Data Model and Definitions

Same data. Different numbers.

It happens to the best of us. Teams use the same words to mean different things. Revenue, customer, and conversion sound obvious until you ask three departments to define them.

A data model defines how those business concepts are structured and calculated, so they can be used consistently across reports and applications.

Talk to us about your data model ↑
A data stack and a person asking what a number means

What is a semantic layer?

 Everyone knows which number they’re looking at, and what it means.

Data Warehouse → 

Take revenue.

Marketing records the order value as soon as a customer checks out. Finance recognizes revenue only after payment. Another report subtracts cancellations and refunds. All three numbers can be technically correct. They're just answering different questions.

A semantic layer gives those terms an agreed definition and calculation that every system working with the data can use.

A marketing dashboard and a finance report both show Revenue, with different numbers. In a semantic layer they become two named metrics: order value from completed online orders, and recognized revenue after payment, cancellations and refunds.

Why doesn't my dashboard match finance?

First, check whether you're actually measuring the same thing.

Customer

Someone who places an order, or someone who pays? Individual people, or accounts?

 
Conversion

An online order, a qualified lead and a signed contract can all be called a conversion, while measuring different things.
 
We take those decisions out of individual dashboards and spreadsheets, and put them in the data model.

The calculation is then reused in Power BI, spreadsheets and other reporting tools, built once.
More than one definition can be valid. The point is knowing which one you're using.
Data governance decides who owns a definition and who can change it. The data model puts it to work.

Data Governance →

An international fashion group asked us for better B2B dashboards. 

Interviews across its teams showed the dashboards weren’t the problem. “No KPI framework for me or my team.” “Sources are not connected.” “Dashboards are too high level.” And whenever a number was questioned, everyone went to the same analyst.

So we started with the definitions: for each role, which KPIs matter, which system they come from and how they are calculated. Once IT had signed those off, we built role-based standard reports on top, plus a sandbox where analysts could go deeper. The same foundation was later rolled out to further markets

Talk to us All services

Who decides the definitions?

You do.

Finance
decides what revenue means

Sales
decides what a qualified lead is

Marketing
decides what counts as a conversion

Our job is to find where the existing calculations differ, get the right people around the table, and build the agreed definitions into the model.

The definition remains yours.
When it changes, we update the model once, and every dashboard, spreadsheet and report follows.

Giving an AI application access to your data doesn't tell it what your organization means by revenue, active customer, or conversion. Those definitions need to exist in the data.

A documented data model gives AI applications definitions they can use consistently. Lineage, data quality and legal basis sit around it.

What does this have to do with AI?

Frequently asked questions
  • Do we need a data warehouse first?

    Not always. It depends on where the data sits and how it's used. A warehouse becomes useful when data from multiple systems needs to be combined, transformed and reused across reporting or analysis.

  • Which tools do you use?

    Dataform by default, or dbt where your team already runs it  The right choice depends on your existing data environment and who will maintain the model after delivery. Every model runs through raw, staging and production layers with automated tests, so a changed source doesn't silently change a number.

Make the definition explicit

If two reports call something the same thing but calculate it differently, you have a definition problem. We find where those differences are, agree where the logic belongs, and build it into the data model.

Data Governance

Agree who owns each definition, which sources count, and who can change them.

Read more

Data Warehouse

Give data from your websites, apps, ad platforms and internal systems one place to work together.

Read more

 

AI-Ready Data

Prepare data structures and definitions for reliable use in AI applications.

Read more
2026-06-02 DIGITAL MARKETING FESTIVAL 0041
Contact details

Cloud Nine Digital
Johann Siegerstraat 24
1096 BH
Amsterdam, Netherlands

T: +31 (0)20 787 1323

E: info@cloudninedigital.nl


Need something else?

Please do reach out. We are happy to help!