TL:DR
Imagine 47 million requests hitting your platform last month. Can you prove who made each one—and invoice with confidence?
If that question tightens your stomach, you're not alone. Metered billing for APIs promises fair, transparent pricing that scales with customer success. But it only works when your measurements are trustworthy, replayable, and finance-grade.
Miscount by even a fraction and you can leak revenue. Or worse—you lose customer trust.
The reality? Counting requests alone often isn't sufficient for modern API businesses. You need billing-grade telemetry that withstands financial scrutiny. This guide reveals the architecture behind bulletproof metered billing models: atomic usage events, idempotency, real-time aggregation, and robust event pipelines
Metered billing for APIs charges customers based on actual consumption—requests made, data processed, or compute time used—rather than flat subscription fees. Think of it like your electricity bill. You pay for kilowatt-hours consumed, not a flat rate regardless of usage.
This model directly aligns value with price. Customers pay for what they use. Nothing more, nothing less.
Three primary models dominate API monetization today:Three primary models dominate API monetization today: Subscription, Metered and Hybrid.
Subscription models offer predictability but risk alienating low-usage customers. Pure metered models scale perfectly with usage but complicate budgeting. Hybrid approaches balance both needs.
Consider real-world examples. Stripe charges a per-transaction fee depending on the payment method used, typically 2.9% + $0.30 per successful charge for most online card payments in the U.S., though rates vary by payment type and region. OpenAI primarily bills per token for its language models, though pricing varies by model type and includes other billing units for different services. GitHub recently shifted certain enterprise plans to pay-as-you-go billing, where eligible customers on specific tiers pay for licenses consumed at month's end rather than pre-purchasing.
Three forces drive this transformation:
Transparency and Fairness
Users pay only for what they use, boosting loyalty as customers can easily adjust usage up or down. Small customers aren't priced out. Large customers pay their fair share.
Revenue Optimization
Metered billing creates built-in expansion revenue—customer growth directly translates to higher spend. Sales teams focus on customer success rather than pushing bigger plans.
Market Expansion
Breaking pricing into increments opens the addressable market. This proves especially relevant for AI and SaaS startups expanding internationally.
The numbers support this shift. The cloud billing market size was estimated at $12.78 billion in 2024 and is projected to grow to $41.3 billion by 2035, exhibiting a compound annual growth rate of 11.25%, according to https://www.marketresearchfuture.com/reports/cloud-billing-market-1557Market Research Future analysis. In the broader SaaS landscape, 85% of surveyed companies either already had usage-based pricing or were planning to adopt it, with 78% of companies with UBP adopting it within the last five years.
Building reliable metered billing requires four essential layers. Each builds upon the foundation below it.
Usage events form your system's foundation. These immutable, append-only records capture every billable action.
What makes a good usage event?
Here's the critical principle: If you can't replay it, you can't trust it.
Your event store must allow complete reconstruction of any invoice from raw events. This provides an unassailable audit trail for disputes or corrections.
Raw events need transformation into billable metrics. Meters handle this aggregation.
Modern metering engines transform usage into "metered features" customized to your business context. An AI voice translation service might track audio duration and language complexity, but bill based on minutes translated or conversations processed.
Common aggregation patterns include:
Time windows matter. Many B2B services bill monthly, though requirements vary:
Your rating engine applies business logic to aggregated usage, transforming raw quantities into dollar amounts.
Common pricing models:
Flat Rate
Tiered Pricing
Volume Discounts
Credit-Based
The final layer generates customer invoices and handles payment. Critical considerations include:
Under ASC 606, revenue is recognized when the customer gains control of the promised goods or services. For usage-based models with a "stand-ready obligation"—the promise to be available on demand—revenue recognition often occurs as usage happens (Cloud Billing Market Size, Share | Growth Report 2035), though base access fees might be recognized straight-line over the period. The specific timing depends on contract terms, performance obligations, and your accounting policies.
There's no universal answer for where to measure. Your architecture dictates the best approach.There's no universal answer for where to measure. Your architecture dictates the best approach.
API gateways offer a natural measurement point.
Advantages:
Limitations:
Tesla's Fleet Telemetry demonstrates one approach: applications receive data they're interested in, vehicles send data when awake and connected, and signals are sent when values change. The specific billing implications depend on individual API configurations and contract terms.
Measuring within applications provides the richest context.
Advantages:
Limitations:
The emerging standard combines both approaches through an event pipeline.
Chargebee reports their system processes up to 200,000 events per second when tracking API calls and AI token usage—though this likely represents peak capacity rather than sustained throughput. Such scale demands dedicated infrastructure.
Architecture Benefits:
Stream usage events to Kafka, Kinesis, or platforms like OpenMeter (Now Konnect Metering & Billing). Use CloudEvents for standardization. This approach provides flexibility, resilience, and auditability.
The difference between "good enough" and billing-grade comes down to edge cases. These details directly impact revenue and trust.
Idempotency prevents double-charging from retries. It's non-negotiable for billing systems.
Stripe emphasizes: "Use idempotency keys to prevent reporting usage for each event more than one time because of latency or other issues—every meter event corresponds to an identifier that you can specify in your request."
Implementation requires three steps:
Common approaches include monotonic sequence numbers per device and idempotency keys derived from device ID + timestamp + record type, with backends storing these keys to reject duplicates, allowing safe resends during unstable connectivity.
