Working with applications in Qlik Answers

Applications allow Qlik Answers to provide responses from structured data sources. Applications can be added as a content source for assistants. Applications can also be individually made available for Qlik Answers.

Qlik Answers can also generate charts and sheets for the user, allowing for rapid creation of application content.

Qlik Answers indexes applications to create internal definitions of application data. These are then used to answer user questions and generate charts and sheets. Applications are indexed when added to an assistant or when made available for Qlik Answers. They are reindexed with every reload.

Qlik Answers uses the following to understand application data:

• The data model

  The fields and data model of the application are the primary information used by Qlik Answers.

• Logical model

  The logical model created by business logic is used by Qlik Answers, but only the information in Overview and Fields & groups.

  Fields hidden in the logical model are not used by Qlik Answers.

• Master items

  Qlik Answers puts a priority on the use of master dimensions and measures when indexing the data model, as they are user-created.

• Application and master item descriptions

  Unstructured data from descriptions is used to provide additional context information for fields and their use.

• Business logic synonyms

  Terms added as synonyms are used to provide alternative terminology for user questions.

For information on preparing your applications for use with Qlik Answers, see Best practices for preparing applications for Qlik Answers.

Best practices for preparing applications for Qlik Answers

While any application can be used for Qlik Answers, time spent preparing the application for use with Qlik Answers improves the quality of responses.

Clarity and context are the most important factors for ensuring good results from Qlik Answers. Clarity makes it easy for Qlik Answers to understand the data in an application. Context helps Qlik Answers interpret and correctly use the data within an application. The following best practices for preparing an application help ensure clarity and context:

• Use unambiguous and descriptive field names

• Streamline your data model

• Check the data formats of fields

• Use master items

• Use business logic vocabulary to add terminology

• Remove unnecessary fields from your data model

Use unambiguous and descriptive field names

Your data model should use fields that are descriptive and unambiguous. Field names should:

• Clarify the business meaning of each field.

• Apply clear, business-aligned naming that outline differences or similarities to other fields.

Try doing the following to help make fields clear to Qlik Answers:

• Use full wording, such as Customer Name instead of CUST_NM. This helps align fields with natural language questions.

• Provide contextual qualifiers in the field names to help disambiguate between fields. For example:

• Make locations clear. Use Customer City and Store City instead of two fields both named City.

• Make data/time clear. Use Order Date and Shipment Date instead of two fields both named Date.

• Try to indicate the role and type of the field. Incorporate words like count, total, amount, or percentage to clarify their aggregative nature. For example, Order Count. If your field names use Booleans, they should read as prepositions, such as by using prefixes such as is_active or has_churned.

• For key fields, prefix them with context. For example, customer_id and order_id is preferable to something like a generic field such as ID or cust_ref.

Avoid using fields names that do the following:

• Use opaque codes or technical jargon.

• Use Flag or a bare adjective such as Active for a Boolean field as this can hinder the interpretation by Qlik Answers.

• Use ambiguous or generic nouns without context. A field called Amount, for example, does not communicate what it is an amount of. Multiple generic fields can make it hard for Qlik Answers to reliably map natural language queries to the correct field.

• Use cryptic abbreviations, such as cust_no for customer number or txn for transaction. These make it more difficult for Qlik Answers to understand the field.

Streamline your data model

Streamlining your data model by removing unnecessary fields produces more accurate, predictable answers from Qlik Answers. A curated selection of fields reduces the chances of incorrect field selection or confusion from Qlik Answers. Streamlined data models are also faster to index. To streamline your data model:

• Hide technical fields.

• Consolidate redundant or near-duplicate fields into a single authoritative version.

Hide technical fields

Your data model should present Qlik Answers with fields that contain real analytic value. Avoid including technical fields that do not contribute to Qlik Answers understanding the application. Technical fields contain information such as: 

• IDs

• Keys

• Load timestamps

• Staging columns

You can remove unnecessary fields by hiding them. Hidden fields are still available for script logic or internal calculations, but are excluded from Qlik Answers analysis.

