Your composable CDP and marketing data warehouse, plus governance
We build your Marketing Data Lakehouse on your own cloud (BigQuery/Snowflake), plus the governance layer on top: binding definitions, automated quality tests, and an access model so the numbers stop drifting apart.
- BigQuery
- Snowflake
- dbt
- funnel.io
- Hightouch
- Census
- Great Expectations
- DataHub
✓ funnel.io partner · full-cycle delivery with Saloid
In short
We build the central data platform where marketing, shop, and CRM data come together, on BigQuery or Snowflake, EU-hosted. Plus the governance layer: binding definitions, automated quality tests, and an access model. The numbers stay reliable, instead of meaning something different in every tool.
- Who it's for
- E-commerce & B2B with data in silos
- What you get
- Lakehouse, reverse-ETL & governance framework
- Entry
- Audit from €4,500, Build in 6 to 10 weeks
01Quick self-check
You're in the right place if:
Tick what applies.
02How it works
From source to activation, in four steps.
One central lake, tested and documented, instead of point-to-point exports.
- funnel.io · dbt
Ingest
funnel.io and custom connectors load shop, ads, CRM, and finance into the lake automatically.
- dbt · marts
Model
dbt normalises, enriches, and sets the business logic for attribution and LTV.
- dbt tests · Slack
Test
Automated checks for completeness, freshness, and plausibility. Slack alert on drift, before a dashboard lies.
- Hightouch · Census
Activate
Reverse-ETL pushes the tested audiences back into ads, CRM, and Klaviyo.
03What we build
Was wir bauen
GA4 has a sampling problem. Meta Ads reports different conversions than GA4. The shop knows its own revenue, the CRM knows lifetime value, and a spreadsheet ties all four together once a week, somehow. Error potential in every row.
The data platform solves this structurally: all relevant sources land in one central, EU-compliant lake on BigQuery or Snowflake, get normalised and enriched, and are ready for dashboards, ML models, and marketing activation.
But a lake alone isn't enough. At some point no one trusts the numbers, because "lead" means three different things and nobody is responsible for implementing one definition everywhere. So we build the governance layer right alongside: binding definitions, automated quality tests, and an access model that protects production dashboards from accidents.
Joint engagement with a clear split. Datascale delivers architecture, pipelines, activation logic, and governance processes, Saloid the EU cloud infrastructure and the technical implementation.
04The difference
All-in-one CDP vs. composable lakehouse.
Architecture: EU data residency in BigQuery Frankfurt or Snowflake EU. Storage stays with you, not with the CDP vendor. CLOUD Act hardening via CMEK is covered in the FAQ.
05The building blocks
Five building blocks. One platform.
From architecture to governance, bookable individually or as a chain.
Data architecture & lake
BigQuery or Snowflake as the central, EU-hosted data source. We decide tool and schema for the concrete context, not from a standard template.
Concretely: Current-state analysis of all sources, target architecture document, BigQuery vs. Snowflake decision with reasoning, cost control.
Tools: BigQuery EU · Snowflake · CMEK
ELT pipelines & modelling
Every relevant source runs into the lake automatically and becomes tested models in dbt. Staging, marts, business logic.
Concretely: funnel.io plus custom connectors for shop, CRM, finance. dbt transformations, daily orchestration with monitoring.
Tools: funnel.io · dbt · Airflow
Marketing activation & reverse-ETL
The tested data goes back where it acts: into ads, CRM, and marketing automation. No export spreadsheet in between.
Concretely: Attribution (data-driven instead of last-click), LTV on your own transaction data, audience segments via reverse-ETL.
Tools: Hightouch · Census · Klaviyo · HubSpot
Definitions & business glossary
If "lead" means three different things across marketing, CRM, and finance, no dashboard fixes it. We define the central metrics once, bindingly.
Concretely: Stakeholder mapping, binding glossary for lead, MQL, SQL, conversion, revenue, mapping onto the fields in GA4, CRM, and ERP.
Tools: Notion / Confluence · dbt docs · DataHub
Data quality & access model
Definitions without tests drift again. Automated checks warn before silent data loss, and an RBAC model protects production pipelines from accidents.
Concretely: dbt tests and Great Expectations for critical pipelines, Slack alerts with diagnostic context, RBAC and schema-change management with review gates.
Tools: dbt tests · Great Expectations · BigQuery IAM
06In plain terms
The key terms, briefly explained.
Composable CDP
Your own warehouse as the central customer-data store, instead of packaging storage, modelling, and activation into a proprietary all-in-one system.
Reverse-ETL
The way back: tested segments from the lake into ads, CRM, and marketing automation, via Hightouch or Census.
Data reliability
The guarantee that a definition means the same thing across systems, and that tests keep it that way after a schema change.
RBAC
Role-based access control. Who may see and change what, so nobody accidentally overwrites a production view.
07Who it is for
When it pays off.
E-commerce with siloed data
Data in shop, ad systems, CRM, and customer service that never connect. LTV and churn analysis on the roadmap, but the foundation is missing.
B2B with a sales cycle
Marketing, CRM, and finance never show the same numbers. CAC and LTV aren't internally calculable because data is locked in silos.
Teams with number conflicts
Meetings start with which numbers are the right ones. That's not a data problem but a governance problem, and it costs more time than the fix.
Pre-BI and pre-ML teams
Power BI, Tableau, or ML models are next and need a scalable, cleanly documented data source as a foundation.
08Deliverables
What you end up with.
