Supervisor Desktop — Webex Contact Center
Contact-centre supervisors had no product of their own. They did the job across three tools built for other people. This is the surface that replaced them — and the process change that made shipping it possible.
One surface, assembled by the person using it. Contacts in progress and in queue, longest current wait, handled and abandoned volumes, service level, average handle time and CSAT — each broken out by voice, chat, email and social — above a live queue table and an agent-state report. The tiles are modular and rearrangeable, so a supervisor's home reflects the floor they actually run rather than a layout someone else fixed for them.
tools replaced by one surface built for the supervisor's actual job.
how far ahead of development design ended up running, one slice at a time.
design phases — high-level and detailed — each with explicit deliverables and exclusions.
interval data that took hours to compile, available instantly from the home page.
- Admin — configuration and permissions
- Agent Desktop — switch into an agent role to monitor or assist
- Analyzer — analytics and historical reporting
-
Name the persona the product forgot
Supervisors stitched the job across Admin, the Agent Desktop and Analyzer — three tools, none designed around their workflow. Competitors already shipped a dedicated supervisor desktop. The gap wasn't a missing feature; it was a missing product.
-
Scope by frequency × criticality
The long-term vision was one role-based product. Chasing it immediately meant inheriting three tools' legacy tech and data constraints wholesale. We scoped instead to what was both high-frequency and high-criticality for supervisors: managing agent teams, queue oversight, controlling agent access.
-
Decide what not to absorb
Analytics mattered daily, but we deliberately did not merge Analyzer in. It served Analysts — a genuinely different persona. The supervisor home surfaces Analyzer data directly instead, configurable by the supervisor. Merging in the name of “one interface” would have blurred a real persona boundary rather than simplified anything.
-
Fix the review, not the reviewer
Triad reviews ran long and dense — edge-case error-message questions arriving during high-level concept reviews, review volume mixing real feedback with premature scope. Rapport-building and documented action items helped but didn't hold; escalation got attention without stopping recurrence. The durable fix had to be structural.
-
A clarifications step, owned by design
We put a “Clarifications on Requirements” step ahead of design work — protected space for designers to surface open questions and hand them back to PM and engineering as their items to close. Unscoped “have you thought about X?” became inputs the triad had to resolve before design started, not during it — and the mechanism for slicing the product into features.
-
Two phases, with explicit exclusions
High-level design stayed locked to happy-path concept validation; error states, edge cases and accessibility detail were deferred by design into detailed design. That gave PM and engineering a legitimate, scheduled place to raise those questions instead of raising them everywhere, always. Research checkpoints were sequenced into the same cadence.
The review friction wasn't the PM's behaviour — it was that we hadn't yet learned to scope a net-new product against real capacity. Milestones were set before anyone sized the work, and effort was reverse-engineered to fit them.What I'd do differently
The supervisor desktop shipped and closed the competitive gap. Reviewers described it as a single view replacing three applications — monitoring, historical metrics and team messaging in one responsive layout built from modular, drag-and-drop widgets. Interval data that had taken hours to compile in the legacy system became instantly available from the home page, and enterprises migrating off a fragmented stack reported smooth transitions with strong adoption and specific praise for the new visibility into queue and agent activity. Operationally, design moved from reactive to a full quarter ahead of development.
Don't merge what serves a different persona
Analyzer stayed separate because Analysts aren't supervisors. Consolidation is only simplification when the things being consolidated belong to the same job — otherwise “one interface” just hides a boundary that still exists.
Give the hard questions a scheduled home
Error states and edge cases weren't excluded from the work — they were deferred to a phase that owned them. Reviewers stop raising a question everywhere once there's a place it is guaranteed to be heard.
Slice first, then run ahead
With clarifications closed upfront, work broke into slices: design takes the prioritised one, hands it to engineering, moves to the next. That compounding is what bought the quarter of runway — not working faster.