Every record is synthetic. This read-only example runs through the deterministic scan rules and shows their actual output.

Synthetic example · Read-only demo

2026-08-01 to 2026-09-01

2 differences to review.

Start with the comparison below. See what each source recorded, understand the difference and check the suggested next step. A difference alone does not prove financial loss.

Fictional test data · Use this report to practice reviewing evidence.

Understand the result

openai · USD · selected comparable provider cost

Reported spend

USD 2,455.75

Observed · eligible provider buckets

Attributed spend

USD 1,840.50

Strong · confirmed exclusive projects

Without a customer link

USD 615.25

Observed coverage gap · review identity

Attribution coverage

74.94%

Attributed / comparable reported spend

What needs your attention

1 usage comparisons supported by the data

Calculated · usage

Usage that doesn't reconcile

380,000 tokens

Your AI provider and your product recorded different usage for the same scope, period and unit.

AI provider recorded

9,020,000 tokens

Your product recorded

8,640,000 tokens

What to check next

Check whether requests are missing from your usage records, whether retries were recorded consistently and whether both sources are complete.

See the calculation & evidence

The calculation

9,020,000 provider − 8,640,000 internal = 380,000 tokens

Next review

Compare the listed ledger events with the provider bucket and verify metering semantics before diagnosing a defect.

What supports this result

Same exact unit, project/model and half-open UTC interval; exclusive confirmed mapping; reviewed closed sources after operator-specified lag

bucket_asterevt_aster_usage

Limits to keep in view

Quantity difference only; no monetary conversion, estimated loss or revenue inference.
Retries, batching, caching, omitted events or differing metering semantics may explain the difference.
Operator closure and matching unit labels are necessary but do not independently prove instrumentation completeness.

Inspect original calculation fields

provider_usage_quantity - sum(internal_usage_quantity)

{
  "bucketId": "bucket_aster",
  "ledgerRecordIds": [
    "evt_aster_usage"
  ],
  "mappingId": "mapping_aster",
  "internalCustomerExternalId": "workspace_aster",
  "provider": "openai",
  "organizationId": "org_sample",
  "projectId": "proj_aster",
  "model": "model_sample",
  "windowStart": "2026-08-01T00:00:00.000Z",
  "windowEnd": "2026-09-01T00:00:00.000Z",
  "unit": "tokens",
  "providerQuantity": "9020000",
  "ledgerQuantity": "8640000",
  "delta": "380000",
  "eligibility": "Same exact unit, project/model and half-open UTC interval; exclusive confirmed mapping; reviewed closed sources after operator-specified lag"
}

Observed · coverage

Spend without a customer link

USD 615.25

The provider reported this cost, but the data do not reliably show which customer it belongs to. Shared project usage may explain it.

What to check next

Check who used the provider project during this period. Confirm a customer link only if the project was exclusive to that customer; otherwise keep it as shared spend.

See the calculation & evidence

The calculation

2455.75 reported − 1840.5 attributed = 615.25 USD without a customer link

Next review

Review missing or shared project identity and confirm customer ownership with source evidence.

What supports this result

Closed source review; nonnegative reported cost; no adjustment, conflicting version or overlapping cost bucket

bucket_shared

Limits to keep in view

Unattributed spend is a coverage gap, not proven loss or waste.
Denominator includes only eligible rows from the selected provider report; excluded rows and adjustments remain in Data Quality.
Source closure is explicitly reviewed by the operator, not independently verified by a connector.

Inspect original calculation fields

comparable_reported_provider_cost - strongly_attributed_provider_cost

{
  "provider": "openai",
  "windowStart": "2026-08-01T00:00:00.000Z",
  "windowEnd": "2026-09-01T00:00:00.000Z",
  "bucketIds": [
    "bucket_shared"
  ],
  "reportedComparableCost": "2455.75",
  "attributedCost": "1840.5",
  "coverageBps": 7494,
  "eligibility": "Closed source review; nonnegative reported cost; no adjustment, conflicting version or overlapping cost bucket"
}

What the data cannot tell us yet

3 scope notes

These limits explain which records could not be compared and which conclusions are unavailable.

Confirm an exclusive project link before comparing customer usage.

Project ownership has not been confirmed.

Stripe revenue and credits are context in this scan; no margin or credit-balance conclusion is made.

View source freshness and scope details

{
  "reviews": [
    {
      "batchId": "demo_ledger",
      "exportedAt": "2026-09-03T00:00:00Z",
      "completeThrough": "2026-09-01T00:00:00Z"
    },
    {
      "batchId": "demo_provider",
      "exportedAt": "2026-09-03T00:00:00Z",
      "completeThrough": "2026-09-01T00:00:00Z"
    },
    {
      "batchId": "demo_revenue",
      "exportedAt": "2026-09-03T00:00:00Z",
      "completeThrough": "2026-09-01T00:00:00Z"
    }
  ],
  "sourceAgeHours": [
    {
      "batchId": "demo_ledger",
      "hours": 24
    },
    {
      "batchId": "demo_provider",
      "hours": 24
    },
    {
      "batchId": "demo_revenue",
      "hours": 24
    }
  ],
  "consolidationLagHours": 24,
  "exclusions": [
    {
      "scope": "bucket_shared",
      "code": "CONFIRMED_COMPARABLE_SCOPE_REQUIRED",
      "recordIds": [
        "bucket_shared"
      ]
    },
    {
      "scope": "bucket_shared",
      "code": "NO_EXCLUSIVE_PROJECT_MAPPING",
      "recordIds": [
        "bucket_shared"
      ]
    },
    {
      "scope": "demo_revenue",
      "code": "REVENUE_MARGIN_AND_CREDIT_RULES_NOT_IMPLEMENTED",
      "recordIds": [
        "ch_sample"
      ]
    }
  ]
}

Covers observed provider spend and comparable usage differences. Revenue is imported as context. Each currency stands on its own; usage differences stay in their original units. Every record and amount in this demo is synthetic.