Usage Based Billing

Dodo Payments Usage-Based Billing

Reference: docs.dodopayments.com/features/usage-based-billing

Charge customers for what they actually use—API calls, storage, AI tokens, or any metric you define.

When to use this skill

• You need to bill customers based on consumption (API calls, tokens, storage, bandwidth).
• You want to combine usage charges with subscriptions or one-time purchases.
• You need to track and aggregate events into billable quantities.
• You're building an AI service, SaaS platform, or infrastructure product with metered features.

Core Concepts

Events

Usage records sent from your application to Dodo. Each event is attributed to a customer and matched to a meter by its event_name.

{
  "event_id": "evt_unique_123",
  "customer_id": "cus_abc123",
  "event_name": "api.call",
  "timestamp": "2025-01-21T10:30:00Z",
  "metadata": { "endpoint": "/v1/users", "tokens": 150 }
}

Meters

Filters and aggregates events into billable quantities. A meter specifies:

• Event name: which events to match (case-sensitive)
• Aggregation type: how to combine events (count, sum, max, last)
• Measurement unit: the billing unit (calls, tokens, GB, etc.)
• Optional filters: conditions events must meet to be counted

Aggregation Types

For sum, max, and last, you specify which metadata property to aggregate.

Pricing

Attach a meter to a product price to charge per unit:

• Price per unit: e.g., $0.001 per API call
• Free threshold: e.g., 1,000 free calls per month
• Charge formula: (usage − threshold) × price_per_unit

Example: 2,500 calls − 1,000 free = 1,500 × $0.02 = $30.00

Meter Lifecycle

Create a Meter

import DodoPayments from 'dodopayments';

const client = new DodoPayments({
  bearerToken: process.env.DODO_PAYMENTS_API_KEY,
  environment: 'test_mode',
});

const meter = await client.meters.create({
  name: 'API Requests',
  event_name: 'api.call',
  aggregation: { type: 'count' },
  measurement_unit: 'calls',
  description: 'Track API calls per customer',
});

console.log(meter.id); // mtr_...

For a sum aggregation, specify the property to aggregate:

const meter = await client.meters.create({
  name: 'Token Usage',
  event_name: 'ai.tokens',
  aggregation: { type: 'sum', key: 'tokens' },
  measurement_unit: 'tokens',
});

List and Retrieve Meters

// List all meters
const meters = await client.meters.list();

// Retrieve a specific meter
const meter = await client.meters.retrieve('mtr_abc123');

Archive and Unarchive

Meters are archived, not deleted. Archived meters stop accepting new events but retain historical data.

// Archive a meter
await client.meters.archive('mtr_abc123');

// Unarchive to resume
await client.meters.unarchive('mtr_abc123');

Event Ingestion

Send Events

async function recordApiCall(
  customerId: string,
  requestId: string,
  occurredAt: string,
): Promise<number> {
  const response = await client.usageEvents.ingest({
    events: [{
      event_id: `api-call:${requestId}`,
      customer_id: customerId,
      event_name: 'api.call',
      timestamp: occurredAt,
      metadata: {
        endpoint: '/v1/users',
        method: 'GET',
      },
    }],
  });

  return response.ingested_count;
}

requestId must identify the underlying API operation and remain unchanged across retries. Do not generate it inside the ingestion attempt.

Event Schema

Idempotency and Deduplication

• Each distinct operation gets one globally unique event_id.
• Derive the ID from an immutable request, job, generation, or snapshot ID and reuse it on every retry.
• Duplicate IDs in a single request reject the entire batch.
• An ID already ingested in an earlier request is silently ignored, making retries safe.

Batch Ingestion

Send up to 1,000 events per request:

async function trackBatchUsage(
  events: Array<{
    operationId: string;
    customerId: string;
    eventName: string;
    occurredAt: string;
    metadata: Record<string, string>;
  }>
) {
  const formattedEvents = events.map((event) => ({
    event_id: `usage:${event.operationId}`,
    customer_id: event.customerId,
    event_name: event.eventName,
    timestamp: event.occurredAt,
    metadata: event.metadata,
  }));

  await client.usageEvents.ingest({ events: formattedEvents });
}

// Batch track multiple API calls
await trackBatchUsage([
  { operationId: 'req_101', customerId: 'cus_abc', eventName: 'api.call', occurredAt: '2026-08-01T10:00:00Z', metadata: { endpoint: '/v1/users' } },
  { operationId: 'req_102', customerId: 'cus_abc', eventName: 'api.call', occurredAt: '2026-08-01T10:00:01Z', metadata: { endpoint: '/v1/orders' } },
  { operationId: 'req_103', customerId: 'cus_xyz', eventName: 'api.call', occurredAt: '2026-08-01T10:00:02Z', metadata: { endpoint: '/v1/products' } },
]);

