← All work

02 · Enterprise SaaS · Engineering

Homepage Dashboard

Client
OpenLM
Role
Lead Product Designer
Year
2024
Tags
Dashboard · Widgets · Customization · B2B
[ The brief ]

A useful homepage.

After the platform redesign shipped, the new homepage launched with two admin controls and a handful of documentation links. Customers used it as a shortcut page on their way somewhere else — and that was it.

The brief was simple to state and harder to deliver: make the homepage actually useful. The answer turned out to be a configurable dashboard built on top of a curated widget catalog, with templates for the most common roles.

Spoiler alert — D30 retention almost doubled. Here's how.
[ Challenges ]

Two hard problems.

01

Building the customization engine

Drag-and-drop grid, resize, per-user persistence, and data plumbing across multiple internal services. The engine had to stay performant on huge environments and forgive any single widget that failed to load.

02

Picking the v1 catalog without a bloat tax

Twenty-seven widget candidates, room for maybe a third. We built a prioritisation gate — ticket volume × interview overlap × engineering feasibility — and let widgets that didn't clear all three slip to the next release.

[ Approach ]

Two tracks, one ship.

We inherited a working design system and modular services from the platform redesign, so the budget could go where the actual risk was: the customization engine and the widget designs themselves.

I split the work into two parallel tracks — a discovery track scoring widget candidates with admins, license managers and IT, and a technical track validating the grid engine with a thin-slice prototype. They converged into one roadmap once we knew which widgets were both wanted and buildable.

Role templates joined as a third track late in the process — a pre-built dashboard per persona, so first-run wasn't an empty canvas.

Roadmap — discovery through design & tests
[ Discovery ]

Tickets + interviews.

What we wanted
  • Make the homepage return-worthy on its own — not a corridor people walk through to reach a real screen.
  • Give each role a useful default the first time they land, with zero configuration.
  • Make customization a real composition tool, not a settings panel disguised as a feature.
How we found out
  • Eighteen interviews across four personas — admins, license managers, IT managers, execs — using the research pool we'd built for the platform redesign.
  • Mined a year of support tickets for the dashboards and signals customers had been asking for in writing.
  • Scored every candidate widget on three axes — ticket volume, interview overlap, and engineering feasibility — before anything entered the v1 list.
71%
Of sessions bounced off the old homepage in under 4s
120+
Support tickets mined for widget signal
9 / 27
Widgets shipped in v1 out of candidates scored
[ Research insights ]

Signals we couldn't unsee.

  • The old homepage had a bounce-equivalent: 71% of sessions left within 4 seconds, straight into the side nav.
  • Only the widgets that surfaced in BOTH support tickets AND interviews survived prioritisation — single-channel signal turned out to be noise.
  • License managers wanted live usage; IT managers wanted compliance status; admins wanted system health. One static layout could not serve all three — that's the whole reason this had to be a dashboard.
Early concept — static shortcuts and tutorials, before the dashboard model won
[ The prioritization gate ]

Feasibility decided v1.

  • Technical feasibility was the deciding criterion for v1 — ship widgets we could build well, then expand.
  • Every widget had to degrade gracefully when its underlying service was slow or unavailable.
  • Layouts had to persist per user and survive across devices without a heavy backend rewrite.
[ Design — old vs new ]

Before, after.

Homepage

Before → After

From a legacy dashboard of raw tables and static charts into a configurable widget surface with role-aware defaults.

Before
After

Concept → Shipped

Before → After

The first concept treated the homepage as a shortcut page — quick links, tutorials, blog. Testing killed it. The shipped version is a real dashboard: KPIs at the top, editable widget grid below, role templates on the side.

Before
After
[ Gallery ]

In context.

[ Impact & conclusion ]
“D30 retention 34% → 58%.”
  • D30 retention 34% → 58% after rollout to the existing customer base.
  • Role templates covered ~80% of new users on day one — the empty-grid problem disappeared.
  • Avg widgets per user reached 6.4 within 30 days — customization is being used, not just available.
  • Homepage moved from the 11th to the 2nd most-visited screen in the platform.

Defaults without customization age fast — power users outgrow them in a week. Customization without defaults dies on first run — most users won't configure anything. The actual win wasn't either side: it was the curated catalog and the role templates that bridged them.