Synthetic Keys Are Why Your Qlik App Can’t Be Trusted

Qlik Sense doesn’t scream when something’s wrong. It quietly links everything to everything.

Author

Glenda Carnate
February 13, 2026

Read time: 2.5 minutes

At first, the application appears to be functioning well. Sheets load quickly. Selections are working as anticipated. Totals are showing as expected. You begin to think you have built a high-quality associative model, even ready for production.

Then, when you navigate to the Data Model Viewer, you find a large number of synthetic keys present in the data model you created. It’s not just one or two... there are so many that you feel as though you have a number of stars in a constellation. Your once “healthy relationship” with Qlik has now turned into a realization that your data model is creating joins without your input.

How to Stop Synthetic Keys From Running Your Model?

1. Avoid allowing Qlik to determine the join logic for your data.

Once auto-generated keys are present in the data model, it is very hard to regain control.
To remedy this, rename/qualify your fields explicitly in the load script.

2. Construct an intentional "star-schema" layout for your data model.

Associative ability does not give you the right to construct a data model without design or structure.
To remedy this, clearly separate your fact and dimension tables.

3. Find and remove circular references as soon as you can.

Circular references may not always cause the application to crash, but they do corrupt the data's meaning.
To remedy this, run the Data Model Viewer just before reloading your data.

4. Gain control of your grain before you can trust your totals.

Multiple grains will create silent errors in your data.
To remedy this, set primary keys and validate record count for your tables.

5. Good governance will prevent hero-developers from having to troubleshoot their applications.

If a change to your application requires detective work to fix, the application will be very fragile and not robust.
To remedy this, document the purpose of each table, the relationship(s) between tables, and the key logic used to create them.

💡Key Takeaway: 

Qlik Sense doesn’t betray you. It exposes modeling shortcuts the moment complexity grows. Synthetic keys aren’t the problem, but undisciplined modeling is.

👉 LIKE this if synthetic keys have ever personally offended you.

👉 SUBSCRIBE now for Qlik Sense truths that protect performance and trust.

👉 Follow Glenda Carnate for clean modeling, governance, and BI discipline that survives scale.

👉 COMMENT with the highest synthetic key count you’ve ever seen.

👉 SHARE this with someone still pretending synthetic keys are harmless.

Reply

or to participate.