Most Qlik Sense apps were built before natural language querying existed. That means a lot of production apps are quietly not ready for Qlik Answers, not because the data is wrong, but because the model was never designed to be read by an AI engine.

Here are the five practices that matter most, and at the end I go deep on the one that is most often missed.

1. Field Naming as a Human-Readable Contract

Every field name in your model is a term the Qlik Answers engine will use to interpret a user's question. SalesAmount, not SA_AMT_01. CustomerName, not CUST_NM.

The engine maps natural language to field names semantically. If your names are cryptic, abbreviated, or inherited from source system conventions, the engine has to guess. And it will guess wrong.

Field naming is no longer just a developer preference. It is part of the interface.

2. Clean, Deliberate Associations With No Synthetic Keys

Qlik Answers traverses your data model to find relevant tables when answering a question. Synthetic keys introduce ambiguity in how that traversal resolves.

Every association in your model should be a conscious, documented decision: one field, one relationship, one reason it exists. A star or snowflake schema built with intention gives the engine a clear, unambiguous graph to walk.

If you see a $Syn table in your model, understand it before you resolve it.

3. A Master Calendar With Consistent Naming Across All Fact Tables

Natural language questions are overwhelmingly time-based. "Show me sales last quarter." "What was revenue in Q1?"

If your date handling is inconsistent, if some tables join on OrderDate, others on InvoiceDate, and your master calendar field is called CalDate, the engine cannot reliably anchor time-based questions.

A single, well-named master calendar with standard period fields (Year, Month, Quarter, Week) connected consistently across all fact tables is foundational. This is one of those things that feels like overhead until Qlik Answers cannot answer a basic time question, and then it feels very important indeed.

4. Measures Defined in the Master Items Library, Not Buried in Charts

Qlik Answers surfaces master items as the vocabulary it uses to answer questions. A measure that only lives inside a chart object is invisible to the engine.

Every KPI a user might ask about verbally needs a corresponding master measure, named in plain language, with a description that reinforces intent.

Think of master items as the glossary you are handing to the AI. If the glossary is empty, the AI is working without one.

5. Facts Separate From Attributes

When the engine receives a question, it needs to understand which tables contain measures and which contain descriptive attributes. Blending transactional data with dimensional data in a single flat table makes this harder to reason about.

Keep fact tables lean (keys and measures). Keep dimension tables descriptive (attributes). Name them accordingly. A table called Sales_Fact and a table called Customer_Dim are self-describing to both a developer and an AI engine.

Deep Dive: Practice 4 in the Real World

This one has the highest ratio of impact to effort, and it is the practice most developers overlook because it feels like housekeeping rather than engineering.

The scenario

You are building a sales performance app for a retail business. The fact table has: OrderID, CustomerID, ProductID, OrderDate, Quantity, UnitPrice, DiscountPct, ReturnFlag.

A developer builds the app and creates the following expressions inside chart objects:

• Total Revenue: Sum(Quantity * UnitPrice * (1 - DiscountPct))
• Net Revenue (excl. returns): Sum(If(ReturnFlag = 0, Quantity * UnitPrice * (1 - DiscountPct)))
• Average Order Value: Sum(Quantity * UnitPrice) / Count(Distinct OrderID)

The charts look great. The app ships.

What Qlik Answers sees

Nothing. Those measures do not exist as far as the engine is concerned.

When a user asks "What was our net revenue last month?" the engine has no master measure called Net Revenue to anchor to. It may attempt to construct something from available fields, or it may return an uncertain or incomplete answer.

What changes when you do it properly

The developer promotes each expression to a master measure with a name and a description:

• Name: Total Revenue
• Description: Gross revenue after discount, across all orders including returns
• Expression: Sum(Quantity * UnitPrice * (1 - DiscountPct))

And separately:

• Name: Net Revenue
• Description: Revenue excluding returned orders, after discount
• Expression: Sum(If(ReturnFlag = 0, Quantity * UnitPrice * (1 - DiscountPct)))

Now when a user asks "What was net revenue in Q1 by product category?" the engine has a named, described measure it can identify with confidence, a master calendar it can use to filter by Q1, and a dimension it can group by. The answer is reliable because the vocabulary is explicit.

The bit most people miss

The description field on a master measure is not documentation. It is a semantic signal to the engine about what the measure represents and when it is appropriate to use it.

