Commerce truth
Shopify powers Commerce Revenue, orders, AOV, and MER.
Case Study
Marketing intelligence / Data product
Relay
A performance workspace built around trustworthy numbers.
Relay turns Meta Ads, Google Ads, and Shopify exports into validated performance intelligence, explainable changes, and a client-ready reporting workflow.
V1 complete / Protected private beta
View GitHub repositorySOURCE DATA
ShopifyMeta AdsGoogle Ads
CANONICAL DATAData Healthprovenance · semantics · completeness
KPI engineChange intelligenceReport
Case Snapshot
The strategic brief
The problem, the system response, the available proof, the strategic value, and the intentional boundary.
- 01Problem
- Performance reporting becomes fragile when exports, metric definitions, data quality, and client commentary are handled as disconnected work.
- 02System
- A deterministic workspace that maps Meta Ads, Google Ads, and Shopify exports into canonical facts, data-health checks, KPI logic, and explainable reporting.
- 03Proof
- A protected V1 with automated test coverage, browser verification, public repository evidence, and a controlled synthetic multi-source workspace.
- 04Value
- Makes the route from source export to client report easier to inspect without pretending platform-attributed and commerce revenue are interchangeable.
- 05Limitation
- Private beta only: manual CSV ingestion, device-specific browser memory, no live provider synchronization, cloud collaboration, or autonomous AI analysis.
01 / Problem
Performance reports are repetitive. Trust is the harder problem.
A reporting cycle can mean exporting files, cleaning fields, reconciling definitions, calculating KPIs, locating movement, writing commentary, and rebuilding the same report again.
Meta, Google, and Shopify each describe a different part of performance. Those numbers can all be useful, but they are not interchangeable. A dashboard that makes them look comparable creates false certainty.
02 / Product principle
A metric should not become more certain as it moves further away from its source.
- 01Shopify remains the source for commerce revenue, orders, AOV, and MER.
- 02Meta and Google keep their own spend, conversions, attributed revenue, CPA, and ROAS.
- 03Provider-attributed revenue is never combined and presented as ecommerce revenue.
- 04Missing values are not turned into zero.
- 05Data quality is evaluated before interpretation.
- 06Narrative is composed from validated structured facts.
03 / System
One deterministic path from source data to a decision.
The same normalized facts power the dashboard, report, and narrative. The interface does not recalculate business logic, and the reporting layer does not reinterpret the dashboard.
- 01Source exports
- 02Detection and mapping
- 03Canonical data
- 04Data Health
- 05KPI engine
- 06Change Intelligence
- 07Narrative
- 08Dashboard and report
04 / Data Health
Before asking what happened, Relay asks whether the data can support an answer.
Relay checks the evidence before it exposes an interpretation. Blocking problems stop downstream analysis. Warnings remain visible instead of being silently repaired.
CHECKS
- Reporting-period coverage
- Expected source completeness
- Currency compatibility
- Mapping integrity
- Provenance and duplicate evidence
- Revenue semantics
WORKSPACE STATE
01Healthy
02Review required
03Blocked
05 / Decision layer
Explain movement without inventing causes.
Once the facts are valid, Relay compares the current period with the previous one and identifies meaningful KPI movement, efficiency changes, top movers, target performance, and areas needing attention.
Direction and interpretation stay separate. A metric can increase without Relay declaring that increase positive. V1 uses deterministic rules over structured facts rather than a generative AI model.
Platform truth
Meta and Google retain provider-specific spend, conversions, attributed revenue, CPA, and ROAS.
Reporting truth
The report uses the same validated analysis as the dashboard. If that analysis is stale, export is disabled.
06 / Verification
A private beta verified as a complete workflow.
Relay V1 closed with automated tests, cross-browser checks, production deployment verification, responsive review, and report/PDF validation.
A controlled synthetic multi-source workspace produced EUR 225 in Shopify commerce revenue, EUR 55 in compatible paid-media spend, a 4.09x MER, two Shopify orders, and 2x ROAS for both Meta and Google. The values are synthetic. The workflow that processed them is real.
07 / V1 boundary
A complete private beta, not a finished SaaS.
V1 deliberately uses manual CSV ingestion and device-specific browser memory. It does not claim live synchronization, cloud collaboration, or autonomous AI analysis.
- No public accounts or cloud client persistence.
- No live OAuth, automatic provider refresh, or credential persistence.
- No scheduled reports, email delivery, or multi-user collaboration.
- No encrypted local memory or durable public abuse controls.
- Browser-native PDF output remains browser owned.
Relay
Faster reporting matters when the answer remains trustworthy.
Relay brings product strategy, data semantics, UX, and engineering together around a reporting workflow that stays defensible from source data to client report.
