Quick answer
Different tools answer different questions. Start with governed transaction data and a defined decision. Add specialist attribution, BI, MMM or incrementality only when the team can operate and validate it.
Definitions
Five measurement jobs
| Layer | Primary question | Current example |
|---|---|---|
| Web analytics | What happened across a site or app? | Google Analytics 4 |
| Ecommerce BI | How is the commerce business operating? | Polar Analytics; Triple Whale |
| Multi-touch attribution | How does a model distribute conversion credit? | Triple Whale; Northbeam; Polar Analytics |
| Marketing-mix modeling | How do aggregate inputs relate to outcomes over time? | Advanced options documented by Triple Whale and Northbeam |
| Incrementality | What happened because of an intervention? | Advanced options documented by Triple Whale, Northbeam and Polar Analytics |
Examples indicate documented product roles, not rankings or completed hands-on validation.
Method before dashboard
Why attribution tools disagree
Tools can ingest different events and costs, resolve identity differently, apply different attribution windows and models, include or exclude view-through activity, and treat refunds, time zones or consent gaps differently. A discrepancy is a question to investigate—not automatic proof that one tool is broken.
Governance
Questions for every evaluation
Source of truth
Name the systems governing orders, refunds, customer records, spend and finance. Decide which discrepancy each tool must explain.
Windows and models
Document lookback windows, eligible touchpoints, conversion time, model logic and whether settings can be audited.
View-through
Ask which impressions qualify, how identity is established and how consent, device changes and windows alter credit.
First-party tracking
Understand collection, identity resolution, denied-consent behavior, implementation dependencies and coverage gaps.
Privacy and consent
Map regional requirements, processor roles, retention, deletion and the reporting effect of missing consent.
Multi-channel and multi-store
Verify currencies, time zones, store entities, offline/POS revenue, marketplaces, subscriptions and B2B flows.
Warehouse and export
Ask whether raw and modeled data can move through files, APIs or a warehouse—and what remains accessible after exit.
Refresh and history
Test latency, late-arriving data, corrections, backfill depth and whether historical data is reprocessed.
Custom metrics and SQL
Check whether analysts can define governed measures, inspect logic and reproduce reported results.
Implementation burden
Account for tagging, connections, QA, identity design, governance, training, support and ongoing administration.
Commercial model
Pricing basis changes the business case
| Basis | Examples | Question to model |
|---|---|---|
| Free / enterprise | GA4 tools free of charge; Analytics 360 unpriced here | Limits, implementation, export and enterprise terms |
| Annual GMV + package | Triple Whale; Polar Analytics | Cost at present and forecast revenue |
| Annual marketing spend / quote | Northbeam | Cost as media spend crosses bands |
Different bases do not prove that one product is cheaper. Include implementation, specialist labor, services, contract length and ongoing governance.
Buy in sequence
Measurement maturity
Foundation
A smaller store may need accurate store records, correctly implemented GA4 or equivalent basic analytics, consent controls and consistent campaign tagging—not a full measurement suite.
Scaling paid media
A team making material cross-channel allocation decisions may investigate dedicated attribution after its transaction, spend and campaign data are governed.
Cross-channel / enterprise
A mature organization may investigate BI, warehouse workflows, MMM and incrementality when it has sufficient data, ownership and analytical capacity.
Restraint
You probably do not need every measurement tool
Overlapping tools create reconciliation work, duplicate implementation and competing definitions. Buy the smallest set that answers named decisions, assign an owner to every metric, and require an exit path for data.
Use the MER & CAC calculator for a blended baseline, then read the Analytics desk, software directory, Stack Builder, and research methodology.
Evidence set