Developers who treat it as optional are leaving the engine to infer intent from an expression alone.

Name it what a user would say out loud. Describe it as if you were explaining it to someone who cannot see the formula. The engine will use both.

The fix for practice 4 takes minutes per measure. The payoff is that Qlik Answers actually knows what your app is measuring.

Start there

Good day,

We would like to inquire whether the same login credentials used for our Qlik on-premise environment can also be used for Qlik Cloud so that users can access both platforms using a single account.

Currently, our authentication setup is as follows:

• Qlik On-Premise: Users log in using their corporate ID (e.g., EMP12345) through Active Directory.
• Qlik Cloud: Users log in using their corporate email address (e.g., john.doe@company.com).

We would appreciate your guidance on the possible integration approaches and the recommended best practice.

Option 1

On-Premises Active Directory → Qlik On-Premise (QSE/QCA) + Qlik Cloud

• Is it possible to use our existing on-premises Active Directory as the single identity source for both Qlik on-premise and Qlik Cloud?
• If so, could you provide the detailed configuration steps and prerequisites?

Option 2

Cloud Identity Provider Microsoft Entra ID → Qlik On-Premise (QSE/QCA) + Qlik Cloud

• Is it possible to use a cloud-based identity provider as the single identity source for both Qlik on-premise and Qlik Cloud?
• If so, could you provide the detailed configuration steps and prerequisites?

Additionally, could you advise which of these two approaches is considered the recommended or best practice in terms of:

• Ease of implementation
• Security
• User experience (Single Sign-On)
• Long-term maintenance and scalability
• Compatibility with Qlik Cloud

Any implementation guides, documentation, or recommendations you can share would be greatly appreciated.

Thank you, and we look forward to your response.

Hello
I have a problem with a client who has disabled webSockets, so no access to the Qlik Sense hub. He talks to me about a "Long polling" alternative.
Is there any alternative to access the hub without enabling webSockets?

We want to work with Parquet files located on an on-premises S3 server. We've been given the URL, secret key, and access key. However, when we create the connection from the Data Load Editor using Amazon S3, we don't have the option to enter the endpoint URL. How can we do this? Any idea?

I have a published app. It is a test version copy of another app that is published in the same stream. I want to delete it. However, the delete functonality is greyed out. I am the owner of the app. 

Hello,

I have Vizlib Writeback table in the app.

I was asked to remove one column and to add another column (which is from real field ).

I have also wizard and also script, I tried to remove the column and add another one both in script and wizard and it throws me errors when i try to update the rows.

Please assist.

Thanks

Inna

Hi,

I have embedded Qlik charts in my mashup project using the <qlik-embed> component. I need to apply an initial selection when the mashup loads.

The requirement is to dynamically determine the highest (latest) value from a field and use it as the initial selection. For example, if the field is Year and it contains values such as 2022, 2023, 2024, 2025, and 2026, I need to identify the maximum value (2026) at runtime and apply it as the initial selection when the mashup opens.

Hi Community,

 

for the past few days, we have noticed an increase in error messages when sending reports via email in Reporting Services. For example, I have a report that is supposed to be sent to five recipients. On certain days, I occasionally encounter one or two error messages with the following text:

 

 

I can't find a pattern as to why the error occurs. What could explain it? In the past, it ran perfectly without any problems.

Kind regards,
Vanessa

Hello! I wonder if someone could help me with a (potentially very basic) question about data updates and Qlik Sense apps connected to Excel files. 

I have created a Qlik Sense app that takes its data from several MS Excel files that are saved on my local computer. My intention is to update the Excel files 5-6 times a year, and I want the app to fetch the latest version of the data every time I update the Excel files. I tested the app for a while and everything was working as intended so I went ahead and published it.

The problem is that, since this app is now published, I no longer have access to the Data Manager function which was how I was planning on feeding it new Excel files. In non-published apps I can simply go into Data Manager and drag the new Excel files to replace the old ones, but this function has disappeared in the published app. I can't seem to unpublish the app either, which was the first workaround I could come up with. I've also tried clicking on the app and then on Load, but that doesn't seem to load the new version of the Excel files for some reason.

Is there some other way in which I can force a published app to fetch new Excel files from my local computer?