Backend Development › Files & Media · also in Product Building Blocks

Data Export

Letting users download their data, often as a background job.

Also known as: data export, data download, export my data

A data export lets users (or admins) download their data — all their records, a report, a full account archive. It’s a common product feature and often a legal requirement (data portability under GDPR). The hard part isn’t the button; it’s doing it for data that’s too large to return in one HTTP response.

user requests export → job generates file → store in object storage → email/serve a link

The standard pattern: the request enqueues a background job that gathers the data, writes it to a file (CSV, JSON, archive) in object storage, and notifies the user with a time-limited download link (see presigned URL). For small exports, it can be synchronous — but “small” is a trap.

The classic mistakes:

• Generating large exports synchronously. Building a big file in the request buffers it all in memory, blocks a worker, and may time out. Use an async job and stream the file.
• Loading everything into memory. Even in a job, holding millions of rows in memory can OOM. Stream rows to the file (see streaming large files).
• Serving the file from the app forever. Large files should live in object storage and be served directly (or via a signed URL), not proxied through app servers.
• No expiry on the link. An export link is a data-access credential; a permanent public link can leak personal data. Make download links time-limited and access-controlled.
• Exporting more than asked. “Export everything” can include other users’ data, secrets, or internal fields; scope exports to the requester’s data and exclude sensitive internals.
• No notification for async exports. If the user has to guess when it’s ready, the feature feels broken. Notify (email/in-app) when the file is ready.
• Forgetting cleanup and cost. Stale exports accumulate in storage and cost money; set retention to delete old export files (see storage lifecycle).
• Ignoring format and encoding. CSV quoting, timezones, Unicode, and large fields trip up naive exporters; test with real data.

How to build it: request → async job → stream the scoped data to a file in object storage → notify with a time-limited link → expire and clean up. It’s a standard background-job-and-storage pattern with real privacy weight — scope narrowly, secure the link, and bound retention. See data migration for the related “import” side.