Google Creative Coverage
I designed AdRoom’s side of the Pixis × Google co-development initiative, now live as Google Creative Coverage, and led my team to ship it to production.
- Client
- Pixis, with Google
- My role
- Technical lead and solution architect
- Team
- My AdRoom team of 6 engineers
- Period
- Jun – Aug 2026
Results
Prototypeto Live
The 5-step flow I prototyped in a week shaped what is now live in AdRoomOct 2026
1 week
To build the clickable prototype that aligned Google and PixisJul 2026
< 2 months
From prototype to production, behind feature flagsJul → Aug
Context
Pixis is an enterprise AI marketing platform, and AdRoom is its AI ad-creative studio. I led AdRoom’s engineering.
In 2026, Pixis ran a technology co-development initiative with Google to help advertisers get the most out of Google’s Demand Gen and YouTube campaigns. What came out of it is Google Creative Coverage, now live in AdRoom. I was technical lead and solution architect on the AdRoom side.
My role
I designed the solution, then led my team to build it.
- Plan. I wrote the product and engineering plan: the user journey, an “exists vs. build” matrix against AdRoom’s existing pipelines, and a feasibility map with a risk rating for each deliverable.
- Specs. I wrote the technical specs and data contracts for each deliverable and for the Google Ads asset integration, plus the credit-pricing model for the new AI pipelines.
- Prototype. In a week, I built a clickable prototype of the 5-step flow shown below: bring assets, coverage check, coverage plan, fill, review. It aligned Google and Pixis stakeholders before implementation started, and the live product follows the same flow.
- Delivery. I led my AdRoom team of 6 engineers through the build, reviewed and merged the work, and presented progress to Google and Pixis stakeholders.
The problem
Demand Gen and YouTube campaigns perform best with a full set of creative: many images across several aspect ratios, multiple videos and optimised text. Most advertisers bring far less, often a few images made for other platforms and no video at all.
Closing that gap by hand is slow and expensive. The initiative set out to measure each advertiser’s creative coverage and fill the gaps with AI, without letting quality slip.
What I built
Google Creative Coverage checks an advertiser’s creative against Google’s requirements for Demand Gen and YouTube Video Reach, plans the gaps, then fills them, with a person approving every asset before it goes out.
- 01Bring assets
- Upload
- Import from Google Ads, YouTube and Meta
- 02Coverage check
- Account and campaign coverage
- Ratios, sizes, file and text limits
- 03Coverage plan
- Keep, adapt or generate, gap by gap
- Credit estimate and spend controls
- 04Fill
- Adapt existing creatives
- Generate from the brand kit
- 05Review
- ABCD feedback on every video
- Approve, then ready to download
- Coverage check and plan. The requirements for each campaign type (ratios, minimum sizes, file limits, text lengths) are modelled as config. Coverage is computed for the account and for each campaign, rolled up into a coverage percentage, and turned into a plan of what to keep, adapt and generate.
- Adapt. Reworks existing creatives, including Meta and TikTok ones, to Google’s specs: crop, pad, blurred pad, trim, or AI regeneration. Cross-orientation video was the riskiest item on the feasibility map, so the user picks the method instead of the system guessing.
- Generate. Creates the missing images and videos from the advertiser’s brand kit.
- ABCD review. Every video is checked against Google’s ABCD framework, and the feedback sits next to it when the user approves. Approved assets go to the Coverage Library, and every run is kept in Run History.
- Integrations. Asset import from Google Ads and YouTube, generation on Vertex AI (Gemini and Veo), all built onto AdRoom’s existing pipelines.
Results
- Live in AdRoom since October 2026, announced by Pixis.
- From prototype to production within two months, behind feature flags: the coverage check and plan, adaptation and generation, and ABCD review.
What I’d do differently
- Prototype before the specs. The clickable demo came after the first specs, and it settled product questions (adapt or generate, spend controls) faster than the documents did. Next time I’d build it first.
- Close the integration decision earlier. How assets would be imported from Google Ads stayed an open question in the plan. Settling it on day one would have taken a dependency off the critical path.