Customer support and sales conversation ratings data
Support and sales conversations with AI agents and human agents, rated by the customers who took part in them. Real people bring everyday problems and purchases, then say whether the problem was solved, whether they felt heard and whether they would buy, and why, in their own words.
Specification
- Conversations
- Customer support and sales by chat, email and phone, with AI agents and human agents
- Topics
- Orders and returns, billing, subscriptions and cancellations, account problems, product questions and buying decisions
- Raters
- Adults in the US and UK who took part in the conversation as the customer, across ages and backgrounds
- Ratings
- Problem solved, effort, empathy, clarity, trust, likelihood to buy or stay, and overall satisfaction (CSAT)
- Annotation
- Reactions turn by turn, the moment the customer gave up or decided, and a reason in the customer’s own words
- Quality check
- Hidden attention checks and repeat items catch careless raters; every conversation is reviewed before it is kept
Example of the data
What you see, and the file you get.
| Problem solved? | no | “they asked for my order number twice” |
|---|---|---|
| Felt heard | 2 / 5 | an apology, then a script |
| Would buy again | maybe | price fine, delivery unclear |
| conversation_id | channel | topic | resolved | empathy_1_5 | would_buy | customer_comment |
|---|---|---|---|---|---|---|
| cv_0001 | chat | return_refund | no | 2 | unlikely | asked for my order number twice and never said when the refund comes |
| cv_0002 | phone | billing | yes | 5 | likely | fixed the double charge and told me what had happened |
| cv_0003 | chat | subscription_cancel | yes | 3 | unlikely | cancelled in the end, but pushed three offers first |
| cv_0004 | product_question | partly | 4 | maybe | clear answer on sizes, nothing about delivery to Scotland | |
| cv_0005 | phone | sales_inquiry | yes | 4 | likely | explained the plans without rushing me |
Sample rows. Also as JSON or Excel.
Contact us to get the dataset
Tell us what you need. We walk you through the data, its formats and how it is checked on a short call.
What it is used for
Training support agents
Teach AI agents what actually solves problems, not only what sounds polite.
Reward models for conversations
Use the customer’s own rating as the signal, not a guess about what customers want.
Better sales conversations
See which answers moved a customer to buy and which lost them.
Comparing agents
Measure AI and human agents on the same situations, rated by the people they served.
Questions
Who rates the conversations?
The customer in the conversation: a real adult who brought the problem or the purchase and rates it straight after, in their own words.
What do the ratings cover?
Whether the problem was solved, how much effort it took, whether the customer felt heard, how clear and trustworthy the answers were, and whether they would buy or stay.
Can I see samples first?
Yes. Request samples with the form on this page and we will send them with the full specification.
Tell us what your model needs.
Request samples and the full specification. We reply within a couple of business days.
Request samples