Observability Doesn't Have to Be Expensive

As software engineers, we learn many best practices. A few of these are non-negotiable: software that follows these practices is strictly better than software that doesn't. For example, version control, unit testing, and observability.

Without version control, collaboration is impossible; without unit testing, software is buggy. Without observability, we can't tell if our users are hitting errors or slowness, or whether our service is completely down.

Unfortunately, there's one big difference between observability and the other best practices: it's orders of magnitude more expensive. Thus, it's not nearly as widely adopted. Many projects avoid observability entirely. Even if a team does use it, they often operate in a failure mode: aggressive sampling, short retention, or turning off features.

Why is it so expensive? Some vendors claim that it's because modern software is too complex, while others say it's because observability itself is expensive to operate. This paints an extremely incomplete picture.

The truth is, almost every observability company is tuned to serve the biggest companies. This is only rational: it is widely assumed that big customers are the best customers. But serving big customers changes you. You hire salespeople, account managers, and many (many) engineers to win big accounts and squeeze as much revenue from them as possible. Inevitably, high prices follow.

It doesn't have to be that way. There are thousands of serious engineering teams that need observability but not the big company sales dance. A business built to serve them is fundamentally different from existing observability companies: lean, efficient, and far less expensive to operate. There is no need to charge extraordinary prices to build a sustainable observability company.

Every service, every Kubernetes cluster, every server should be observable. We've been told that's too expensive. It doesn't have to be.

Why observability loves big customers

Observability companies skew towards big customers for the same reason all B2B businesses do.

It is no secret that most software companies optimize for revenue growth. This is especially true if they take investment.

When you optimize for revenue growth, it's frequently advised to seek big customers. To name just a few reasons:

• A big customer can bring as much revenue as 100 smaller ones, but is not 100x more difficult to win.
• They have lower churn, e.g., because they are less likely to fail.
• They have more opportunity for account expansion.

Consider Datadog. When they filed their S-1 to go public, customers spending more than $100k a year already represented 48% of ARR. By 2025, that had grown to 90%.¹ Clearly, the money is in big customers.

Serving big companies changes you

It's not enough to simply want big companies. You have to shape your company around them.

To win big customers, you need salespeople. If someone is going to commit $100k a year, they want to talk to you. They want a contract. They want to know you can handle security review, compliance, support. They want to know you will cater to their needs.

Since each large customer represents a meaningful chunk of revenue, losing one hurts. You need account managers and customer success staff, too.

This means building out a large sales and marketing organization. In 2019, 39% of Datadog's employees were in sales and marketing, slightly more than in R&D.²

That isn't all. Once a company is oriented around large customers, it is not enough to win those customers. Each account must be expanded. This is especially important if you need to grow fast and/or have investors.

A great way to grow a customer is to build more features and charge for them. Because the customer is big, they likely have a need for those features, and they likely have the money to spend.

This again takes a lot of people. In 2019, Datadog had 462 R&D employees.²

Of course, those engineers do a lot more than build enterprise features. But once a company is built around large accounts, the roadmap bends towards serving them. The core product inevitably suffers.

This is the standard B2B SaaS playbook. And it worked: Datadog's market cap of tens of billions of dollars is eye-watering. The observability market supports many other successful companies, seemingly all of them targeting the biggest customers.

Unfortunately, that leaves most of the market rationing telemetry, under-instrumenting, or priced out entirely.

An observability company for the rest of us

What does a company built to serve the rest of us look like?

First of all, it is lean. As we have seen, large teams lead to large costs.

Second, it is efficient. Find ways to get infrastructure costs down.

Third, it is focused. Nail the core observability features we all need.

Finally, shape it around self-serve customers. Make pricing transparent. Write great documentation. Let people buy without excruciating sales calls.

Such a company can solve the core promise of observability without being sucked down the enterprise pricing gravity well.

Observability should be a no-brainer

With all the discussion of observability's high price, something gets forgotten: observability is great.

Who doesn't want to sleep well at night knowing their service is healthy?

All software benefits from observability, and we all benefit when software runs well. We should want observability to be widely adopted.

We've been told that's too expensive.

It doesn't have to be.

Footnotes

1. Datadog, Inc., 2025 Form 10-K. Used for the 2025 customer count and ARR concentration. ↩

2. Datadog, Inc., Form S-1/A filed September 9, 2019. Used for the 2019 customer count, ARR concentration, S&M/R&D headcount, and quoted expansion strategy. ↩ ↩²