Stop arguing about which channel actually worked, measure it

The Problem

Modern marketing runs across many channels (Google, Meta, TikTok, Yandex, programmatic, organic, email, SMS) and each platform tells you its version of “its” performance. Each platform self-reports the conversions it claims credit for. Sum the platform numbers and you get more attributed conversions than actual conversions, because two or three platforms each claim the one sale.

CMOs trying to allocate budget face an impossible task: which channel actually drove the growth? Was it the Meta retargeting that got the click, or the Google brand search that got the conversion, or the YouTube view three weeks earlier that planted the intent? Without unified attribution, the answer is opinion. The data cannot settle it.

The conventional solutions, Google Analytics’s default last-click model, the platform self-reports, occasional spreadsheet exercises, are all wrong in known ways. Real marketing attribution requires unified data infrastructure: pull everything into one warehouse, model the full conversion path, apply a defensible attribution methodology, and surface the results in a BI tool where the marketing team can actually use them.

What the Solution Does

A unified data and analytics infrastructure for marketing attribution.

1. Ingest, pull data from every ad platform, web/mobile analytics, CRM, and order system.
2. Unify, match identities across platforms (cookies, login IDs, deterministic and probabilistic stitching).
3. Model, apply attribution methodology (last-click, first-click, multi-touch, data-driven, your choice, or multiple in parallel).
4. Surface, channel-level ROI, conversion path analysis, cohort behavior, in a BI tool your marketing team uses daily.
5. Iterate, methodology and reporting evolve with the marketing strategy. This is operational analytics that keeps running. A one-time consulting deliverable goes stale.

Where It Fits

This makes sense if you…

• Spend > €100k / year on advertising across multiple channels
• See ongoing arguments / uncertainty about which channels are driving results
• Have CRM / order data that can be linked to web/mobile sessions
• Want a unified analytics infrastructure in place of spreadsheet-driven monthly reports
• Are willing to invest in data engineering (the value sits in the unified layer, the dashboards just read from it)

This is probably not the right time if you…

• Have one channel, single-channel attribution is much simpler
• Don’t have technical resources to support the data integration
• Need privacy-strict attribution where cross-channel identity stitching isn’t permitted (some EU contexts)
• Aren’t ready to act on the data (insights without decision-changes have no ROI)

Business Value

Spend allocation discipline. When CMOs can see which channels actually contribute to revenue, the budget reallocation conversations become data-driven. Typical mis-allocation reduction: 15-30%, depending on baseline attribution discipline.

Decision speed. “Should we increase Google Ads spend by 50%?”, a question that used to take a week of spreadsheet wrangling and three meetings becomes a 10-minute look at the dashboard.

Cross-team alignment. Marketing, sales, finance, and product teams stop arguing about whose numbers are right. The data warehouse becomes the shared source of truth.

Cohort and conversion-path insights. Multi-touch attribution surfaces how customers actually convert, which is rarely the linear funnel marketing teams imagined. Higher-quality strategic decisions follow.

How It Works

The architecture below is the pattern we use for end-to-end attribution and marketing dashboards. It generalizes across e-commerce, SaaS, travel and content businesses.

1. Data ingestion

Connectors to every relevant data source:

• Ad platforms: Google Ads, Meta, TikTok, Yandex, Snapchat, etc.
• Web/mobile analytics: Google Analytics, GA4, Yandex Metrica, in-house event tracking.
• CRM: HubSpot, Salesforce, custom.
• Order/billing: e-commerce platform exports, payment processor records, custom DB.

We use Datapipe for the ETL pipeline, incremental processing, dependency tracking, automatic recovery from upstream API changes.

2. Data warehouse

ClickHouse is our default for high-volume analytical workloads. BigQuery, Snowflake, or Postgres also fine, depending on team preference and cost profile.

3. Identity stitching

Cross-device and cross-channel identity. Deterministic (login IDs, customer IDs, email matches) plus probabilistic (cookie graphs, device fingerprints where legally permitted). The accuracy of attribution depends sharply on this layer, bad identity stitching produces bad attribution.

4. Attribution model

We typically deploy multiple in parallel:

• Last-click (Google Ads default), for compatibility with platform reporting.
• First-click, for top-of-funnel channel analysis.
• Multi-touch (weighted), for realistic cross-channel value distribution.
• Data-driven (Shapley values on conversion paths), for the most rigorous answer, when data volume supports it.

Marketing teams typically compare last-click against multi-touch to see where channel value is being misattributed.

