Every CTO eventually has this conversation with their CFO: "Why did our infrastructure spend go up again this quarter?" Often, there's no single answer, just a vague sense that costs creep upward while the team's output doesn't grow at the same rate.
The truth is that DevOps costs are rarely just the cloud bill. They show up in engineering hours spent on maintenance, in delayed releases that cost market opportunity, and in incidents that take longer to resolve than they should. This post walks through the signs that your setup is costing more than it needs to, even if the monthly invoice looks "normal."
Sign 1: Your Cloud Bill Grows Faster Than Your Usage
What to Look For
If your AWS or GCP spend has increased 20-30% over the last two quarters, but your user base, traffic, or feature set hasn't grown proportionally, that gap is worth investigating.
What's Usually Behind It
This pattern almost always traces back to a small number of causes: resources provisioned for a traffic spike or migration that were never scaled back down, environments (staging, QA, demo) that run 24/7 despite only being used during business hours, or instance types chosen early on that no longer match actual usage patterns.
Why It Matters
Cloud overspend doesn't just cost money directly, it often signals that nobody owns infrastructure cost as an ongoing responsibility. If nobody's job is to notice when a resource is oversized or idle, the bill simply grows until someone finally asks why.
Sign 2: "Temporary" Fixes Have Become Permanent Infrastructure
What to Look For
Look for scripts, manual processes, or workarounds that were introduced as quick fixes, often during an incident or deadline crunch, and are still in use months or years later with everyone aware they're "not the real solution."
What's Usually Behind It
Under deadline pressure, teams reach for the fastest fix rather than the right one. A manual deployment step, a hardcoded configuration, or a one-off script to patch data becomes part of the regular workflow because replacing it was never prioritized once the immediate pressure passed.
Why It Matters
These workarounds carry hidden costs: they're usually undocumented, often depend on one specific person knowing how they work, and tend to break in ways that are disproportionately disruptive relative to how small they originally were.
Sign 3: Releases Require a Specific Person to Be Available
What to Look For
If deployments only happen when a particular engineer is online, or if "we'll wait until [person] is back" is a common phrase around release timing, that's a structural risk, not just a scheduling inconvenience.
What's Usually Behind It
This typically happens when deployment knowledge lives in one person's head rather than in documentation or automation. It's common in smaller teams where one engineer set up the original CI/CD pipeline and has continued to be the only one who fully understands it.
Why It Matters
Beyond the obvious risk if that person is unavailable during an urgent fix, this bottleneck quietly limits how often the team can ship. Releases get batched and delayed to fit around one person's schedule, which slows down everything from bug fixes to feature launches.
Sign 4: Engineers Spend Significant Time on Infrastructure Instead of Product
What to Look For
Ask engineers directly: in a typical week, how much time goes toward debugging pipeline issues, fixing environment problems, or dealing with deployment friction, versus building product features? If the answer is consistently more than 15-20%, that's a meaningful drag on output.
What's Usually Behind It
This is often the cumulative effect of several smaller issues, flaky tests, slow builds, environment drift, each of which seems minor individually but adds up to a significant tax on engineering time across the team.
Why It Matters
This is the most expensive item on this list, and the hardest to see on a budget spreadsheet. If you're paying senior engineers $80-150k+ a year and 20% of their time goes to infrastructure friction rather than product work, that's a substantial hidden cost that never shows up as a line item, it just shows up as slower roadmap progress.
Sign 5: Incidents Take Longer to Resolve Than They Should
What to Look For
Track how long it takes from "something is wrong" to "it's fixed and verified." If this regularly takes hours for issues that, in hindsight, had an obvious cause, that gap between detection and resolution is a cost center.
What's Usually Behind It
Slow incident resolution is usually a visibility and rollback problem, not a skill problem. If logs are hard to search, if it's unclear which recent deploy might be responsible, or if rolling back takes as long as the original deploy, engineers spend most of the incident just figuring out what's happening rather than fixing it.
Why It Matters
Every hour of downtime or degraded service has a real cost, in lost revenue, support tickets, or customer trust, that's typically much higher than the engineering hours spent resolving it. Slow incident response multiplies the cost of every issue that occurs.
Sign 6: New Team Members Take Weeks to Become Productive
What to Look For
If it consistently takes a new engineer two, three, or more weeks before they can ship their first meaningful change independently, the setup itself is likely part of the problem, not just the learning curve of the codebase.
What's Usually Behind It
This usually points to undocumented setup processes, environments that are difficult to replicate locally, or tooling that requires tribal knowledge to use correctly. New hires spend their first weeks debugging their own environment rather than learning the product.
Why It Matters
Onboarding friction is a recurring cost, it happens every time someone joins the team, and it compounds as the team grows. A setup that takes two weeks to onboard costs roughly the same in lost productivity each time, regardless of how good the new hire is.
How to Quantify These Costs
Most of these signs are qualitative on their own, but they can be roughly quantified:
-
Engineering time on infrastructure: estimate the percentage of time spent, multiply by team size and average salary
-
Idle or oversized cloud resources: most cloud providers offer cost recommendation tools that flag underutilized resources directly
-
Incident cost: multiply average resolution time by the number of incidents per month, then by an estimated cost-per-hour of downtime or degraded service
-
Onboarding cost: multiply weeks-to-productivity by the number of new hires per year and their salary
Putting even rough numbers against these signs is often what turns "something feels off" into a concrete case for action, particularly when making that case to a CFO or board.
What to Do Next
None of these signs, on their own, mean your setup needs to be rebuilt. Most of them point to specific, addressable issues: cost monitoring that isn't happening, knowledge that isn't documented, or pipeline friction that hasn't been prioritized because it's never urgent enough on any single day.
The first step is the same regardless of which signs you're seeing: get a clear picture of where time and money are actually going. Our guide on running a DevOps audit covers how to do this systematically.
