What we use Google Cloud for
The most common reason is a simple one: the company already works with Google Workspace, the reporting sits in BigQuery, and a new application needs to reach that data. The second reason is individual services — Cloud Storage for media, Cloud SQL for managed databases, the Vision or Translation services when an application has to process images or language. The third reason is a corporate policy that leaves no room for discussion.
In practice we mostly build applications there that come out of our own development: Symfony or Node.js services in containers on Cloud Run, plus Cloud SQL, Cloud Storage and Pub/Sub for messages between systems. Interfaces to ERP, CRM or shop work the same as anywhere else — the platform changes the location, not the task. GKE only comes into play when there really are many services to orchestrate.
How we work with it
Set up as code, with Terraform, split into separate projects for test and production. Rights run through IAM and service accounts, not through personal accounts that nobody can attribute after the next change of staff. Regions are set to Europe, as a rule Frankfurt or Belgium. We manage certificates, DNS and domains as well — including when the application itself sits at Google and everything else with us.
We treat costs as part of operations, not as a surprise at the end of the month. Budgets, alarms and a regular look at which resources are actually in use belong to that. BigQuery in particular repays the attention: in the on-demand model you are billed by the volume of data processed, so an unfortunate query gets expensive quickly; in the capacity-based editions you pay by slot hour instead. Partitioned tables and bounded queries are not a fine detail here, they are the basis. Support and maintenance are governed by individual service level agreements.
Limits and alternatives
Cloud for its own sake does not add up. A TYPO3 site with steady traffic and a Shopware shop with predictable peaks do not need a platform built for worldwide load distribution. They need an appropriately sized environment, backups and someone who does the updates. That is exactly what our own servers in Germany provide — predictable in cost and without dependence on a provider whose price list can change.
Where Google Cloud is hard to beat: analysis across very large volumes of data, and services you could barely run sensibly yourself. Then the question is not whether but how tightly you tie yourself in. So we keep application logic portable and use platform-specific services deliberately, where the benefit justifies the later cost of moving. Which route fits, we establish before the quote.