Federated analytics computes aggregate statistics across many devices without collecting the underlying individual data centrally. Each device computes locally and contributes only to an aggregate.
What it is
Derived from federated learning, the approach suits engagement measurement well: the useful output is usually 'what proportion of learners disengaged at section 4', which can be computed from local contributions without any central record of individual sessions.
Differential privacy techniques can be layered on to bound what an aggregate reveals about any contributor.
Why it matters
It offers a route to content-level insight with materially less personal data processing — often none at all, if aggregation happens before transmission.
A common confusion
Aggregating after central collection is not federated analytics, and does not carry the same benefit. If individual data reached your server, you processed it.
Related
See also client-side inference, air-gapped deployment and dpia (data protection impact assessment). See the full glossary for the rest, or our security and data handling page for the wider context.
See where your content loses people
Book a walkthrough and we will show you the engagement data on your own content.