Work completed under a non-disclosure agreement with Procore Technologies. Internal metrics, stakeholder details, and product data have been modified to protect confidentiality. Design decisions, process, and outcomes shown reflect my own contribution.
"I have to pull from three places every Monday morning. It takes forever just to get a picture of the week."
Safety Director, usability research
One dashboard that replaced three fragmented tools
Safety directors were spending 2 to 3 hours a week manually pulling data from three separate Procore tools. Safety Hub consolidated Observations, Inspections, and Incidents into a single card-based analytics dashboard.
Worked with Procore's design lead, PM, and engineering team across research, definition, and handoff.
Research
Sessions were facilitated by Procore's UX research team. My role: observe sessions, participate in synthesis, and translate findings into design direction. The three directions below came directly from that process.
What we heard
"I don't look at what's newest. I look at what's most dangerous."
Safety Director, Session 3
Directors scan for risk clusters, not recency. A week-old observation from a high-risk project outweighs a today-logged item from a low-risk one.
Card hierarchy must surface risk severity first, not date. This ruled out any direction built on chronological sorting.
"I built my own spreadsheet because Procore couldn't give me one view."
Safety Director, Session 1
The Excel workaround was widespread. Directors were already doing the aggregation manually, which meant the dashboard had to fully replace that ritual, not just improve it.
The solution must be a complete replacement for the Monday spreadsheet, not a supplement to existing tools.
"I only care about my projects. Everything else is noise."
Superintendent, Session 6
Superintendents and Directors are looking at the same data through completely different lenses. One view cannot serve both without failing both.
A Company/Project toggle is not a nice-to-have: it is the core architecture decision the product depends on.
What the research ruled out
Aggregate Observations, Inspections, and Incidents into one scrollable list sorted by most recent activity.
Still chronological. Same mental model problem in a single tool instead of three.
Show a summary tile for each tool: Observations count, Inspections count, Incidents count, with links to drill down into each.
Still siloed. No cross-signal between tools. Directors still have to synthesize manually.
Cards organized by urgency and risk pattern, not date. Cross-tool signals surfaced. Company and Project views with the same card system at different scopes.
Matches how directors think. Risk-first, pattern-aware, scope-aware. Confirmed by sessions 1, 3, 5, and 7.
Information Architecture
Before any visual design started, four structural decisions shaped the dashboard. These were not aesthetic choices. They were the calls that made the product work or fail for the people using it.
Company Dashboard
Six projects, three modules, half a Monday: that was the cost of weekly safety review before this existed. Every card was designed around a specific question the Director asks at the start of the week. The order, the ranking logic, the urgency coding: all of it answers that question before the Director has to look for it.
Project Dashboard
Marcus checks this before the toolbox talk. He needs to know who hasn't closed their corrective actions, what's on today's inspection schedule, and whether his site is trending safe. The cards here answer superintendent questions, not director ones: accountability by person, not by project. The Activity Feed closed a gap competitive review had flagged: dedicated field safety apps had established the 24h activity log as a standard for shift handoff.
Mobile Design
Mobile entered scope after the desktop dashboards reached handoff. Superintendent feedback made the case: field users check safety status on-site, between conversations, not at a desk. The mobile views were designed in a second sprint. They are not a port of the desktop: every card and navigation pattern was rethought for field context.
Company Dashboard
Project Dashboard
Scope toggle: primary navigation
No sidebar on mobile: the toggle is the only scope control. Switching view replaces every card with content for that persona.
2x2 KPI grid
The desktop KPI row folds to a grid so all four signals stay readable at arm's length without horizontal scroll.
Projects ranked by risk share
Same ranking logic as desktop: the site with the highest observation share appears first, not the one filed most recently.
"See all" defers depth
Each card shows 3 to 4 critical items. Everything else is one tap away, keeping the most urgent signal visible without scrolling.
Context band replaces the sidebar
The orange band anchors project scope persistently: the job the sidebar does on desktop, without consuming horizontal space.
Assignee queue: pre-toolbox check
Who hasn't closed their corrective actions, sorted by overdue first. The superintendent sees accountability problems before the crew arrives.
Activity feed: shift handoff log
A timestamped 24h log replaces the verbal briefing. The superintendent arrives knowing what happened on the previous shift.
Inspection schedule: this week only
Narrowed from the full calendar to this week. Completed items show green, overdue show red. No scanning required.
From desktop to mobile: what changed and why
Desktop sidebar nav
Bottom nav + context band
Sidebar takes too much horizontal space. Bottom nav handles modules; context band handles scope.
Horizontal KPI row
2x2 KPI grid
Four cards side by side at phone width forces tiny numbers or horizontal scroll. Grid keeps all four readable.
Full card rows
3 to 4 items + "See all"
Showing everything on first load creates a wall on mobile. The most urgent item stays visible without scrolling.
Back nav + toggle
Toggle only
Both controlled the same scope switch. Keeping both created redundant affordances. The toggle alone is enough.
Design Decisions
The risk with a two-view dashboard is that both views become reskins of each other: same layout, same hierarchy, different data labels. Four decisions prevented that and made each view genuinely distinct for the persona using it.
Impact & Outcomes
Procore launched Safety Hub in open beta April 2026, aligned with the two-view framework this design established. The core IA decisions held through production.
The ~40% figure reflects self-reported estimates from beta director feedback, not a published or internally validated Procore metric.
What worked
- Grounding every card in a named persona meant every design review pushback had a research-backed answer
- Separating the two views structurally forced genuine differentiation, not just a data swap
- Competitive review surfaced the Activity Feed gap early, giving time to design it properly
What I'd do differently
- Scope mobile for the Project View from day one: superintendents check this on-site, not at a desk
- Include a usability test round before handoff to validate the two-view toggle with real users
- Push for a shared corrective actions module earlier: it was the gap both personas flagged but got deprioritized