• In load script or Data manager, add a % prefix in their name (for example, %Discount2)

• In the logical model in business logic, set their visibility to Hidden.

  For more information, see Visibility.

Consolidate and rename fields

You should consolidate redundant or near-duplicate fields into a single authoritative version. Ambiguous fields make it difficult for Qlik Answers to interpret data correctly.

Rename or consolidate fields to make the data model as clear as possible for Qlik Answers to understand.

Example: Streamlining the data model by fixing ambiguous fields

Consider the following field names from a data model:

• Discount_Amount

• Discount_Value

• Discount1

• Discount2

These field names create a number of issues for Qlik Answers when it tries to interpret them:

• Multiple fields compete for the term discount, creating ambiguity for Qlik Answers.

• Numeric suffixes (1, 2) and vague field names provide no clear business meaning.

• Mixed naming conventions reduce clarity.

To fix these issues, the fields should be renamed when preparing the application for use with Qlik Answers. If the fields represent different concepts, they should be renamed to be clearer to their use and purpose. For example:

• Product Discount

• Promotional Discount

• Coupon Discount

• Loyalty Discount

If they represent the same concept, they should be consolidated into a single authoritative field, such as Discount Amount. If any of these fields are technical or legacy, they should be hidden.

Format data/time fields loaded as plain text

Some fields contain date/time information, but are loaded as plain text in the data model. As they are not classified as date/time fields, but rather text fields, they will not be used correctly in Qlik Answers analyses as they will not be treated as true date fields.

If a field containing date/time information is tagged or stored as text, convert it to the proper format during the load, either with tools in Data manager or with date functions in the load script. This ensures that:

• The field is recognized as containing dates.

• The auto-calendar generation works.

• Time-based questions from users map correctly to the data/time fields.

• Sorting and filtering behave correctly.

Use master items

Master items improve the ability of Qlik Answers to interpret application data. When interpreting questions, Qlik Answers weighs master items more heavily than fields in the data model as they are user-created. They are important because a user thought they were important enough to create them.

Master items add clarity and reduce ambiguity by creating a single, trusted version of each important metric or field in a data model. This also helps keep similar answers aligned across users. When someone asks about Profit margin and there is a corresponding master measure, the answer is based on the same definition, no matter who asks or who the question is phrased.

One of the most important elements of master items are descriptions. Descriptions are used by Qlik Answers to provide context for interpreting master items. Strong descriptions in master items clearly explain:

• Intent

• Meaning

• Business context

Master item descriptions make it easier for Qlik Answers to understand a master dimension or measure and pick the correct metric, even if users ask in an unexpected way.

For additional best practices for master items and Qlik Answers, see Writing master item descriptions for Qlik Answers.

Example: Useful master item description

Master measure: Customer Acquisition Cost

Description: Average cost to acquire a new customer. Calculated as total marketing and sales spend divided by the number of newly acquired customers. Excludes retention or renewal spend. Also known as CAC.

Use business logic synonyms to add value

Business logic synonyms help you to refine how Qlik Answers interprets terminology. While Qlik Answers understands common business language, your data model may include terminology that a LLM would not naturally recognize or interpret correctly. Synonyms help Qlik Answers understand the terminology unique to your organization's data. The following categories of terms benefit from adding synonyms:

• Organization-specific jargon or acronyms

• Internal KPI nicknames

• Product or process codes that double as business terms

• Metrics that appear similar but have distinct internal definitions

• Domain-specific language not widely used outside your industry

Try to avoid synonyms that do the following:

• Add ambiguity. Including top or bottom, for example, can cause issues as they are not clear. For example, 0 could mean the top 5, the top 10%, top by revenue or number of deals.

  Adding synonyms that duplicate values from fields can also add ambiguity.

• Duplicate vocabulary for the same terms, for example, adding the synonym sales to two separate fields.

• Include stop words, as they can cause questions to be disallowed.

To learn more, see Adding synonyms.