Best practices for categories

To create optimal categories, review the following best practices.

Recommended number of terms / phrases per category

Typically, up to thirty (30) terms / phrases per category produce the best category accuracy. Although 30 (thirty) is not a hard system limit, usually ten (10) to twenty (20) terms yield optimal results. Using over thirty (30) terms typically produce less precise categories and may impact performance.

Analyze the goal for the category

Gather information on the business issue you want to investigate and track. Questions you may want to ask include:

Identify your key business goal.
Determine the type of interactions the business issue generates. For example, the reason for interactions: a specific campaign, customer service, technical support, channel deflection, sales opportunities, among others.

Define search criteria that meets the goals set for the category

Once you define the needs for the business issue, define the search criteria that most accurately represents the business issue. The criteria can be as extensive as you need.

Avoid meaningless terms

When adding terms or phrases to a category, avoid the use of short connecting words that are less likely to be recognized. Use topics and relations instead.
Focus on the issue, not the root-cause

To avoid creating categories with a bias, focus on the terms that surface the issue and not the root-cause.
Create concise and accurate categories

Try to hit the "sweet spot" — the category should not be too broad as to include false positives, and not too narrow that it excludes information.

Test for precision and recall

Test the category to verify that it retrieves the interactions that match the business issue, and modify and reprocess as needed (precision).

In the Analyze > Interactions workspace, review the interactions to verify that they represent the business issue you want to capture through the category.