Nine Entertainment
2025-2026
Reimagining TV Ads Scheduling at Nine
Nine's traffic schedulers decide which ads run in every commercial break across its TV network. I led the design of a new Traffic Application that turns legacy systems into one single application that improves efficiency and usability while creating the design system for multiple applications to scale and future proof.
Overview
Two legacy systems into one, built on a shared design system
Schedulers were splitting their work across two legacy systems (Saturn & Phoenix), each with its own data, layouts and workarounds. As design lead on Nine's platform rebuild, I designed TMS to bring that work into one application. Alongside it, I started and led a design system that now powers TMS and a second new product, so both look and behave the same.
Problem
Two systems, dense screens, expert users
Split workflows: schedulers moved between two systems to get one job done, using workarounds such as email or slack messages between teams or build their own excel tools.
Dense screens: the legacy schedule was one table with no hierarchy, so checking a single break meant scanning row after row.
High stakes: every spot has to fit its break, land inside the advertiser's requested window and follow placement rules.
No shared foundation: each new product would have built its own components, duplicating dev work and giving people who use both tools two different experiences.
Research
Designing with the traffic team, not for them
I ran workshops and feedback sessions with the traffic team to see how they scan, compare and fix schedules across both systems. What shaped the design:
Breaks first: schedulers check duration and fill status before anything else.
Detail on demand: client, product and campaign matter once a break needs attention.
Rules in context: request windows need to sit next to the spot, not in a separate lookup.
Side by side: comparing schedules is core to the job.
Design
One application, one design system
Consolidating two systems. I mapped workflows across both legacy systems and designed TMS around the tasks schedulers actually do, rather than recreating each old screen.
The Traffic Schedule. The core view shows each break's spots, duration and fill status at a glance, with client, product, campaign and request windows alongside. After the first showcase, scope grew from 7 to 20 columns. A shift in priority to WIN schedules then meant fitting seven schedules on one screen instead of two, so I reworked the layout to stay scannable at that density.

Dashboards. Driven by user workflows and tested extensively to only display what is really needed. One of the main considerations was to make sure to reduce screen clutter and information and only to display what is really needed in a particular moment. Using concepts such as progressive disclosure helped to reduce noise and keep the user focused.

Exploring fast with AI. The schedule started in early discovery as "Sequencing and Smoothing". I fed workshop outputs into AI via a PRD to generate early concepts, then reworked them by hand: AI framed the idea, but couldn't handle the density and rules of an expert tool.
A design system for more than one product. I initiated and built a shared design system and Figma library for TMS. Working closely with the designer of a second new product (a sales product), I led the effort to make it the single source for both.



Design-to-code handover. I set up handover for AI-assisted development using Figma Code Connect, design tokens and the Figma MCP server, so developers' AI tools build from real components rather than screenshots.
Impact
One system to learn, one source to build from
Two legacy systems consolidated into one application.
One design system shared by two new products: developers build from a single source, and people who use both products get a familiar experience.
Seven schedules visible on one screen, up from two.
Validated with the traffic team through workshops and feedback rounds.
Outcome
What I'd keep, and what I'd change
Starting the design system early paid off: when a second product came along, it built on what already existed instead of starting from scratch. The hardest part was scope. Concepts went to stakeholders, then a wave of new requirements caused rework. Next time I'd lock requirements earlier, share low-fi prototypes sooner and get feedback before polishing. AI sped up exploration, but high-fidelity design for expert users still needs a designer's judgement.

