Identity and scope
Source ID, provider account, store or property, order or transaction ID, event ID, record status, and test flags.
Interactive decision tree
Build a prioritized reconciliation checklist
This tool does not declare one platform correct. It narrows the likely mechanisms and links to a detailed guide. Inputs stay in your browser.
Discrepancy matrix
Classify the gap before changing the report
The same total can differ for several reasons at once. Use identifiers to measure coverage, then isolate one mechanism at a time.
| Class | Typical signature | Evidence to compare | Can it reconcile? |
|---|---|---|---|
| Tracking coverage | Analytics or ad events are lower than backend orders, often uneven by device, browser, consent state, or market. | Order ID, transaction ID, event ID, tag diagnostics, consent state, event timestamp. | Often partially. Missing historical events may not be recoverable. |
| Duplicate events | Event counts or revenue exceed order records, with repeated transaction or event identifiers. | Duplicate transaction IDs, browser/server event IDs, retry logs, event timestamps. | Usually, when a stable deduplication key exists. |
| Attribution | Platform conversions exceed or lag store orders; several platforms each claim credit for the same purchase. | Conversion event, click/view eligibility, attribution model, lookback window, campaign touchpoint. | Not to a single universal platform total. Preserve attributed and observed facts separately. |
| Revenue basis | A stable percentage or amount gap remains after order IDs align. | Gross and net sales, discounts, tax, shipping, tips, refunds, cancellations, status. | Yes, if both systems expose the required components. |
| Date and timezone | Daily differences reverse on adjacent days while the longer-period total is closer. | Event timestamp, order date, click date, conversion date, refund date, timezone, inclusive boundaries. | Usually, after choosing a declared date basis. |
| Currency | Market-specific gaps or a percentage gap that follows FX movements. | Original currency, reporting currency, FX date, provider rate, settlement currency. | Yes for reporting, but provider and internal historical rates may remain different views. |
| Scope and status | Specific stores, properties, test records, channels, or order statuses are consistently missing. | Store/property/account IDs, order source, test flag, draft/cancelled status, filters. | Usually, by making scope explicit rather than silently changing totals. |
| Report grain | Totals inflate after a join or change when rows are grouped at a lower level. | Primary keys, base entity, row grain, relationship cardinality, metric grain. | Yes if a safe aggregation or scoped allocation exists; otherwise keep the metrics separate. |
| Late changes | Recent periods keep moving after initial reporting. | Ingestion watermark, provider restatement, refund/cancellation events, late attribution, refresh time. | Often, after the period stabilizes and the replacement policy is understood. |
Manual workflow
A seven-step reconciliation that preserves meaning
Do not overwrite one source with another to make a dashboard look tidy. Document the semantic difference and decide which view answers the business question.
- Freeze a small, stable window.Use three to seven completed days first. Avoid diagnosing an actively changing current day.
- Write each metric as a full sentence.Include source, event, status, revenue basis, attribution, date, currency, and scope.
- Export or query shared identifiers.Compare order ID, transaction ID, and event ID coverage before comparing totals.
- Build matched, source-only, and analytics-only sets.Quantify missing and duplicated records instead of reasoning from percentages alone.
- Recompute a common revenue basis.Align discounts, refunds, tax, shipping, cancellation state, and currency using explicit component fields.
- Align time without erasing provenance.Choose an analysis date while retaining the source timestamps that explain reclassification.
- Publish the residual gap with a reason code.Tracking loss, attribution, timing, scope, and unresolved should remain visible.
Required contract
The minimum fields behind a credible comparison
A total-only screenshot is not enough evidence. These fields let an analyst determine what is missing, duplicated, reclassified, or inherently different.
Time
Source timestamp, order date, event date, click or impression date where relevant, refund date, timezone, ingestion time, and last update.
Value components
Gross amount, discounts, refund amount, tax, shipping, tips or duties where relevant, net amount, original currency, and reporting currency.
Marketing context
Conversion event, platform, campaign, ad set or ad, attribution model, lookback window, touchpoint type, and spend.
Inspect the contracts before comparing the fields
The Metric Hive Semantic Contract Library exposes registered report entities, grains, time keys, canonical mappings, aggregation behavior, and field visibility. Review the relevant Shopify, Google Analytics, Meta Ads, or Google Ads connector page too. Registration is inspectable evidence, not a guarantee that every connector report is customer-available or production-smoked.
Primary sources: verify current provider behavior
Provider definitions, interfaces, and reporting behavior can change. Use these official documents to validate the current source behavior behind a discrepancy, then record the settings used for your own extract.
Source-specific guides
Continue with the closest discrepancy
Each guide keeps observed, tracked, attributed, and derived facts distinct and includes a focused checklist.
Shopify vs. GA4 revenue
Reconcile backend orders with tracked ecommerce events across identifiers, revenue components, scope, date, and currency.
Meta conversions vs. Shopify orders
Separate platform-attributed purchase events from observed orders, then inspect event coverage and deduplication.
Platform vs. blended ROAS
Compare attribution-dependent channel efficiency with observed business revenue over total paid-media spend.
Keep both truths when the facts are different
Metric Hive is designed around source-specific and canonical semantic contracts so an attributed platform metric does not silently become observed commerce truth. It can make definitions and compatible comparisons explicit; it cannot recover events that were never collected or prove incrementality from attribution data alone.