Amplitude

Attribution reads your Amplitude data from an Amazon S3 export. You make sure a page view event is tracked, point Amplitude's S3 destination at a bucket you own, grant Attribution read access to that bucket, and our team connects the import on our side.

Make sure page views are tracked

Attribution builds its model from page views with a URL and a referrer. Most Amplitude installs already log such an event; check that yours does. If it does not, log one after the Amplitude SDK initialises:

amplitude.getInstance().logEvent('Loaded a Page', {
  url: window.location.href,
  referrer: document.referrer
});

The event must fire on every page load, or at least on every landing page, on each domain you run ads to.

Export Amplitude to S3

1. Create an S3 bucket in your AWS account.
2. In Amplitude, add the Amazon S3 destination and point it at that bucket.
3. When Amplitude asks which data to include, select Export events ingested today and moving forward and leave the event filter empty, so every event reaches the export. Amplitude writes a file about once an hour.

Give Attribution access to the bucket

1. Open the bucket in the AWS S3 console.

2. Open the Permissions tab.

3. Under Bucket policy, click Edit.

4. Add this statement to the Statement array, next to the statement Amplitude's setup already added. Replace bucketName with the name of your bucket in both resource lines. The import needs write access as well as read: it marks each file it has loaded so nothing is imported twice.

   JSON

   {
     "Sid": "Stmt1441164338000",
     "Effect": "Allow",
     "Action": [
       "s3:*"
     ],
     "Resource": [
       "arn:aws:s3:::bucketName",
       "arn:aws:s3:::bucketName/*"
     ],
     "Principal": {
       "AWS": [
         "arn:aws:iam::348454833122:root"
       ]
     }
   },

5. Click Save changes.

Tell us the bucket name

Send the bucket name to [email protected]. We configure the import on our side and confirm once data appears in your project.