Query Events

// List events for a customer
const events = await client.usageEvents.list({
  customer_id: 'cus_abc123',
});

// Retrieve a specific event
const event = await client.usageEvents.retrieve('evt_abc123');

Pricing Models

Per-Unit Pricing

The only currently documented and operable pricing model. A meter attachment uses:

• price_per_unit: decimal string (max 5 integer digits, 12 decimal places)
• free_threshold: optional integer (usage below this is not charged)

Charge formula: (usage − threshold) × price_per_unit

The product-catalog-management skill is the canonical source for the complete product creation request. It defines the singular price object with type: 'usage_based_price' and its nested meters array; do not define a parallel product schema here.

Note: Tiered, graduated, volume, and staircase pricing models are not currently documented in the Dodo Payments API. Use per-unit pricing with free thresholds for now.

Instrumenting Your Application

Track API Calls

Persist the event before reporting the operation as complete, then ingest it from a retrying worker. The outbox or queue implementation must durably store the payload before persist resolves.

type PersistedUsageEvent = {
  event_id: string;
  customer_id: string;
  event_name: string;
  timestamp: string;
  metadata: Record<string, string | number | boolean>;
};

interface UsageOutbox {
  persist(event: PersistedUsageEvent): Promise<void>;
  nextBatch(limit: number): Promise<PersistedUsageEvent[]>;
  markIngested(eventIds: string[]): Promise<void>;
}

async function completeApiOperation(
  outbox: UsageOutbox,
  operationId: string,
  customerId: string,
  occurredAt: string,
): Promise<void> {
  await outbox.persist({
    event_id: `api-call:${operationId}`,
    customer_id: customerId,
    event_name: 'api.call',
    timestamp: occurredAt,
    metadata: { endpoint: '/v1/users', method: 'GET', status: 200 },
  });
}

async function ingestUsageOutbox(outbox: UsageOutbox): Promise<void> {
  const events = await outbox.nextBatch(1000);
  if (events.length === 0) return;

  await client.usageEvents.ingest({ events });
  await outbox.markIngested(events.map((event) => event.event_id));
}

If the worker crashes after Dodo accepts the batch but before markIngested, retry the same persisted events with the same IDs. Dodo ignores the already-ingested IDs.

Track AI Token Usage

async function callAI(
  customerId: string,
  generationId: string,
  prompt: string,
  completedAt: string,
) {
  const response = await openai.chat.completions.create({
    model: 'gpt-4',
    messages: [{ role: 'user', content: prompt }],
  });

  // Track tokens after completion
  await client.usageEvents.ingest({
    events: [{
      event_id: `generation:${generationId}`,
      customer_id: customerId,
      event_name: 'ai.tokens',
      timestamp: completedAt,
      metadata: {
        tokens: response.usage.total_tokens.toString(),
        prompt_tokens: response.usage.prompt_tokens.toString(),
        completion_tokens: response.usage.completion_tokens.toString(),
        model: 'gpt-4',
      }
    }]
  });

  return response;
}

Track Storage Usage

For snapshot-based metrics (current state), use the last aggregation:

async function updateStorageUsage(
  customerId: string,
  snapshotId: string,
  bytesUsed: number,
  capturedAt: string,
) {
  await client.usageEvents.ingest({
    events: [{
      event_id: `storage-snapshot:${snapshotId}`,
      customer_id: customerId,
      event_name: 'storage.snapshot',
      timestamp: capturedAt,
      metadata: {
        bytes: bytesUsed.toString(),
        gb: (bytesUsed / 1024 / 1024 / 1024).toFixed(2),
      }
    }]
  });
}

// Call periodically or after storage changes
await updateStorageUsage(
  'cus_abc',
  'snapshot_01K1M4D2K9',
  5368709120,
  '2026-08-01T10:30:00Z',
); // 5GB

Querying Usage for Display

Retrieve Usage History

const usage = await client.subscriptions.retrieveUsageHistory(
  'sub_abc123',
  { page_size: 100 }
);

console.log(usage.items); // Array of billing-period usage records

This returns aggregated usage per meter for the subscription's current billing period.

Credit-Based Billing Integration

Usage events can deduct from a customer's credit balance instead of charging per-unit. See the credit-based-billing skill for full details on credit entitlements, balances, and ledger management.

To link a meter to credits:

1. Create a credit entitlement (e.g., "AI Credits").
2. Attach the credit entitlement to the same product.
3. On the meter, enable Bill usage in Credits.
4. Set credit_entitlement_id and meter_units_per_credit (e.g., 1,000 tokens = 1 credit).

