Hi ​@maxime.psbpm,

Indeed, Celonis does not natively support 0:N → 0:N modeling, Celonis treats N:N as 1:N → 1:N, which can cause unintended data loss when filtering. This is because the default filtering logic in Celonis are INNER JOINS, where only matching records remain.

So in your case, This means contracts that do not have a linked asset get unintentionally filtered out.

To avoid losing unrelated contracts, in the past I modeled the assets in a separate table instead of relying solely on an intermediate table.

Your asset table would be something like this.

SELECT DISTINCT Asset 
FROM CONTRACT_ASSET 
WHERE Asset IS NOT NULL;

This table would then contain only the assets, without the link to contracts.

And then you would have to Modify your contract_asset table to ensure there are only valid links.

This ensures:

• Only actual contract-asset links exist in the join table
• Contracts without assets are handled separately.

In the end, you can apply the filtering on assets, not contract_assets. this ensures C3 and C4 aremain in the dataset because we didn’t filter them out through Celonis’ default INNER JOIN.

This is pretty hard to explain textually, apologies. If you need some more help, let me know.