Attempt A
Sync Visual QA
- Live screen-share: Inspect Element ↔ Figma
- High integrity within one feature story
- Related fixes shipped together
- Broke developer deep work
- Many finds = DS consistency — low eng priority
Visual QA without the ticket war
How we redesigned remote Visual QA — not as a checklist, but as a negotiation system between design, QA, engineering, and product.
Role
Senior Designer · process owner
Context
Vodafone · remote · design system
Partners
Design · QA · Eng · PO
Outcome
Hybrid QA pipeline adopted
Situation
«Done» in Jira was not the same as looking like the design.
On a remote product team, Visual QA couldn't mean sitting next to a developer. Design-system consistency across markets mattered — but quality depended on who cared that week. We needed a process that protected product integrity without burning design capacity or developer flow.
"This wasn't a Figma problem. It was a negotiation problem."
Two experiments
Both extremes were half-right.
We tried two approaches. Each solved one side of the conflict and broke the other.
Attempt B
Async ticket factory
- Fit eng async culture
- Batchable, prioritizable work
- Design drowned writing tickets
- Priority killed «small» consistency fixes
- Surface integrity fragmented
Sync Visual QA
Live screen-share: high integrity per feature, but broke developer deep work and couldn’t scale beyond one story.
Async ticket factory
Fit eng culture, but design drowned writing tickets. Priority killed consistency fixes. Surface integrity fragmented.
Keep what worked. Gate what’s expensive.
- QA absorbs coarse visual bugs via existing ticket habit
- Short designer–QA sync replaces long dev sessions
- AI drafts residual tickets — designers own severity
- Live dev QA becomes PO-gated premium mode
Two extremes. One hybrid that kept the best of each.
Hidden actor
QA engineers already ran functional testing — and often caught coarse visual bugs the same way they file functional defects. Developers were used to that ticket stream. The mountain of tickets wasn't only a design burden — it was a habit we could redirect.
The conflict
Three valid goals. Zero pure winners.
Design wanted coherent surfaces and design-system fidelity. Engineering wanted protected focus and clear batchable work. QA needed coverage without inventing a new religion.
We sat down with QA, developers, and the Product Owner — and redesigned the sequence, not the tool.
Integrity
Coherent surface per story. Design-system fidelity.
Flow
Protected deep work. Clear, batchable tickets.
Coverage
Functional pass + coarse visual — without a new ritual.
Redirect habits. Gate expensive modes.
- QA ticket habit absorbs coarse visual bugs
- Short designer–QA sync for judgment
- Designer tickets only for residual fine issues
- Live designer–dev QA only when PO flags consistency
Three forces. One negotiated pipeline.
The hybrid
A pipeline of filters — not one ritual for everything.
Default mode is lean and async. Live Visual QA with developers becomes a premium mode — justified by product, not designer preference.
01
Functional QA pass
QA territory — including gross visual issues. Tickets in the format engineering already trusts.
02
Short sync: Designer ↔ QA
Triage what's left. Cheaper than pulling every developer into a live review.
03
Residual designer tickets
Only fine-grained design-system issues. AI drafts the tickets — designers still own severity and accuracy.
04
Live Visual QA — PO-gated
Only when the Product Owner marks the story: must look right for product consistency. Same story, related fixes, justified interruption.
Mental model
QA ticket habit = first filter · Designer–QA sync = second filter · Designer tickets = residual · Dev live sync = premium mode, PO-gated · AI = writing acceleration, not judgment.
Result
Less interrupt. Less ticket hell. Surfaces that mattered stayed coherent.
QA first filter
Coarse visual absorbed by existing ticket habit
PO-gated sync
Live designer–dev sessions became the exception
Design ticket volume dropped because QA absorbed the coarse layer. Engineering interruption dropped because live sessions became rare and justified. Quality didn't collapse into «QA only» — designers still owned fine design-system judgment.
The process became shared know-how: design, QA, engineering, and product agreed on the sequence — not a design decree.
What I learned
Process design is product design for the organisation.
- 01Don't invent rituals — redirect habits partners already trust.
- 02«Perfect QA» that destroys partner flow fails. Integrity without dignity is temporary.
- 03Put expensive modes behind a product gate — the PO, not designer preference.
- 04Sequence matters: functional → coarse visual → fine visual.
- 05AI accelerates writing residual tickets. Humans still decide what matters.