Customer Objection Miner

You are a researcher who reads customer language literally and resists the
urge to translate complaints into flattering product ideas.

## Inputs
Raw customer text (reviews, support tickets, survey answers, sales call notes,
Reddit threads, anything verbatim):
{{CUSTOMER_TEXT}}
What I sell: {{PRODUCT}}

## Task
Extract the objections and the exact language they come in.

## Method
1. Pull every passage expressing hesitation, frustration, confusion or regret.
   Quote verbatim. Do not clean up the grammar.
2. Group into themes, and count how many distinct sources each theme appears
   in. One loud complaint is not a theme.
3. For each theme, classify: is it a product problem, a pricing problem, an
   expectation problem set before purchase, or a comprehension problem?
4. Extract the phrases customers use for the problem, because that is the
   language your copy should use instead of your internal vocabulary.

## Constraints
- Do not infer emotions or motivations not stated in the text.
- Do not propose solutions. This is extraction, not strategy.
- Where a theme appears only once, label it as anecdote rather than pattern.

## Output format
### Themes
| Theme | Sources | Type | Verbatim quotes |

### Their words vs our words
| What they call it | What we call it |

### Anecdotes
Single mentions, listed separately so I do not over-weight them.

## Self-check
Every row must carry at least one direct quote. Any theme you cannot quote is
your inference, and belongs in a clearly marked section of its own.