5. BI surface

We use Metabase by default (open source, marketing-team-friendly, fast to deploy). PowerBI / Looker / Tableau also fine. Standard dashboards: channel ROI, conversion paths, cohort retention, attribution comparison across models.

6. Continuous evolution

Marketing attribution isn’t done. New channels appear; old ones fade; campaigns change. The data pipeline and dashboards evolve. We typically retain ongoing engagement at a small monthly retainer for the first year.

Stack

Datapipe for ETL, ClickHouse / BigQuery / Snowflake for the warehouse, Metabase / PowerBI / Looker for dashboards, Python for custom transforms, native connectors for ad platforms. Identity stitching: custom logic per client based on available identifiers and regulatory constraints.

What You Need to Make This Work

Data. Read access to ad platforms (each has its own API and permission model). Web/mobile analytics export. CRM / order data. Optional: in-house event tracking for deeper insights.

Integrations. Outbound integration to your data warehouse (we set up if you don’t have one). BI tool of your choice.

Hardware. Cloud-based. Warehouse hosting cost scales with data volume.

Team. A marketing data lead who knows the channels and the business questions. A data engineer for the integration (around 40-60 hours during pilot). Marketing managers who’ll use the dashboards (their feedback shapes what we build).

Implementation Roadmap

1. Test (6-8 weeks)

Identify priority channels and conversion goals. Set up ingestion. Build the warehouse. Implement initial attribution model. Deliver the first dashboard. Output: working data pipeline for the priority channels, measured channel-level attribution, and recommendations for pilot scope.

2. Pilot (3-5 months)

Expand to all channels. Implement multi-touch attribution. Build comprehensive dashboards. Tune identity stitching. Output: production-grade attribution infrastructure with documented business impact.

3. Production (6-9 months)

Continuous evolution. Add new channels as they enter the marketing mix. Quarterly methodology review. Your team owns the dashboards and decisions; we stay on for technical maintenance and new-channel onboarding.

Keep in Mind

Known limits:

• Attribution methodology is a judgment call. No attribution model is “correct”. Each captures a different aspect of channel contribution. The value is in consistency: pick a methodology, apply it consistently, compare to alternatives.
• Identity stitching has accuracy limits. Cross-device matching is good but not perfect. Privacy regulations are pushing this harder over time. We surface confidence in attributions.
• Platform reports will not agree with your unified attribution. Each ad platform claims more credit than it deserves because they all measure overlapping conversions. This is the right outcome, but expect to explain it to stakeholders repeatedly.
• Data quality is the ceiling. Bad CRM data, untracked offline conversions, missing UTMs, all degrade attribution. Sometimes the most useful finding is “fix the tagging”.
• Privacy / cookie deprecation matters. Third-party cookie deprecation, GDPR / CCPA constraints, all affect what’s measurable. We design with privacy-by-default and explain the trade-offs.
• Attribution doesn’t tell you “why”. It tells you which channels were involved in conversions. “Why did Channel X convert this user?” is a qualitative question.
• CMO-level adoption matters more than technical accuracy. A dashboard the CMO ignores has no value. We co-design with the user.

FAQ

Can this work with Google Analytics 4 only?

GA4 is a good source for top-funnel data. Real attribution requires going beyond it, connecting GA4 sessions to CRM identities, to order data, to ad-platform spend. The unified warehouse is what makes this possible.

What about cookie deprecation?

Third-party cookie deprecation reshapes attribution. We design for the post-cookie world: first-party data primacy, server-side tracking, deterministic identity stitching where legally permitted. The methodology adapts; the architecture is forward-compatible.

How does this compare to attribution platforms (Funnel.io, Supermetrics, etc.)?

Commercial attribution platforms are excellent for organizations that want a ready-made product they configure and run. Our approach is the right choice when you need: custom attribution methodology, integration with non-standard data sources, fully on-prem deployment, or substantially lower per-month cost at scale.

Can the dashboards integrate with our existing BI?

Yes. We work with Metabase, PowerBI, Looker, Tableau. The data warehouse is the source; your BI tool reads from it.

Does this support B2B / SaaS attribution where the funnel is months long?

Yes. Long-cycle attribution is part of the design. Multi-touch attribution across multi-month journeys is exactly what data-driven attribution handles.

What about offline conversions?

Yes, if you can get the data into the warehouse, point-of-sale records, in-store conversions, phone-call conversions. We integrate from anywhere your business records conversions.

Ready to Discuss?

If you spend meaningfully on multi-channel marketing and your attribution is currently a guess-driven exercise, this is a worthwhile investment. We’ll walk through your channels, your data infrastructure, and your reporting cadence, and tell you what to expect from a pilot.

Discuss your project