Automated usage metering is the pipeline that turns what customers consume in your product into usage records a cloud marketplace accepts and bills, with no export, re-keying or re-sending by hand. On a marketplace it has one job a spreadsheet cannot do: meet each marketplace’s reporting window, every hour, for every customer.
Usage-based revenue usually lives in four places at once: the product that counts the usage, the billing system that prices it, the marketplace that invoices the buyer, and the ledger finance closes the month from. Each keeps its own copy. When the copies disagree, it is rarely a pricing problem. It is a timing problem: a record sent late, sent twice, or never sent at all.
Automation fixes that only when it is built around the marketplaces’ rules rather than your billing calendar. Those rules are short, strict and different on each cloud.
What does automated usage metering have to do?
Automated usage metering has to do four things on every cycle, for every active buyer:
- Collect usage from the system that counts it, attributed to the right customer and the right agreement.
- Aggregate it into the dimensions the listing was published with. A record for a dimension the buyer’s plan does not have is rejected; Microsoft returns
InvalidDimensionfor it. - Report it inside the marketplace’s window, exactly once.
- Reconcile what the marketplace accepted against what the product counted, and act on the difference.
Steps 1 and 4 are where the silos are. Steps 2 and 3 are where the marketplace rules are, and they are what a monthly batch job gets wrong. Choosing the dimensions themselves is a decision you make once, before you publish; designing metering dimensions covers it.
How late can usage be reported on each cloud?
Every major marketplace closes its window within a day of the usage, and Google Cloud’s closes within an hour. The rules, from each provider’s own documentation:
- AWS Marketplace (SaaS). Usage goes to
BatchMeterUsage, up to 25 records and under 1 MB per request. Records aren’t accepted 24 hours or more after the event, and the previous month’s records are accepted only until 06:00 UTC on the 1st. Identical requests are idempotent, so a retry with the same records is safe. New SaaS products from June 1, 2026 must identify the buyer byCustomerAWSAccountIdand the agreement byLicenseArn, because a buyer can now hold concurrent agreements for one product. - Microsoft commercial marketplace. The metering service takes one usage event per resource, dimension and hour of the day. Units consumed within the hour are summed into that single event, events can be sent for the past 24 hours only, and a second event for the same hour is rejected with
409 Conflict. The batch API takes up to 25 events. - Google Cloud Marketplace. Providers must report usage within one hour of it being generated. At month end, usage reported after 1 AM Pacific Time the following day might not reach that month’s invoice.
Two consequences follow. A pipeline that reports once a day is already out of bounds on Google Cloud and at the edge everywhere else, so report hourly, or as close to the event as you can. And the month boundary is a hard line: usage discovered on the 2nd for the 31st is past every cloud’s window for that month.
Where automated metering usually breaks
- A nightly job. Usage counted once a day and sent once a day misses Google Cloud’s window outright and leaves no margin for a failed run anywhere.
- A retry that changes the record. On AWS, resending the same records is safe; regenerating them with a new timestamp is not an identical request, and can bill the same usage twice. On Microsoft, two events for the same hour collide, so sum within the hour before sending. The AWS error codes and batching strategy are in when usage records fail.
- Dimension drift. The product starts counting something the listing never published, and those records are rejected until the listing or the mapping catches up.
- The wrong agreement. With AWS concurrent agreements, a report that names only the product and the customer account no longer tells AWS which agreement the usage belongs to.
- A rejection nobody sees. A record your pipeline sent can still be refused by the marketplace. If nobody owns that queue, the revenue stays in the product and never reaches an invoice. Fixing usage that was already billed is a separate process, compared in marketplace usage correction.
How Fours handles this
Fours sits between your usage source and the marketplaces. Your service sends usage records for an active entitlement on AWS, Azure, Google Cloud or Oracle to the Fours Metering API, or Fours picks them up through usage-sync integrations such as Orb, Metronome and Chargebee. Each record group is validated on arrival: an inactive entitlement, an unknown dimension or a negative quantity is refused with a 400, and so is an id already received for that entitlement in the last 15 days. Every hour Fours sums each entitlement’s unreported usage per dimension and sends one report per entitlement in that marketplace’s format. On AWS it keeps a separate entitlement for each concurrent agreement and adds the agreement’s LicenseArn itself. A report the marketplace rejects shows as Report Failed in the console, with the cause and the next step, and dimension conversion maps your own dimension names onto the listing’s with a multiplier.
What it leaves to you: counting usage in your product and sending it promptly. Fours accepts late usage, but on AWS and Azure it reports it in the hour it is sent, which can put it on the next month’s bill, so the marketplace windows above still set your deadline. AMI and container products are outside it entirely; AWS requires those to meter from inside the instance or container through MeterUsage. And if you use AWS concurrent agreements, report through the Fours API rather than the Orb or Metronome sync, which do not yet tell a buyer’s concurrent agreements apart.
The commercial side, from usage pricing to invoices, is on the billing and metering page.
Frequently asked questions
What is automated usage metering? Automated usage metering is the pipeline that collects what customers consume in your product, aggregates it into a listing’s pricing dimensions and reports it to the marketplace that bills the buyer, with no manual export or re-entry in between.
How late can usage be reported to a cloud marketplace? AWS does not accept SaaS usage records 24 hours or more after the event. Microsoft accepts usage events for the past 24 hours. Google Cloud asks providers to report usage within one hour of it being generated.
Can the same usage be billed twice? Microsoft rejects a second usage event for the same resource, dimension and hour with a 409 Conflict. AWS treats identical BatchMeterUsage requests as idempotent, so retrying with the same records is safe.
Sources
Primary sources for the platform rules cited above. Last verified October 28, 2026. Cloud providers change fees, eligibility, and program terms without notice — check the source before relying on a figure.
- AWS Marketplace API Reference: BatchMeterUsage — Records not accepted 24 hours or more after the event; previous month until 06:00 UTC on the 1st; idempotent identical requests; 25 records and 1 MB per request; LicenseArn and CustomerAWSAccountId for new SaaS products from June 1, 2026
- Microsoft Learn: Marketplace metered billing APIs — One usage event per resource, dimension and hour; past 24 hours only; 409 Conflict on a duplicate; InvalidDimension; batch of up to 25 events
- Google Cloud Marketplace: Best practices for usage reporting — Report usage within one hour of it being generated; month-end cutoff of 1 AM Pacific Time the following day
- Fours docs: Usage Metering — Fours behaviour only: validation rules, 15-day duplicate id check, the hourly per-entitlement report, late-usage handling, LicenseArn for concurrent agreements, Orb, Metronome and Chargebee syncs, dimension conversion, AMI and container exclusion
Keep reading
Stay Updated
New posts, product updates and marketplace strategy are shared on LinkedIn as they publish.
Follow Fours on LinkedIn