Latest / The Cloud Business Podcast with Fexingo: AWS, Azure, GCP, and Enterprise Infrastructure / The Cloud SLA Audit That Could Save Your Enterprise Millions
Transcript
- Lucas: So there's this number that most enterprise cloud buyers never think about until it's too late: the SLA credit. And I don't mean the uptime percentage on the marketing page — I mean the actual contractual credit you're owed when the provider falls short. Luna: Right, because everyone talks about 'five nines' but nobody actually reads the fine print on what happens when you don't get them. Lucas: Exactly. And here's the thing — a Fortune 500 company I spoke with recently did an audit of their AWS usage over a twelve-month period. They found that AWS owed them roughly one point two million dollars in unclaimed SLA credits. AWS never proactively offered a cent. Luna: Wait — AWS just held on to that money? Even though they knew they hadn't met the SLA? Lucas: That's the thing. The SLA is structured as a credit you have to claim. It's not automatic. Most providers — AWS, Azure, Google Cloud — they all require you to submit a request within a certain window, usually sixty to ninety days from the incident. If you don't file, you don't get paid. Luna: So it's like a rebate you have to know about and chase down. And of course, the provider isn't going to send you a reminder. Lucas: Precisely. And for that Fortune 500 company, the credits came from things like a three-minute EC2 outage in us-east-1 that dropped their compute availability below 99.99 percent for the month. Individually, each credit is small — maybe a few thousand dollars. But when you add up every eligible incident over a year, it becomes real money. Luna: So how do you even know when an incident qualifies? Because I assume not every blip triggers a credit. Lucas: Right, the SLA is defined per service. For EC2, the standard monthly uptime percentage commitment is 99.99 percent for a single instance. If you're running a hundred instances across multiple availability zones, the math gets more complex. But the key is that you need to track uptime at the instance level, not just the region level. Luna: So you need tools that can log every instance's uptime and compare it against the SLA. Most enterprises probably don't have that in place. Lucas: Exactly. The company I spoke with used a third-party SLA monitoring tool that cost them about forty thousand dollars a year. They recovered one point two million in credits. That's a thirty-to-one return. And they also used the data to renegotiate their enterprise discount with AWS. Luna: Because now they had proof that the service wasn't as reliable as advertised. That's leverage. Lucas: Exactly. And it's not just AWS. Azure has a similar structure — for virtual machines, they commit to 99.95 percent uptime for single instances, 99.99 percent if you use availability zones. Google Cloud's Compute Engine SLA is 99.95 percent for a single instance. And they all require you to file within a window. Luna: So the first step is just knowing the SLA terms for each service you use. Which is surprisingly hard because the SLA documents are long and legalistic. Lucas: Right. And they change. AWS updated its SLA for EC2 in 2023 to make the credit calculation more granular. If you're not tracking those changes, you might miss a new eligibility condition. Luna: So what's the concrete three-step audit you mentioned? Lucas: Step one: inventory every cloud service you're paying for and map it to the specific SLA document. Don't assume all services under a provider have the same SLA. RDS has different terms than EC2. Lambda has a different SLA than EKS. Luna: Right, so you need a service-by-service mapping. That's work, but it's one-time. Lucas: Exactly. Step two: set up monitoring that tracks uptime per service per region at the granularity the SLA defines. For EC2, that means per-instance uptime. For RDS, it's per-database instance. You need logs that show the exact minute an instance was unreachable. Luna: And then step three is actually filing the claims, which has its own process. Lucas: Yes. You need to submit a support ticket within the claim window — typically 60 days for AWS, 90 for Azure. You need to include the affected resource IDs, the outage start and end times, and sometimes a log file. The provider may push back, but if you have the data, you usually win. Luna: And do the credits come as cash? Or just service credits? Lucas: They're service credits — applied to your future bills. So it's not cash in hand, but it reduces your net spend. For a company spending ten million a year on cloud, recovering one percent through SLA credits is an extra hundred thousand dollars you didn't budget for. Luna: That's the kind of thing that would get a FinOps team noticed by the CFO. Lucas: Absolutely. And the interesting thing is that the providers know this is happening. They've priced the SLA credit risk into their margins. But they're counting on most customers not claiming. It's a behavioral economics play — the friction of filing outweighs the perceived benefit for most teams. Luna: So the ones who do file are essentially capturing value that the provider has already accounted for. It's like an arbitrage. Lucas: Exactly. And the tools to automate this are getting better. There are startups now that integrate with your cloud billing and monitoring APIs, automatically detect SLA breaches, and file the claims on your behalf. They take a percentage of the recovered credits — usually around twenty percent. Luna: So you can outsource the entire process. That seems like a no-brainer for larger enterprises. Lucas: For the Fortune 500 I mentioned, they did it in-house because they wanted to keep the full credit. But for a lot of companies, the twenty percent fee is worth it to avoid the hassle. And the ROI is still enormous. Luna: It reminds me of the old telecom audit firms that would find billing errors and split the savings. Same model, different infrastructure layer. Lucas: Exactly the same. And the cloud market is now mature enough that this kind of optimization is becoming standard practice. Every major enterprise should be doing an SLA audit at least annually. Luna: So for listeners who want to start, what's the first concrete step they should take today? Lucas: Pull your cloud billing data for the last three months and identify your top ten services by spend. Then go read the SLA for each one. Just that exercise will probably reveal at least one instance where you could have claimed a credit. And if you're curious, start tracking uptime for those services manually or with a tool. Luna: That's actionable. And honestly, if this episode saves one listener ten thousand dollars, it's worth the time we spent. Lucas: Which is kind of the whole point of this show — finding those hidden optimizations that actually move the needle. And speaking of that, a quick honest thing: a handful of listeners chip in monthly through buy me a coffee dot com slash fexingo, and that's literally what funds making this many of these episodes. No ads, no sponsors — just listener support. So if you've gotten value from episodes like this one, that's the way to keep it going. Luna: Yeah, it's a small group, but it makes a huge difference. Keeps us free to dig into whatever angle we think is actually useful. Lucas: Exactly. And back to the audit — once you've done the top-ten service review, the next level is to automate the monitoring. There are open-source tools like Cloud Custodian that can track SLA metrics across AWS and Azure. You don't need a paid tool to start. Luna: And I imagine the process is similar for Google Cloud and other second-tier providers. Lucas: Yes. Google Cloud's SLA credits are also claim-based. Their compute engine SLA has a 99.95 percent uptime commitment for standard VMs. And they have a 'financial credit' request form you have to fill out within 30 days. That's a shorter window than AWS or Azure, which is another reason to monitor proactively. Luna: So if you're multi-cloud, you need different processes for each provider. But the principle is the same. Lucas: Exactly. And the one thing I want listeners to take away is this: cloud providers are not going to give you money you don't ask for. The SLA credit is a contractual right, but it's not a gift. You have to pursue it. Luna: And if you're not monitoring, you're essentially leaving money on the table — month after month after month. Lucas: Yeah. And that's the kind of inefficiency that, once fixed, just keeps giving. Set it up once, and it becomes a recurring savings line item on your P&L.