Cost as a first-class SLO: building a cost-aware culture

Here’s the pattern almost every organization repeats: the bill gets scary, someone runs a big optimization push, savings land, everyone moves on — and six months later the bill is back where it started. The cleanup worked. The culture didn’t. Cost drifted right back up because nothing about how the team works day-to-day had changed.

The teams that stay lean don’t run better cleanups. They’ve made cost a first-class engineering signal — treated like reliability, owned by the people who create it, and watched continuously instead of rediscovered quarterly.

Reliability got this right; cost can borrow it

A decade ago, reliability was somebody else’s job — a separate ops team you threw releases over the wall to. SRE fixed that by making reliability an engineering concern with owners, error budgets, and continuous signals. Cost is where reliability was before that shift: outsourced to finance or a FinOps team, addressed in retrospect, disowned by the engineers who actually drive it.

The fix is the same move. Cost isn’t a report you receive; it’s a property of the systems you build, and it belongs with the people who build them.

What “first-class” actually means

• Owned — every service has a team accountable for its cost, the same way it’s accountable for its uptime. This is why attribution is the foundation: no owner, no ownership.
• Budgeted — teams get a cost or unit-cost target, like an error budget. A budget turns “spend less” (nobody’s job) into “stay under this” (a clear goal).
• Continuous — cost shows up in the normal flow of work: a regression flagged on the deploy that caused it, not in a quarterly review three months later when nobody remembers the change.
• Visible — the numbers are in front of the team that moves them, in units they understand, not buried in a finance dashboard they never open.

Make the feedback loop short

The reason quarterly cleanups don’t build culture is the lag. A cost regression discovered three months after the deploy that caused it teaches nobody anything — the context is gone, the author has moved on, the lesson doesn’t land. Shorten the loop to days and the same regression becomes a learning moment: this change moved that number, caught while it’s still fresh. Short feedback loops are how any engineering discipline gets internalized, and cost is no exception.

Continuous by design, not by willpower

The honest problem is that “watch cost continuously” is not something busy teams sustain by discipline. Manual vigilance decays; the cleanup-and-regress cycle is what you get when the watching depends on someone remembering to look. Culture sticks when the continuous signal is built into the tooling, so staying lean is the default path, not a heroic effort.

That’s the role we built Lizrd to play — it keeps the loop running so cost behaves like any other engineering signal: continuously reading utilization and spend, attributing it to owners, and surfacing regressions and opportunities as concrete, reviewable changes in the flow of work. A one-time cleanup lowers the bill. A cost-aware culture keeps it low — and culture is what you get when the signal never stops.