo Research. Dashboards for every researcher, notebooks for your analysts, plain-language answers for everyone — and nothing leaves without review. The export habit ends here.

Mahalo Workbench overview showing governed data views, dashboards, notebooks and egress status

01 · The problem

The export is where governance ends.

Modern studies capture data under consent, audit and access control — then export it to analyze it. Every CSV on a laptop is an ungoverned copy: unversioned, unauditable, stale by morning. Workbench brings analysis to the data instead, so findings carry the same chain of custody as the record they came from.

02 · Governed views

One source of truth, with a paper trail.

All analysis reads named, versioned, approved views — never raw tables. Each view publishes its data dictionary and its lineage: operational record, pseudonymisation, governed store, view. Role-based row and column access resolves at query time, and any aggregate below the study's threshold renders as "<10", everywhere, automatically.

  • Versioned view definitions, released like software
  • Lineage visible on every view
  • Small-cell suppression enforced in the product
Mahalo Workbench data views showing versioned governed datasets, column definitions, and lineage

03 · Dashboards & cohorts

Answers without code.

Researchers and clinicians explore the study in dashboards and a point-and-click cohort explorer — recruitment, retention, completion, data quality, questionnaire scores, wearable trends. Cohorts are saved as versioned filter definitions, not lists of people. There is no download button; the export action opens an egress request.

  • Operational and scientific dashboards on live views
  • Cohort counts that respect suppression
  • Filters recompute live, per site, wave or period
Mahalo Workbench wearable trends dashboard showing governed study charts
Mahalo Workbench cohort explorer showing suppressed, role-aware participant filters

04 · Analyst workspaces

Real tools for your analysts.

JupyterLab and R workspaces run inside the boundary with views mounted read-only, packages pinned from an internal mirror, and internal git. Regression, longitudinal models, signal processing and machine learning happen on live governed data — not on last month's extract. Workspaces have no general internet egress, because an open outbound connection is an unreviewed exit.

  • Per-study workspaces, right-sized compute
  • Pinned environments, reproducible by design
  • Outputs leave only through egress review
Mahalo Workbench notebooks showing JupyterLab and R environments inside a governed study boundary

05 · Governed answers

Ask in plain language. Verify in one click.

The Workbench assistant answers questions about study data with the sources named: every answer cites the views and versions it read and shows the generated query, so a statistician can reproduce it in a notebook. It sees only what the asker may see, applies suppression to its answers, and runs on a pinned, change-governed model version. Every interaction is logged.

  • Answers cite views and versions
  • Generated query shown for reproducibility
  • Descriptive answers only — never clinical advice
Mahalo Workbench AI assistant showing a descriptive answer, cited governed views, and generated query

06 · Reviewed egress

Nothing leaves unreviewed.

There are two exits. Analysts request egress; automated checks — identifier scan, row-level detection, smallest-cell computation — run first, then a reviewer approves release to a named destination with an expiring link. Scheduled deliveries ship governed views to institutional trusted research environments with manifests the receiving side verifies. Rejections travel back with the reviewer's note, so the control teaches instead of just refusing.

  • Automated checks before human review
  • Separation of duties: producers can't approve
  • Every decision an audit record
Mahalo Workbench egress review showing automated disclosure checks and reviewer decision controls
Mahalo Workbench trusted research environment exports showing governed scheduled deliveries

07 · Reproducible reporting

Findings that stay reproducible.

The report builder pins every chart to the view version and as-of date it was computed from; reopening a report recomputes nothing silently. Scheduled reports deliver recurring snapshots through templates egress review approved once — each delivery logged with its data timestamp and recipients.

Mahalo Workbench report builder showing charts pinned to governed view versions
Mahalo Workbench scheduled reports showing approved-template deliveries

08 · Portfolio analysis

One workbench across your portfolio.

Every study you can access, one place — with your role resolved per study. Data never mixes between studies; cross-study analysis requires a governed multi-study view approved by each study's governance. Deployed in-region where residency law requires it.

Mahalo Workbench all studies view showing a governed research portfolio

Featured whitepaper

Analysis without the export

Why study-data governance breaks at analysis — and how governed views, notebooks, AI answers, and reviewed egress close the gap.

Get the whitepaper

Workbench inherits the platform's audit, access and security model — review it in the Trust Center.

See your study in Workbench.

Bring one study — its questionnaires, its wearables, its team — and we'll walk it through the workbench end to end: views, dashboards, a notebook, and an egress request.

Frequently asked questions

What is Mahalo Workbench?
Mahalo Workbench is a secure analysis environment built into Mahalo Research. Researchers analyze study data through governed views using dashboards, a cohort explorer, JupyterLab and R notebooks, and a governed AI assistant, without exporting data. Outputs leave only through reviewed egress or scheduled delivery to institutional trusted research environments.
Is Workbench a trusted research environment (TRE)?
Workbench applies TRE-style controls — pseudonymised data, governed views, small-cell suppression, egress review, full audit — and integrates with institutional TREs through scheduled, manifest-verified deliveries. Institutions keep their own TRE; Workbench feeds it on governed terms.
Who can see identifiable participant data in Workbench?
No one. Identifiers are removed by a pseudonymisation checkpoint before data enters Workbench, and the environment holds no re-identification key. Identifiable data stays in Mahalo Research under its own permissions.
Do analysts need to write code?
No. Dashboards and the cohort explorer cover descriptive analysis without Python, R or SQL. Analysts who do write code get JupyterLab and R workspaces inside the same boundary.
How does data get out of Workbench?
Through two audited routes: analyst egress requests, which pass automated checks and human review before release via an expiring link, and scheduled governed deliveries to institutional trusted research environments. There is no direct download.