Project 03 — 2023 / 2026
Liftoff
Rebuilding the creative uploader for a mobile-advertising platform — reworking the product’s most-avoided surface by testing patterns, not defending one.
Liftoff runs advertising for thousands of mobile apps — a product merged from many ad-tech tools over the years. Across the Vungle lineage I’ve designed campaign tooling, advertiser and publisher surfaces, and the design system beneath them. The one this case study digs into is the creative uploader — where advertisers hand over the videos and images that become ads, and the most-avoided surface in the product. Rather than commit to a single redesign, I built three working prototypes — free-form, wizard, and canvas — and put the patterns in front of real users to see which one earns it.
Three clickable prototypes — free-form, wizard, and canvas — built fast with AI, so we could test the patterns with real users instead of debating mockups.
Each flow as a sequence, not a screenshot — how the pattern moves from list to modal to done. Four flows across the creative tools.
Campaign Attach
Attach creatives to campaigns in bulk — with live creative counts and conflict warnings.
Campaign Detach
Pull creatives off campaigns with the same table-to-modal-to-confirm pattern.
Endcard Selection
Pick VAST assets and assign endcards while approving a batch of creatives.
Concept Grouping
Group assets into concepts in the Asset Library for cleaner campaign reporting.
A subset of the component system I specified and maintained across the creative tools.
Liftoff is a patchwork — ad-tech tools from Vungle, Algolift, GameRefinery, Jetfuel and more, built by different teams and merged into one platform. I’ve designed across several of those systems at different points in that trajectory: the advertiser platform and campaign tooling shown here, the uploader above, on a design system that still threads through the product today.
The problem
Uploading creative meant wrestling a free-form tool with two asset libraries, unclear format requirements, and no feedback until something failed. Advertisers avoided it — which meant fewer creatives tested and less spend.
The approach
I mapped the real workflows — starting with the vertical-video path that drives 44% of revenue — then built three prototypes around one question: how much structure does a reluctant uploader actually want?
Across all three, the same bets stay constant: you add your sources once and the system derives every required size, plain-English guidance sits inline, and a live preview shows exactly what will run before submitting — so the test isolates the pattern, not the polish.
“The best tool for a reluctant user is one that quietly makes them look like an expert.”