To get the best performance from Fin when categorizing conversations, it’s important to create clear and distinct attribute names and descriptions. Following these best practices will help Fin interpret prompts more reliably and increase your classification success ratio.
Best practices
Create short, descriptive value names
Keep your attribute names concise and easy to understand. A good name immediately signals the attribute's purpose.
Keep names under approximately 25 characters.
Use short, descriptive language (e.g., “Login Issues”, “Billing Questions”).
Avoid internal jargon, ticket IDs, or ambiguous terms.
Write detailed descriptions for values
A detailed description is crucial for helping Fin understand the nuances of each attribute value.
Describe what the value represents and when Fin should choose it.
Include keywords or phrases the customer might use.
Provide examples of questions or intents (e.g., “How do I reset my password?”).
Add clarifications on what doesn’t belong in the attribute, if necessary.
Make attributes values distinct
Ensure there is minimal overlap between your attribute values. If attributes are too similar, Fin may struggle to choose the correct one.
Before finalizing, ask yourself: “Would a human find it hard to choose between these attributes?”
Avoid creating redundant or overly narrow attributes that could be combined.
Create an ‘Other’ attribute value to prevent incorrect classification
We recommend including a general-purpose ‘Other’ attribute. This gives Fin a safe option to use when a conversation doesn’t clearly fit into any of your defined attributes, which helps prevent it from forcing an incorrect match.
This is especially useful when your other attributes should only apply to a specific subset of conversations (e.g., conversations about a particular product or feature).
By following these guidelines, you’ll help Fin interpret prompts more reliably, reduce confusion between attributes, and increase your classification success ratio.
Examples of good attribute value descriptions
Example 1
Account Access
This value covers conversations where customers cannot log into their account, have forgotten their password, or are locked out for security reasons. Customers often express urgency since they cannot use the product until access is restored.
Applies if the customer:
- Says they cannot log in to their account.
- Reports a forgotten or incorrect password.
- Mentions being locked out or account suspended.
Does not apply if the customer:
- Is asking about subscription or billing without mentioning login.
- Is requesting account deletion or GDPR/Privacy help.
- Is reporting a technical bug inside the app after logging in.
Likely Keywords: login, password, locked out, sign in, access denied
Example 2
Refund Requests
Covers conversations where customers explicitly request a refund, mention overcharges, or ask about cancelling payments. This includes disputes over being incorrectly charged or needing money returned.
Applies if the customer:
- Requests a refund directly (“I want my money back”).
- Mentions being charged incorrectly (“I was charged twice”).
- Asks about refund status or refund timelines.
Does not apply if the customer:
- Is reporting a failed payment or declined card.
- Is asking about subscription cancellation but not refunds.
- Is asking about financial hardship or restructuring options.
Likely Keywords: refund, charged, cancel payment, overcharged, money back, refund status
Pro tip: Try passing your attribute names and descriptions to a writing tool like ChatGPT or Claude to help you define them more clearly.
Example prompt: Write comprehensive descriptions for all of the attribute values listed - Include all relevant details about what belongs in the attribute. Think about every type of conversation that should fall under this attribute and describe them in the description. Providing a detailed description will help our AI Agent classify support conversations correctly. Include keywords and examples of what not to include if relevant.