If you have ever opened a Datadog invoice and felt your stomach drop, you are not alone. Datadog is a capable platform, but its pricing is built to grow with your data, and telemetry only ever grows. This guide covers how to reduce Datadog costs in practical terms: where the bill actually comes from, which levers work, which levers quietly cost you data, and how a different architecture removes the trade-off between Datadog cost optimization and keeping your telemetry complete.
Why Datadog bills grow the way they do
Datadog meters most of what matters by volume and cardinality: ingested and indexed logs, custom metrics and their cardinality, ingested and indexed spans, and a per-host component on top. The result is that your bill tracks your success. More traffic, more services, more custom metrics, and more regions all push the number up. The usual advice for reducing your Datadog bill reflects that structure:
- Index fewer logs (ingest-then-filter, exclusion filters).
- Cut custom metric cardinality.
- Turn down trace sampling and retention.
- Trim per-host and APM footprints.
Some of this is healthy hygiene. But notice the pattern: nearly every lever asks you to keep less data. That is the same false choice we cover in sampling vs. lossless: you lower cost by lowering fidelity, and you find out during the next incident which logs you should have kept.
The cost line most teams miss: egress
There is a charge that never appears on the Datadog invoice but shows up on your cloud bill instead. Because Datadog is a SaaS platform, your agents ship telemetry out of your cloud to Datadog’s. Those are data egress fees you pay your cloud provider, on the highest-volume data you produce, every single day. When you model the true cost of a SaaS observability platform, egress belongs right next to ingestion and indexing. We break it down in the hidden cost of observability data egress.
A different way to cut the bill: compress, don’t drop
The real reason Datadog cost optimization feels like a fight is that volume is coupled to cost. Break that coupling and the fight ends. Sasquatch Labs reduces observability and SIEM cost with patent-pending lossless compression, shrinking logs, metrics, and traces by roughly an order of magnitude or more while keeping every byte. The data stays complete and fully queryable, so you are cutting cost by lowering the cost per byte, not by throwing telemetry away.
Two architectural differences do most of the work:
- It runs in your own cloud. Telemetry never egresses to a vendor, so those cloud transfer fees disappear and your data stays under your control. See BYOC observability.
- Query happens in place. The Snowman engine reads compressed data directly and speaks LogQL, SPL, TraceQL, and PromQL, so there is no re-ingestion and no separate search-compute meter.
A sensible plan to reduce Datadog costs
- Do the free hygiene first. Kill genuinely useless custom metrics and duplicate pipelines. That is real waste, not data you need.
- Stop paying to drop. Before you accept sampling and index exclusions as permanent, price the alternative: keeping everything, compressed.
- Put egress on the whiteboard. Count the cloud transfer cost of shipping telemetry out, not just the SaaS invoice.
- Model cost per byte, not just volume. The number that changes the economics is the compression ratio. See the math of observability cost reduction.
You can reduce a Datadog bill by keeping less data. You can also reduce it by keeping the same data for a fraction of the cost. Only one of those leaves you able to answer the next hard question. See how the second option works at sasquatchlabs.io.