User feedback is information people provide about their experiences, needs, or evaluations of a software system. For chatbots, it can expose frustrations that metrics and conversation logs alone may not reveal. section 3.4.1
Make feedback useful by:
- Collect detailed user descriptions and group them into recurring pain points; compare them with conversation transcripts to diagnose problems. section 3.4.1
- Use concise in-product ratings, such as thumbs up or down, and clarify whether they measure an answer, a process and its outcome, or the overall conversation. section 4.2.3
- Invite participation with incentives or a genuine thank-you, but look for repeated problems rather than treating isolated reports as statistically significant. section 3.4.1
Table 3.6 Survey responses leading to identified pain points (part 1)
|
User
|
Survey response
|
Identified problem
|
|---|---|---|
|
1
|
I tried asking about vaccine eligibility in different states, but the bot couldn’t provide clear information. Not knowing if I qualify for the vaccine when I plan to travel is frustrating. The responses seemed generic and didn’t address the complexity of eligibility criteria in various locations
|
PharmaBot understood the basic intent (eligibility) but failed to provide state-specific information. The chatbot did not consider the user’s location or the state they inquired about, leading to an unhelpful response.
|
|
2
|
I attempted to schedule a vaccine appointment, but the process felt confusing. The bot’s instructions were unclear, and I felt unsure if my appointment was successfully booked. It would be helpful if the bot could provide more guidance throughout the scheduling process.
|
The scheduling workflow lacked clarity, causing users to feel uncertain about whether their appointment was successfully booked.
|
|
3
|
(No survey response given; the user left the chat)
|
Overcomplicated steps discouraged users from completing the process. User 3 did not get to the survey.
|
Survey responses about vaccine eligibility and appointment scheduling show how user reports can be categorized into concrete chatbot problems. section 3.4.1