With the power of Kong + OpenMeter, Konnect's Metering & Billing you can automate this three-step process at the infrastructure level. By mapping the gateway’s unique request IDs to the CloudEvent ID field, you ensure that even if a network hiccup causes a dual submission, the metering engine performs a final deduplication check against its stateful window. This "Ingress-to-Invoice" alignment ensures that the usage reported to Stripe is mathematically guaranteed to be unique, fulfilling the non-negotiable requirement of billing accuracy without requiring custom deduplication logic in your application code.
Real systems must handle out-of-order and delayed events.
Acceptance Windows
Clock Synchronization
Common problems include API retries creating duplicate events that can lead to overbilling and usage near midnight getting recorded in the wrong month. Solutions include idempotency keys and UTC standardization with edge case testing around period close.
Disputes happen. Design for them upfront.
Correction Mechanisms:
Best Practices:
Conflating rate limiting with billing measurement is a dangerous mistake. They serve fundamentally different purposes.
Cloud providers explicitly acknowledge this distinction:
The key insight? Rate limiting decisions and billing measurements operate independently. Whether blocked requests are billable depends on your specific product terms and pricing model. Separate enforcement from accounting.
As usage grows, consider these advanced patterns:As usage grows, consider these advanced patterns:
As your API ecosystem expands, scaling your billing infrastructure requires moving beyond simple logging to a more sophisticated, distributed architecture. Using Kong Konnect Metering & Billing allows you to offload this complexity to the control plane while maintaining high-performance data planes.
Modern billing systems must balance the need for immediate user feedback with the absolute precision required for financial settlement. High-volume environments often separate these concerns:
For global APIs, usage data must be collected as close to the user as possible to avoid latency, then reconciled centrally for billing.
Scale brings the risk of "telemetry tax"—where the cost of monitoring usage rivals the cost of the service itself.
What is metered billing for APIs? Metered billing charges customers based on actual consumption—such as the number of API calls, data throughput, or specific AI tokens—rather than a flat monthly fee. This "pay-as-you-go" model aligns the customer's costs directly with the value they derive from your services.
How do I track API usage accurately for billing? Tracking requires an event-driven architecture that captures usage at the point of ingestion. By using Konnect Metering and Billing, you can automatically transform API traffic into verifiable usage events. This ensures that every request is logged with a stable timestamp and a unique ID to maintain a permanent, audit-ready record.
What's the difference between API rate limiting and metered billing? Rate limiting is a protective measure that throttles traffic in real-time to prevent service degradation. Metered billing is a financial process that aggregates that same traffic over a billing cycle to generate an invoice. While rate limiting says "no" to excess traffic, metered billing says "yes, and here is the cost."
What are billing-grade telemetry requirements? To be "billing-grade," telemetry must be idempotent, resilient to network failures, and fully auditable. This involves implementing strict deduplication, a defined acceptance window for late-arriving data, and the ability to replay events if a downstream billing provider like Stripe or Lago experiences an outage.
Where should I measure usage—gateway or application? It depends on the metric. API Gateways are ideal for measuring network-level usage (requests, bandwidth, or latency). However, for domain-specific metrics—like "messages sent" or "compute minutes"—the application is the better source. Konnect allows you to unify both sources by acting as the central collector for gateway-native metrics and custom application events.
How do I handle late or duplicated usage events? Deduplication is handled via idempotency keys (often derived from the Request ID or a custom header) that ensure an event is only counted once. For late data, you must establish an "acceptance window"—typically 24 to 72 hours—where the system can ingest delayed events and back-fill the usage meters before the final invoice is cut.
What is CloudEvents and why does it matter for metered billing? CloudEvents is an industry-standard specification for describing event data. It provides a consistent "envelope" for metadata like the event source and timestamp. By adopting this standard, tools like OpenMeter (integrated into Konnect) can seamlessly ingest data from different parts of your stack while maintaining a uniform audit trail.
Remember those 47 million illustrative requests? With billing-grade telemetry in place, you can now confidently state: "Yes, we can prove every single one. Here's the invoice, backed by an immutable audit trail."
Metered billing for APIs isn't just about counting requests; it’s about building a trust system that accurately captures the value exchange between you and your customers. Data shows this pays off: companies using hybrid models (subscription + usage) report a 21% median growth rate, outperforming both pure subscription and pure usage-based models.
By leveraging Kong Konnect Metering & Billing, you ensure that accurate, auditable data is baked into your infrastructure. This builds customer trust, unlocks flexible consumption-based pricing, and eliminates revenue leakage through automated reconciliation.
The path forward is clear:
Ready to turn your API traffic into revenue?
Stop guessing your usage and start metering with confidence. See how the integration of Kong and OpenMeter provides a seamless, "Ingress-to-Invoice" solution for your platform.
The Shifting Economic Landscape: The AI token economy in 2026 is evolving, and enterprise leaders must distinguish between low-cost input tokens and high-premium output tokens to maintain profitability. Agentic AI Financial Risks: The transition t
The challenges of an acquisition frequently appear in a number of critical areas, especially when dealing with a platform as important as Kafka: API Instability and Change : Merged entities frequently rationalize or re-architect their services, whic
Managed Redis cache is a turnkey "Shared State" add-on for Kong Dedicated Cloud Gateways. It is designed to combine the performance of an in-memory data store with the simplicity of a SaaS product. When you spin up a Dedicated Cloud Gateway in Kong
We've talked about why 2026 is the year of AI unit economics . There, we explored the "2025 hangover" where organizations realized that without financial governance, AI isn't just a science project but has become a margin-bleeding cost center. But
Modern organizations rely on APIs to power their digital customer experiences. This can lead to stronger brand loyalty and higher revenues — if they play their cards right. The driving factor in delivering personalized content is connectivity to mor
