Why is my subscription's current usage higher than expected?

If a billable metric's charge has filters configured (also called pricing group keys), Lago calculates usage separately for each unique combination of filter values instead of pooling it. A negative adjustment event only reduces usage inside its own bucket, so if current usage looks too high, the likely cause is that a negative event landed in a different filter bucket than the positive event it was meant to offset.

How filtered usage works

When a charge has filters, for example user_id and workspace_id, each unique combination of those filter values becomes its own independent usage bucket, aggregated and billed separately. A bucket can never go below zero: if a bucket's total would be negative, Lago floors it to zero, and the unused negative amount is not carried over to any other bucket.

For example, a charge filtered on user_id and workspace_id that receives an initial provisioning event of +12 units with no filter values, an additional seat event of +1 unit filtered by user_id only, and five adjustment events of −1 unit each filtered by both user_id and workspace_id, produces three separate buckets. Lago reports 13 total units, not 8, because the −5 bucket floors to zero and has no effect on the other two buckets.

How to confirm the cause

1. Check which filters are configured on the relevant charge, in the billable metric's charge settings.

2. Compare the filter values on the negative adjustment events against the filter values on the positive events they are meant to offset. A negative event with {user_id: X, workspace_id: Y} only reduces usage for events sent with that exact combination. If the original event had no filter values, the adjustment needs no filter values either.

How to resolve the issue

Align filter values between positive and negative events going forward. For usage already ingested with mismatched filter values, the overcounted amount for the current billing period can be corrected with a credit note. Contact support if you need help with this.

This article reflects guidance drawn from customer case resolutions. It is not officially supported documentation and may not apply to all situations.