Review how well every finished support conversation was handled

Fields

• Channel (email, web chat, chat app, SMS, call)
• The whole conversation, oldest message first

Questions

1. Does the conversation contain a reply from an agent of the company? Yes / No

   The other questions mean something only when the answer is yes. Yes: At least one substantive reply written or spoken by an agent is present. No: Only customer messages appear, or the only company content is an automated acknowledgement.

2. Did the agent address the customer's main request? Yes / No

   A refusal counts when the agent gives a specific reason. Yes: The agent replies answer or act on the main request, or decline it with a specific reason. No: The agent replies ignore the main request, answer a different question, or refuse without a reason.

3. Did the agent deal with every separate question or request the customer made? Choice

   Count each distinct question or request the customer made anywhere in the conversation.

   • All answered Every question or request was answered, acted on or declined with a reason.
   • Some missed At least one question or request was left without an answer.
   • Single request The customer made only one request.

4. Did the agent make clear what happens next, when anything was left to do? Choice

   Judge the position at the end of the conversation.

   • Clear The next action and who does it were stated.
   • Not stated Something was left to do and the next step or its owner was not stated.
   • Nothing left The request was completed and nothing remained.

5. Did an agent say something that contradicts a fact stated earlier in the conversation? Yes / No

   Judge only against the conversation itself, not against outside knowledge. Yes: An agent message conflicts with an earlier fact given by the customer or by an agent. No: No such conflict appears.

6. Did the customer have to repeat a request or detail because an agent missed it? Yes / No

   Yes: The customer restates a request or detail they already gave, after an agent reply that missed it. No: The customer never had to repeat themselves, or repeated only to add something new.

7. Were the agent replies respectful throughout? Yes / No

   Yes: Every agent reply was courteous. No: At least one agent reply was dismissive, sarcastic, blaming or insulting.

8. How completely did the agent replies, taken together, answer the customer's request? Scale

   • Not addressed The replies are unrelated to the request or do not address it at all.
   • Partial The replies address the request with significant gaps left unanswered.
   • Mostly The replies substantially answer the request, with minor gaps.
   • Fully The replies answer the request fully, with a clear next step where one is needed.

9. How did the agent respond to the customer's situation? Scale

   Judge acknowledgement of what the customer described, not politeness formulas, and not whether the outcome was good. A routine request needs none; rate it as neutral.

   • Dismissive The agent brushed off the customer's situation or blamed the customer.
   • Neutral The agent was correct and polite without acknowledging the customer's situation.
   • Acknowledged The agent recognised the customer's situation, difficulty or frustration.

Lens columns

has_agent_reply, has_agent_reply_probability, request_addressed, request_addressed_probability, every_part_answered, every_part_answered_probability, next_step, next_step_probability, agent_contradiction, agent_contradiction_probability, customer_repeated, customer_repeated_probability, respectful, respectful_probability, completeness, completeness_average, empathy, empathy_average