Usage under the free threshold is excluded. Approximately every minute, a background worker aggregates new usage, converts it using the meter-to-credit ratio, and consumes the oldest non-expired credit grants (FIFO). When credits run out, configured overage behavior applies.

Webhook Integration

Usage events trigger webhooks for monitoring and reconciliation. See the webhook-integration skill for webhook setup and verification.

Common Mistakes

1. Using Unstable or Reused Event IDs

Generate one ID from the immutable business operation. Reusing an ID for a different operation drops usage, while generating a timestamp or random ID on every retry can bill the same operation twice.

// WRONG — a retry creates a new billable event
await client.usageEvents.ingest({
  events: [{
    event_id: `api-call:${Date.now()}:${crypto.randomUUID()}`,
    customer_id: 'cus_abc',
    event_name: 'api.call',
  }],
});

// CORRECT — retry request req_123 with this same ID
await client.usageEvents.ingest({
  events: [{
    event_id: 'api-call:req_123',
    customer_id: 'cus_abc',
    event_name: 'api.call',
  }],
});

Do not reuse api-call:req_123 for a distinct request. This matches credit ledger guidance: a timeout is not permission to generate a fresh idempotency key.

2. Using Fire-and-Forget Ingestion

Do not start ingestion after responding without first persisting the event. The process can crash after the user receives success but before usage reaches Dodo.

// WRONG — an acknowledged operation can lose its usage event
app.post('/api/generate', async (req, res) => {
  const result = await generateAI(req.body);
  res.json(result);
  void client.usageEvents.ingest({ events: [result.usageEvent] });
});

// CORRECT — durable persistence completes before success is returned
app.post('/api/generate', async (req, res) => {
  const result = await generateAI(req.body);
  await usageOutbox.persist(result.usageEvent);
  res.json(result);
});

A retrying worker ingests the persisted payload with its original event_id, as shown in Track API Calls. If persistence fails, return an error so the operation can be retried rather than silently underbilling.

3. Clock Skew in Timestamps

Timestamps must be within one hour in the past or five minutes in the future. Ensure your server clock is synchronized.

// WRONG — timestamp is 2 hours old
const oldTime = new Date(Date.now() - 2 * 60 * 60 * 1000).toISOString();
await client.usageEvents.ingest({
  events: [{
    event_id: 'evt_123',
    customer_id: 'cus_abc',
    event_name: 'api.call',
    timestamp: oldTime, // Rejected
    metadata: {}
  }]
});

// CORRECT — use current time
await client.usageEvents.ingest({
  events: [{
    event_id: 'evt_123',
    customer_id: 'cus_abc',
    event_name: 'api.call',
    timestamp: new Date().toISOString(),
    metadata: {}
  }]
});

4. Ingesting on the Client Side

Never send events from client-side code. Always ingest from your backend to avoid exposing your API key.

// WRONG — client-side
const trackUsage = async (eventName: string) => {
  await fetch('https://test.dodopayments.com/events/ingest', {
    method: 'POST',
    headers: {
      'Authorization': `Bearer ${process.env.DODO_PAYMENTS_API_KEY}`, // Exposed!
      'Content-Type': 'application/json'
    },
    body: JSON.stringify({ events: [...] })
  });
};

// CORRECT — send the event from a backend worker using its persisted payload
await client.usageEvents.ingest({
  events: [{
    event_id: 'api-call:req_123',
    customer_id: 'cus_abc',
    event_name: 'api.call',
    timestamp: '2026-08-01T10:30:00Z',
    metadata: { endpoint: '/v1/users' },
  }],
});

5. Mismatched Event Names

Event names are case-sensitive and must match the meter's event name exactly.

// WRONG — meter expects "api.call", event sends "API.CALL"
const meter = await client.meters.create({
  name: 'API Requests',
  event_name: 'api.call',
  aggregation: { type: 'count' },
  measurement_unit: 'calls',
});

await client.usageEvents.ingest({
  events: [{
    event_id: 'api-call:req_123',
    customer_id: 'cus_abc',
    event_name: 'API.CALL', // Won't match
  }]
});

// CORRECT
await client.usageEvents.ingest({
  events: [{
    event_id: 'api-call:req_123',
    customer_id: 'cus_abc',
    event_name: 'api.call', // Matches exactly
  }]
});

Resources

• Usage-Based Billing Guide
• Meters Documentation
• Event Ingestion API
• Create Meter API
• Usage-Based Billing Integration Guide
• Credit-Based Billing
• Webhook Integration