Cardinality: Why Your Datadog Bill Is Out of Control

What Is High Cardinality — and Why Does It Cost So Much? In the Datadog world, cardinality refers to how many unique combinations of tags or attributes your telemetry data generates — particularly in custom metrics within the Infrastructure and Datadog APM modules. Take a metric like api.request.count. Seems harmless — until it’s tagged with env, region, user_id, and container_id. You now have thousands (or millions) of unique time series metrics. Each of those gets billed. Every Datadog metric query across them slows down. And nobody realizes it until the graphs started lagging and the Datadog pricing model kicks in, resulting in high Datadog cardinality costs.

The issue gets even more expensive with Datadog logs. If every log line includes dynamic fields — session tokens, headers, request payloads — you’re ingesting a massive volume of indexed log data. Even if it’s rarely queried, you’re still paying to ingest, store, and process it. And now, increasingly, Datadog tracing data is adding to the problem.

Datadog trace search is essential for understanding how requests move through your system — but traces are also full of dimensions. Every span can carry tags like:

• Service name
• Endpoint
• Status code
• Container, pod, or host ID
• User context
• Feature flags
• Region and environment

Multiply that by thousands of traces per minute, and suddenly your Datadog APM pipeline is ingesting millions of unique tag combinations from traces alone. Because traces often reflect dynamic, real-time request data, the cardinality footprint is unpredictable — which makes cost control in Datadog harder and surprise bills more likely.

The financial impact is obvious: more data, more cardinality, more cost. But the operational impact is just as painful:

• Dashboards slow down or time out
• Queries become unreliable
• Engineers lose trust in the tools
• Teams start turning things off to save money — and fly blind
• Eventually someone says, “Should we just leave Datadog?”

But switching platforms isn’t the solution. The real question is: How can we make Datadog more cost-efficient?

That’s where ControlTheory comes in. We help you understand what’s actually flowing into Datadog — not just telemetry volume, but telemetry shape:

1. Where high cardinality is coming from
2. Which dimensions are driving up Datadog costs
3. How traces, logs, and metrics contribute

This gives you the visibility you need — a bird’s-eye view of your telemetry. Once you can see what’s happening with your Datadog cardinality, you’re in a position to optimize it. We work with you to implement Datadog cost optimization strategies through telemetry pipelines that:

1. Filter out low-value data before ingestion
2. Downsample or aggregate metrics
3. Remove volatile dimensions (like user_id)
4. Offload raw logs to cold storage

These policies are implemented between Datadog agents and the backend, often using OpenTelemetry for transformation and control.

“Increasingly complex systems and ballooning telemetry volumes have made observability costs and processes an operational challenge for many organizations. Concepts like controllability aim to address these issues and necessarily evolve how we think about observability by focusing on actively governing, shaping, and optimizing telemetry rather than just collecting it.”

Kelly Fitzpatrick Senior Analyst at RedMonk

“By monitoring and routing logs, traces and metrics as they move across different data silos and to leading observability platforms, customers get application visibility while controlling costs and reducing data vendor lock-in.”

Jason Englishpartner and principal analyst at Intellyx

“ControlTheory’s approach merges observability with feedback-driven control-ability, using a closed-loop control plane that balances cost and value, delivering exactly the data you need, precisely when you need it.”

Brian DucharmeVMBlog.com

“ControlTheory is pushing the boundaries of observability by introducing the crucial concept of Controllability, which empowers businesses to immediately manage costs, optimize performance, and position themselves for the AI-enabled future.”

Kip McClanahanGeneral Partner at Silverton Partners

“We’re excited to welcome ControlTheory to the CNCF as a new member. The future of observability is open—with projects like OpenTelemetry leading the way as the 2nd highest velocity open source project behind Kubernetes. ControlTheory’s innovative approach to controllability empowers organizations to regain control of their current observability, optimize existing stacks, and accelerate their journey toward an open, interoperable future.”

Chris AniszczykCTO, CNCF

Here’s the smart part: just because data isn’t going to Datadog doesn’t mean it’s gone. We route non-critical telemetry to low-cost S3 storage, where it’s retained for compliance, rehydration, or ingestion into other tools like OpenSearch or ClickHouse. You’re not deleting — you’re taking control of your observability pipeline.

1. Datadog becomes faster and more responsive
2. Trace and log data is curated, not chaotic
3. Engineers spend less time wrangling noisy dashboards
4. Your bill drops — without losing visibility
5. You get sustainable observability without overpaying

If your Datadog bill feels disconnected from the value you’re getting, you’re not alone. But you’re not stuck, either. ControlTheory helps you optimize Datadog cardinality and usage, reduce cost, and build a smarter observability stack.