To reduce your Datadog bill, cap high-cardinality custom metrics, index only the logs you actually query (and tier the rest to cheaper storage), trim host count and product sprawl, and commit annually to your baseline usage. Datadog prices each product independently and bills mostly per host and per GB, so costs compound fast — most teams find their invoice runs 2–3× their original estimate.
This guide breaks down where the money leaks and how to plug it without losing the visibility you actually need. For the full pricing structure, see Datadog pricing explained; to compare options, see best cloud monitoring tools.
Key takeaways
- Datadog is a “utility bill with many meters” — Infrastructure, APM, Logs, RUM, and more each bill separately and stack.
- The biggest hidden drivers are custom metrics cardinality and log indexing volume.
- The $15/host headline is the floor; a typical Infra + APM + Logs host runs $40–$70+.
- On-demand usage costs 20–50% more than an annual commitment to the same baseline.
Why Datadog bills surprise teams
Datadog doesn’t have one price — it has many. Infrastructure monitoring is billed per host (around $15/host/month on the annual Pro plan, more on-demand or Enterprise). APM adds roughly $31/host on top, and crucially, APM hosts must also be licensed as infrastructure hosts, so the two scale together. Logs are billed separately for ingestion and indexing. Custom metrics, RUM, Synthetics, and security products each have their own meter.
The result: enabling a second and third product doesn’t add cost linearly — it compounds across independent dimensions you can’t see until the invoice arrives. That’s why teams routinely budget $2,000 and receive $5,000+.
The real cost drivers
- Custom metrics cardinality. Each plan includes an allotment per host (around 100 on Pro, 200 on Enterprise). A single high-cardinality tag — per-user, per-request, per-container — can spawn thousands of custom metrics from one service overnight, and overages bill per increment.
- Log indexing. Datadog charges ~$0.10/GB to ingest logs and a much larger ~$1.70 per million events to index (make them searchable). Indexing is usually the biggest single lever.
- Host count, including ephemeral hosts. Short-lived containers, autoscaling nodes, and forgotten instances all get billed. Spiky traffic inflates host counts without warning.
- APM span ingestion and retention beyond the per-host allotment.
- Product sprawl. Every extra product (RUM, Synthetics, Security) is another meter running.
- On-demand overages. Anything above your annual commitment bills at a premium, typically 20–50% higher.
10 ways to reduce your Datadog bill
1. Cap and clean up custom metrics
Find your highest-cardinality metrics and remove tags you never query. Use Datadog’s “Metrics without Limits” to keep ingesting data while indexing only the tag combinations you actually need — this decouples ingestion cost from the much pricier indexed volume.
2. Index only the logs you query
This is the single biggest lever for most accounts. Ingest broadly for archiving and search-on-demand, but index only high-value logs. Indexing every log “just in case” is where bills explode.
3. Tier low-value logs to Flex storage
Move high-volume, low-value logs (debug, staging, verbose access logs) to Flex storage, which is dramatically cheaper per event than standard indexing. Reserve standard indexing for logs that need active monitoring and alerting.
4. Filter logs at the source
Don’t ship debug and staging logs to Datadog at all if you don’t need them there. Route them to cheaper object storage. A single verbose microservice can generate hundreds of GB a month.
5. Trim host count and exclude ephemeral hosts
Aggressively exclude short-lived containers and non-essential hosts from billing. Don’t pay full host price for instances that exist for five minutes.
6. Right-size APM sampling and retention
You don’t need to index and retain 100% of traces. Keep error, high-latency, and business-critical traces; sample the rest. RED metrics (requests, errors, duration) are computed on 100% of traffic regardless of sampling, so you keep accuracy while cutting cost.
7. Prune product sprawl
Audit which products are actually used. Turning off an under-used RUM, Synthetics, or security product removes an entire meter.
8. Downgrade non-production environments
Dev and staging rarely need Enterprise-tier monitoring or long retention. Use cheaper plans and shorter retention for non-prod.
9. Commit annually to your baseline
On-demand rates run 20–50% above annual. Commit to the baseline you’re confident you’ll use and let overages flex — don’t commit to your peak. Negotiate overage caps in the contract.
10. Give the bill an owner and allocate it
Untracked observability spend is unmanaged spend. Tag usage by team and allocate cost back to who generates it, and connect it to business value — cost per service, per customer. This is the unit economics and allocation discipline of FinOps. A dedicated platform can automate Datadog cost allocation; see best FinOps tools.
The core trade-off to manage
Datadog forces a tension between cost and visibility: indexing less and sampling more saves money but reduces what’s instantly searchable during an incident. The goal isn’t less observability — it’s focusing observability on high-value signals. Decouple ingestion (cheap, keep it broad) from indexing (expensive, be selective), and you keep the visibility that matters while cutting the spend that doesn’t.
Frequently asked questions
Why is my Datadog bill so high?
Because Datadog meters each product separately and bills per host and per GB, costs stack quickly. The usual culprits are high-cardinality custom metrics, indexing too many logs, ephemeral host counts, and product sprawl — which is why bills often land 2–3× higher than estimated.
What is the biggest Datadog cost driver?
For most accounts it’s log indexing (~$1.70 per million events) and custom metrics cardinality. Indexing only the logs you query and capping high-cardinality metrics are the two highest-impact fixes.
How can I reduce custom metric costs in Datadog?
Remove unused high-cardinality tags and use “Metrics without Limits” to index only the tag combinations you query, while still ingesting the rest. A single misconfigured tag can generate thousands of billable metrics.
Does reducing hosts always lower my Datadog bill?
Not necessarily — logs, custom metrics, and APM are billed on separate meters, so cutting hosts alone may not move the bill much if those other dimensions are the real drivers.
Is an annual Datadog commitment cheaper?
Yes. Annual commitments are typically 20–50% cheaper than on-demand. Commit to your baseline usage and negotiate overage cap terms rather than committing to peak.
Written by the FinOpsDaily team. Last updated: 2026. Datadog pricing varies by plan, commitment, and product mix and changes over time — verify current rates on Datadog’s pricing page before deciding. This is a sensitive area for budgets; treat figures as directional.
