What changed in Q1.
A field note from our customer base. The trends we're watching, the trends that surprised us, and the ones we're ignoring on purpose.
Every quarter we pull aggregated patterns from our customer base. We don't look at any single customer's data — we look at the shape of the work, across thousands of teams.
Some of what we see is obvious. Most of it isn't.
Three things we noticed this quarter
Deploys got smaller, and faster. Median deploy size across the customer base dropped 18% quarter over quarter. Median deploy duration dropped 22%. We didn't ask anyone to do this. The platform incentivizes it — smaller deploys, faster feedback — and over time teams converge on the local optimum.
Observability replaced staging. Customer accounts spinning up new staging environments dropped by half. Customer accounts using production traffic replay against canaries doubled. Staging is still a thing, but for fewer teams it's the primary safety mechanism.
Incident response shifted into design review. This was the surprise. Customers are increasingly catching the failure modes in their design review process, weeks before the code ships. We didn't build a feature for this — they just have more legible runtime data now, and they're using it to ground architectural conversations.
What we're not chasing
We're not building an AI assistant for incident response. We're not building a "data lake explorer" for the data layer. Our customers ask for both. We say no because the data show they have time now — what they need is not more tools.
The trend we're watching most closely is what teams do with the time they get back. That's where the next platform features will come from.
Iris writes about platform engineering, post-incident learning, and the operational habits of teams that ship without burning out.