Divestiture · 2022–23
Carving TV Guide and Metacritic out of a shared platform
TV Guide and Metacritic shared a multi-tenant CMS platform with CNET, ZDNET and Download.com. When the company sold both brands, I served as their sole engineering contact and led their technical carve-out—giving each brand its own cache, build and deploy pipelines, and codebase free of company-private packages while the shared platform kept serving the remaining three brands. The separation finished on schedule without delaying the sale.
Context
One small platform team ran a shared CMS and API for five brands. The brands shared one codebase, one multi-site database with each brand's content keyed by a site ID, a shared cache and search tier, and shared build and deploy pipelines. I was the only engineer the TV Guide and Metacritic teams worked with directly. We met weekly to build and ship their features.
Before the sale
In early 2022 I onboarded Metacritic onto the CMS as a new site. It had its own Kubernetes and Helm configuration and its own CI pipelines, and ran beside the shared platform rather than inside it. It went live that February. Metacritic reused TV Guide's content code instead of having its own, so the two brands' code had to leave as one unit.
The carve-out
The shared platform had to keep serving CNET, ZDNET and Download.com while the two brands left it.
- Split the shared cache. TV Guide and CNET shared memcached instances. I gave TV Guide its own and moved CNET's cache settings out of the shared configuration, so either side could change without affecting the other.
- Cut a standalone release branch. I cut a TV Guide release branch with CNET-only components stripped out and removed dependencies on internal company packages so the codebase could build and run independently.
- Separate pipelines. Metacritic already had its own pipelines from 2022. Just before the sale, I gave TV Guide its own build and deploy pipelines for each environment on that release branch and detached the brand from the shared pipelines.
- Stood it up with their teams. I worked with each brand's team to get its standalone setup running.
- Cleaned up the shared side. After the sale I removed TV Guide's scheduled jobs from the shared platform. The team then deleted the two brands' code, about 700 files, from the shared codebase.
Both brands left with code free of the company's private packages, their own cache and pipelines, and a standalone deployment their engineering teams stood up alongside me. The CNET side of the platform moved from Docker Swarm to Kubernetes the following year: see that case study.
My part: I was the engineering contact for both brands and led their technical carve-out: the cache split, the release branch and dependency removal, getting TV Guide off the shared pipelines, the hand-off work with their teams, and the clean-up on the shared platform.
What I carry forward
Map every shared dependency before the deal date: cache, private packages, pipelines and scheduled jobs. Code that has to leave together, like Metacritic on TV Guide's code, moves as one unit, so plan the split around those ties instead of brand by brand.