Best European Alternatives to Datadog

Why people leave Datadog

Let us dispose of the argument you were probably expecting. Datadog runs nine sites, one of which, EU1, is in Germany at app.datadoghq.eu. Choosing it costs nothing extra and is available on any plan. If your only requirement is that the telemetry stays on European soil, Datadog already does that, and a page that pretends otherwise is wasting your afternoon.

What people actually leave for is the invoice, and specifically its shape. Infrastructure Pro is $15 per host per month on an annual commitment, $18 without one. APM is a further $31 per host on top of that infrastructure.

Logs are $0.10 per ingested gigabyte and then $1.70 per million events to index them for fifteen days. One ordinary machine with infrastructure and APM attached is $46 a month before a single log line arrives, and nobody ever has one machine.

The second reason is that the choices get harder to reverse over time. The site is picked when the account is created and Datadog documents no path between sites. The agents are Datadog's own, so instrumentation is an asset that only spends in one shop. And the open-source projects European teams were assembling their own stacks from keep being bought: Quickwit, announced on 9 January 2025, is now Datadog's.

• The bill multiplies rather than adds Every product is priced per host and the products stack. Infrastructure at $15, APM at $31, logs by the gigabyte on the way in and by the million events on the way into the index, synthetics by the check run, real user monitoring by the session. Each line looks reasonable on its own, which is precisely why the total surprises people. The only lever you are given is to collect less, and collecting less is the opposite of what you bought.
• Indexing is a second, separate meter The detail most teams discover during an incident: ingesting a gigabyte of logs costs from $0.10, and making those logs searchable costs $1.70 per million events for fifteen days of retention. So the question "why is this log line not in the index" has a price attached, and the answer is usually an exclusion filter somebody added last quarter to keep the bill down. A store you own does not present that trade-off at three in the morning.
• Your site is a one-time decision The nine sites — three in the United States, two federal, EU1 in Germany, UK1, and two in Asia-Pacific — are separate installations, and the documentation describes how to choose one rather than how to move between them. Anyone who started on US1 years ago and now needs Frankfurt is not changing a setting; they are standing up a second account and rebuilding the dashboards, monitors and integrations inside it.
• The company is American wherever the bytes are Datadog, Inc. is at 620 8th Avenue in New York, certified under the EU-U.S. Data Privacy Framework with an Irish entity appointed as its Article 27 representative. That is a well-run compliance posture and it does not change which legal system reaches the company. EU1 answers a question about storage. It does not answer a question about disclosure, and no amount of region selection will.

What you have to replace, not just match

Datadog is not one product, and the two European tools on this page are not a set that adds up to it. Being honest about that is more useful than pretending.

The layers worth separating: infrastructure metrics from hosts and containers; application performance, meaning distributed traces; log management, meaning ingest, search and retention; synthetic monitoring, meaning scripted checks against your endpoints and journeys; and real user monitoring with session replay. Below those sits the integration catalogue, which is the reason Datadog installs in an afternoon.

Elastic APM takes the first three, provided you are willing to run or rent the store they live in. Checkly takes the fourth, and takes it further than Datadog does. Nothing here takes the fifth, and nothing here matches the catalogue — a European stack usually ends up as two tools plus Prometheus for the parts neither covers.

The alternatives compared

How each alternative compares to Datadog

What actually breaks when you switch

Agents first. Datadog's agent is proprietary and everything downstream assumes its tag conventions, so replacing it means re-instrumenting with OpenTelemetry and Prometheus exporters and then discovering that half your dashboards grouped by a tag the new pipeline spells differently. Decide the tag scheme before the first service moves, not after the fortieth.

Then the built artefacts. Dashboards, monitors, SLOs, composite alerts and notification routing represent years of accumulated judgement, and none of it exports into another product. The pragmatic approach is to rebuild only the dashboards someone opened in the last quarter — which, in most organisations, is a startlingly short list.

Finally, the overlap has to be paid for. Running both stacks for a quarter is the only safe way to do this, and for that quarter you have two bills. Teams that try to avoid the double payment cut over during a quiet week, lose visibility during an incident, and roll back — which costs more than the overlap would have.

If Datadog has a German region, what is left to object to?

Three things, none of them geography. The first is the pricing model: per host, per product, with logs metered twice and every additional capability arriving as another line. The second is that Datadog, Inc. remains a New York company, so a disclosure order reaches it regardless of which of the nine sites your account lives in.

The third is reversibility. The site cannot be changed after the account exists, the agents are proprietary, and the dashboards, monitors and SLOs you have built are expressed in Datadog's own configuration. The longer you stay, the more expensive the next decision becomes — which is an argument for making it deliberately rather than under budget pressure.

If none of those three bothers you, staying on EU1 is a defensible choice and this page is not trying to talk you out of it.

Will an open-source stack actually be cheaper?

Cheaper in licence, not automatically cheaper in total. A self-managed Elastic cluster turns a per-host subscription into servers, storage and a person who understands shard sizing. For an estate of ten hosts that trade is usually bad. For an estate of four hundred, where Datadog infrastructure and APM alone would be over $18,000 a month, it is usually good, and the crossover sits somewhere in between depending on how noisy your logs are.

The saving that people underestimate is retention. Datadog charges $1.70 per million events for fifteen days in the index; storage you own is a disk, and the decision about how long to keep something becomes an engineering decision rather than a purchasing one.

The cost people underestimate is the operator. Someone has to own capacity planning, upgrades and the 2 a.m. cluster problem. If that person does not exist and is not being hired, the honest comparison is against Elastic Cloud in an EU region, not against a cluster you will not staff.

What happens to the integrations?

This is where Datadog is genuinely hard to replace. Installing the agent and switching on an integration is a five-minute job for hundreds of common services, and that catalogue is most of why teams adopt it in the first place.

The European answer is OpenTelemetry and Prometheus exporters, which now cover the mainstream — databases, queues, web servers, Kubernetes — and cover the long tail patchily. The difference is not usually whether the data can be collected; it is whether someone has already built the dashboard and the sensible default alerts on top of it.

Budget for that rebuilding. Two or three weeks of one engineer's time for a mid-sized estate is a realistic figure, and it is the single most under-counted line in every migration plan of this kind.

Does Datadog buying European projects matter to me?

It matters if your plan was to build on them. Quickwit, the open-source search engine that several teams were using to keep log costs down, was acquired by Datadog with the announcement on 9 January 2025; its founders were split between Paris, Tokyo and New York, and the code is being relicensed to Apache 2.0.

That relicensing is genuinely good news for anyone already running it, and it is also the pattern worth noticing: the cheap component of your European stack can become a strategic asset of the company you were trying to leave.

The defence is not to avoid open source but to prefer projects whose licence already converts on its own terms, and to check who owns the entity behind a tool before it becomes load-bearing in your architecture.

Which one to pick

If your estate is large and the per-host multiplication is what brought you here, Elastic APM self-managed is the answer that changes the mechanism rather than the supplier, and the number to work out first is your own crossover point in hosts.

If you want the store without staffing it, Elastic Cloud in an EU region keeps the same data model and moves the operational burden back to a vendor — a Dutch one, which is a different answer to the jurisdiction question than a German region of an American company.

If what you actually spend money on is synthetic checks, Checkly is a straightforward upgrade and not merely a substitution, because reusing your Playwright suite fixes the decay that makes recorded checks useless after a redesign.

And if you depend on real user monitoring, session replay and the integration catalogue together, accept that this page has no replacement for that combination. Switch to EU1 if you have not, understand that jurisdiction still follows a New York company, and spend the effort on the log indexing bill instead — that is where the money is.