Changelog

Various bug fixes, inclusion of .xdsl file type usability, and the ability to save model files from the Workspace tab.

Behind-the-scenes efficiency pass on the analytical engine: faster rebuilds when repeated requests hit the same model, a methodologically tighter rule for building the model structure, and a leaner deployment with more memory headroom. No visible change to the workspace, the Calibration tab, or your reports.

Power-user controls for the CSV upload flow. Advanced settings now let you choose how the model structure is built from your data, from the recommended default to several specialised options. Outcomes can now influence each other in the model, and the home page's returning-user strip gained the same Recommended / Advanced controls as the onboarding flow.

Quality and accuracy pass. Your data now has more room to depart from the research-derived starting values (small and medium samples were previously over-anchored), the large-sample CSV path can optionally use a research-informed framework as a soft anchor, and the Calibration tab gained a clustering-column picker for CSVs with multiple rows per team or site. Under the hood, the engine moved to its latest core library and a new regression suite locks in the behaviour.

Adds credible intervals across the Workspace so every estimate now carries a confidence range, a new Calibration tab and Model Fit panel that show how well the model fits your data, sharper rankings that reflect what would happen if you actually moved a factor rather than just observed it, confidence-tier framing in the generated report, and a new Training Course section in the sidebar with Beginner and Advanced tracks of eight chapters each.

Full launch after the soft-launch period. Adds the experimental Budget Lab, a clearer Diagnostics 'From → to' view of where each factor's probability is moving, and a 'For investigators' paragraph in both reports naming the top factors to check when a fresh signal appears. An example generated report is now linked from the Report page.

Pre-launch positioning pass. The headline now leads with the differentiator - showing how psychosocial hazards interact - instead of the price point, with 'free' demoted from a primary value chip to a quieter trust signal alongside browser-only and no-account. The 'Why is this free?' FAQ has been reframed around continuing the legacy of People at Work and lifting the bar further; a new FAQ entry explains where Prior Work fits relative to the platform's June 2026 retirement. A 'With thanks' section on the About page acknowledges the original People at Work academic team and the soft-launch reviewers. Page meta descriptions rewritten across the site so crawlers and academic searchers can find the technical and methodological vocabulary that body copy intentionally keeps out of view.

Workspace rebuilt as three opinionated tabs - Overview shows how the workforce is sitting on each factor, Scenarios surfaces the workplace factors that keep showing up across the strongest combinations, and Diagnostics says where to look first when an outcome has shifted adversely. The main report tightened from seven sections to five (with the prioritisation matrix folded into a new Where to Focus section and the report-mode toggle removed in favour of a single auto-detected path), and the pulse report from six sections to four. AI prompts rewritten with lighter scaffolding - example pairs replacing rule lists - which produced noticeably sharper section prose in test runs without losing the consulting voice.

Process streamlined from 9 steps to 7 - the Interventions Library now lives inside Action Planning, and Pulse Survey and Monitor & Evaluate have merged into a single Track & Monitor page. The Introduction page has been refocused on why Prior Work exists and the architecture behind the measurement set, with the full 20-construct detail moved to a new Glossary page. Model-fitting copy has been updated to reflect the current work stress research foundation.

Polish release - pulse reports now expose a download button for the companion consultation guide, the Licence / FAQ / Changelog / Privacy pages have been widened to match the rest of the site, and frontend dependencies have been patched to clear five security advisories.

Pulse data now updates the whole model - recording new evidence on the factors you measured shifts the connected factors too, not just those directly measured. Baseline and pulse reports now share a single consultant voice grounded in current work-stress research and the SWA Code of Practice. Reports moved to Claude Opus 4.7 with a cost estimate shown before generation.

Report architecture overhaul - both the baseline and pulse reports now draw on the full model, recommend an intervention set, and generate a companion consultation guide automatically. The Workspace merges the Where to Focus and What-if views into a single Explore & Focus tab with dominant-state highlighting.

Soft-launch review pass covering Interventions, Pulse Survey, Pulse Report, and About. Jurisdictions removed from the interventions library; pulse reporting now distinguishes chronic vs emerging hazards; About page updated with author bio and self-serve positioning.

Small-sample inference fix and site-wide content review covering legal accuracy, comms strategy, survey page, and home page framing.

Survey instruments released; demographic handling added to CSV upload; Option C redesigned for instrument flexibility; Glossary, FAQ, Privacy, and Changelog pages added.

Pulse survey and pulse report released; sidebar and nav footer consistency fixes; ISO 45003 hazard list expanded on the Survey page.

Initial soft-launch build - 10-step process scaffold, data-model workspace, Report page with AI narrative generation, and sector benchmark models.