Skip to content
Neruba
Usage pricing, credits, and subscriptions
OverviewWhat Neruba owns, who it fits, and what to inspect first.CapabilitiesInspect ingest, billing runs, balances, payments, and operator workflows.
Compare
Pricing
DocsJump to quickstart, examples, operations, and rollout guides.API examplesCopy auth, setup, ingest, credits, and billing-read request flows.ImplementationMap rollout sequencing, migration work, and launch readiness.Engineering NotesUse the technical lesson sequence when the team needs the patterns first.
Trust
Plan your rollout
← Back to all engineering posts
6 min read

Usage aggregation windows (handling late events)

A strong aggregation model keeps two promises at once: finance gets stable invoice periods, and customers still receive fair corrections when delayed usage appears later.

On this page
Window diagramUse a customer-safe modelDecide your late-events policy

Start with the right mental model

Raw usage is durable history. Rollups, invoice-period totals, and reconciliation views are derived outputs. That distinction is what makes delayed events fixable without rewriting the past.

Inline engineering diagram
At a glance

How usage totals stay correct when events arrive late

Invoice correctness comes from keeping raw usage as immutable history and treating period totals as derived outputs that can be recalculated when reality changes.

Step 1
Raw usage history

Ingest immutable events with tenant, meter, source id, effective time, and ingestion time.

Step 2
Window policy

Aggregate by day, week, or month with a clear late-arrival buffer before treating totals as final.

Step 3
Rebuild derived totals

Recompute invoice-period usage deterministically when backfills or delayed events land.

Step 4
Visible evidence

Keep reconciliation artifacts per period so support and finance can see why a number changed.

The goal is not to finalise as fast as possible. The goal is to finalise with confidence, leave room for late arrivals, and preserve a clear correction trail when the totals change later.

Use a model that stays fair after the invoice closes

  • Store raw usage events immutably.
  • Aggregate into billing windows with a short late-arrival buffer before treating totals as final.
  • Allow deterministic recomputation when delayed events arrive.
  • Keep reconciliation artifacts for every invoice period.

Decide the correction policy before operations need it

Decide what happens when an event arrives after an invoice has closed: credit note, next-invoice adjustment, or a controlled reopen. The implementation should enforce that policy instead of inventing one during an incident.

Previous article
Webhook reliability: at-least-once delivery in practice
A webhook path that stays calm when providers retry, reorder, or redeliver the same event more than once.

Keep reading on the site, or start the guided email sequence if you want the same lessons delivered in order.

Plan your rolloutStart technical briefings
On this page
Window diagramUse a customer-safe modelDecide your late-events policy
Plan your rolloutStart technical briefings
Built for product, finance, and security teams

Ready to move from review into a concrete rollout conversation?

Use the platform, docs, trust, and implementation pages to get the right people aligned. When the project becomes active, share your pricing model, deployment posture, and migration constraints so the reply starts with your environment.

Plan your rolloutImplementation guide
Technical briefingsNeruba Engineering Notes
Neruba
Usage pricing, credits, and subscriptions

Usage ingest, ledger-backed billing, and operator-ready recovery for teams that need the money model to stay explainable.

© 2026 AspectSoft
Product
OverviewCapabilitiesSolutionsBuying paths
Developers
QuickstartImplementationDocsAPI examplesOperationsBlog
Trust
Trust CenterSecurityPrivacyStripe comparison