Cloudflare Docs
Analytics
Analytics
Visit Analytics on GitHub
Set theme to dark (⇧+D)

Sampling

​​ Overview

In a small number of cases, the analytics provided on the Cloudflare dashboard and GraphQL Analytics API are based on a sample — a subset of the dataset. In these cases, Cloudflare Analytics returns an estimate derived from the sampled value. For example, suppose that during an attack the sampling rate is 10% and 5,000 events are sampled. Cloudflare will estimate 50,000 total events (5,000 × 10) and report this value in Analytics.

​​ Sampled datasets

Cloudflare Analytics builds the following datasets from sampled data:

Data setNodes
Firewall Activity LogfirewallEventsAdaptive firewallEventsAdaptiveByTimeGroups
Firewall AnalyticsfirewallEventsAdaptiveGroups
Firewall Rule PreviewfirewallRulePreviewGroups
Network AnalyticsipFlows1mGroups*
ipFlows1hGroups*
ipFlows1dGroups*
ipFlows1mAttacksGroups*
Network Analytics v2
for Magic Transit customers
magicTransitNetworkAnalyticsAdaptiveGroups
dosdNetworkAnalyticsAdaptiveGroups
dosdAttackAnalyticsGroups
flowtrackdNetworkAnalyticsAdaptiveGroups
magicFirewallNetworkAnalyticsAdaptiveGroups
Network Analytics v2
for Spectrum customers
(Enterprise plans only)
spectrumNetworkAnalyticsAdaptiveGroups
dosdNetworkAnalyticsAdaptiveGroups
dosdAttackAnalyticsGroups
Workers MetricsworkersInvocationsAdaptive
Magic Firewall AnalyticsmagicFirewallSamplesAdaptiveGroups

* These nodes are deprecated. Refer to Deprecated data nodes for more information.

The presence of sampled data is called out in the Cloudflare dashboard and in the description of the dataset in the API.

​​ Why sampling is applied

Analytics is designed to provide requested data, at the appropriate level of detail, as quickly as possible. Sampling allows Cloudflare to deliver analytics within seconds, even when datasets scale quickly and unpredictably, such as a burst of Firewall events generated during an attack. And because the volume of underlying data is large, the value estimated from the sample should still be statistically significant – meaning you can rely on sampled data with a high degree of confidence. Without sampling, it might take several minutes or longer to answer a query — a long time to wait when validating mitigation efforts.

​​ Types of sampling

​​ Adaptive sampling

Cloudflare almost always uses adaptive sampling, which means the sample rate fluctuates depending on the volume of data ingested or queried. If the number of records is relatively small, sampling is not used. However, as the volume of records grows larger, progressively lower sample rates are applied. Firewall Analytics and the Firewall Event Log follow this model. Data nodes that use adaptive sampling are easy to identify by the Adaptive suffix in the node name, as in firewallEventsAdaptive.

​​ Fixed sampling

The following data nodes are based on fixed sampling, where the sample rate does not vary:

Data setRateNotes
Firewall Rules Preview

Nodes:
firewallRulePreviewGroups

1%Use with caution. A 1% sample rate does not provide accurate estimates for datasets smaller than a certain threshold, a scenario the Cloudflare dashboard calls out explicitly but the API does not.
Network Analytics

Nodes:
ipFlows1mGroups
ipFlows1hGroups
ipFlows1dGroups
ipFlows1mAttacksGroups

0.012%Sampling rate is in terms of packet count (1 of every 8,192 packets).
Network Analytics v2

Nodes:
dosdNetworkAnalyticsAdaptiveGroups
dosdAttackAnalyticsGroups
flowtrackdNetworkAnalyticsAdaptiveGroups
magicFirewallNetworkAnalyticsAdaptiveGroups
magicTransitNetworkAnalyticsAdaptiveGroups
spectrumNetworkAnalyticsAdaptiveGroups

Depends on the node.Sample rate is 1/100 for Magic Firewall, 1/10,000 for dosd, and dynamic for flowtrackd. Refer to the NAv2 migration guide for details.

​​ Access to raw data

Because sampling is primarily adaptive and automatically adjusts to provide an accurate estimate, the sampling rate cannot be directly controlled. Enterprise customers have access to raw data via Cloudflare Logs.