Data architecture
- Current-state analysis of all relevant data sources (completeness, quality, freshness)
- Target architecture document: which data goes where, which schema, which granularity
- Tool decision BigQuery vs. Snowflake with reasoning for the concrete context
Pipelines & modelling
- ELT pipelines for all sources (funnel.io plus custom connectors for shop, CRM, finance)
- dbt transformations: normalisation, enrichment, business logic
- Daily refresh with monitoring and alerting on failures
Governance & data quality
- Business glossary with binding definitions for lead, MQL, SQL, conversion, revenue
- Automated tests (completeness, freshness, plausibility) plus Slack alerts with diagnostic context
- RBAC model and schema-change management with review gates for breaking changes
- Data catalogue with owner assignment per table
Activation & handover
- Attribution and LTV model on your own data
- Audience segmentation for reverse-ETL (Google Ads, Meta, Klaviyo, HubSpot)
- Dashboard (Data Studio, Power BI, or Tableau) plus data-model documentation
- 30-day post-launch support
What we do NOT do
Engagement depths
Three depths. Clear scopes.
No retainer trap.
Audit Sprint
We audit what is wrong. Prioritised report + action plan.
plus statutory VAT
Included
- Full analysis of the existing setup
- Prioritised report with concrete actions
- 90-minute walkthrough with your team
Not included
- Implementation (follows in the Build Sprint)
- Code in your app or website
When it fits
When the setup runs but the numbers are questioned internally.
Fixed price by scope (Audit Sprint / Audit Sprint Plus). 50% credited toward a Build Sprint if commissioned within 30 days.
Request an Audit Sprint →Build Sprint
Fresh build or restructure, built to spec.
plus statutory VAT
Included
- BigQuery / Snowflake setup + funnel.io connectors
- dbt models + tests, business glossary, RBAC
- Reverse-ETL (Hightouch / Census) + monitoring
Not included
- Campaign execution, media buying, creative
- Tool licences (billed directly, no markup)
When it fits
When a clean rebuild beats patching in production.
Final fixed price after scope definition.
Discuss a Build Sprint →Managed Evolution
Ongoing partnership. Analytics as a product.
plus statutory VAT
Included
- Monthly development + roadmap
- QA on every release deploy
- Slack support, < 4 h response (Mon–Fri)
- Monthly report + executive summary
Not included
- 24/7 on-call rotation
- Campaign operations
When it fits
When analytics has to keep growing with you.
Monthly cancellation after the minimum term.
Request Managed Evolution →All prices net, plus statutory VAT. For companies in Germany, Austria and Switzerland.
What is the difference between a data warehouse and a data lake?
A classic data warehouse (Redshift, on-premise systems) is highly structured and optimised for SQL queries. A modern data lake or lakehouse (BigQuery, Snowflake) combines the flexibility of a lake with the query performance of a warehouse. For marketing use cases BigQuery is the better choice in most setups today.
Do we really need a data lake, or is funnel.io alone enough?
funnel.io alone is sufficient for most marketing dashboards (→ Revenue Intelligence). A data lake becomes necessary when shop transaction data, CRM data, and marketing data need to be unified, when LTV or attribution models should be calculated, or when ML applications are planned.
Can we host BigQuery in the EU?
Yes. BigQuery offers EU regions (europe-west3 Frankfurt, europe-west4 Netherlands). Datascale configures all projects in EU regions by default. All data stays in the EU region.
How do you mitigate the US CLOUD Act in a GCP-based architecture?
BigQuery is configured exclusively in an EU region (europe-west3 Frankfurt or europe-west4 Netherlands). Data residency is contractually assured by Google. Residual risk: the CLOUD Act targets US parent companies. For highly sensitive data we combine BigQuery encryption with CMEK (Customer-Managed Encryption Keys), optionally with an External Key Manager from EU vendors like Fortanix or Thales. The decryption key then sits outside CLOUD Act reach. For maximum sovereignty: Snowflake on AWS Frankfurt with the same External-Key setup, or an open-source lake on StackIT or IONOS.
What does a Composable CDP cost compared with an all-in-one CDP?
An enterprise CDP (Segment, mParticle, Tealium) typically runs €80,000 to €250,000 per year, depending on MTU volume. Plus implementation. A composable setup on BigQuery EU sits in a different order of magnitude: BigQuery storage and compute together stay under €1,500 per month for most DACH mid-market companies, dbt Cloud Team plan from €100 per month, Hightouch Starter from €350 per month. Implementation as a Build Sprint from €7,500 net. Year-one total cost is typically €35,000 to €60,000. The biggest difference is not the price. It is data control: storage belongs to you, not to the CDP vendor.
What's the difference between data reliability and classic data quality?
Data quality measures individual values (completeness, format, range). Data reliability guarantees that a definition means the same thing across systems, and that automated tests keep it that way when someone changes a schema or connects a new source. Data quality is a snapshot; data reliability is a process.
Do we already need a finished data lake for the governance layer?
No. We start with the sources that already exist. GA4, Plausible, CRM exports, ad APIs. If the audit shows a central layer is missing, that becomes a recommendation, not an upfront investment. Most reliability problems can be addressed without a finished lakehouse, as long as definitions and tests are properly documented once.
Who takes on the ongoing data steward role after the setup?
Someone internal, from the data or analytics team. We build the framework, document the processes, and onboard the first steward, but governance only works if it's owned internally. In our experience, externally-staffed steward roles don't last 12 months.
How long until the first quality alerts go live?
After a 2-week Audit Sprint the 3–5 most critical pipelines have alerts in place. The full build with data catalogue, RBAC model, and governance processes typically runs over 6 to 10 weeks, together with Saloid for the technical implementation.
Integrations
From the integrations catalog
These are the tools we wire up for this service. Each catalog page rates the tool honestly: setup effort, GDPR classification and when it fits.
Next step
Data platform and governance: architecture conversation.
Strategy call about lakehouse architecture, reverse-ETL, and governance. Full-cycle implementation together with Saloid.
- Entry
- €4,500–7,500 net
- Delivery
- 6–10 weeks
- Scope
- 5 modules