Rebiha

Home › Datasets › Customer Support & Sales

Customer Support & Sales Fine-Tuning Dataset

Support tickets, de-escalation, troubleshooting, and sales conversations.

Fine-tune a model on this dataset

$29 for the dataset. Training is priced separately by model, starting at $47.

What is in this dataset

900 instruction-response pairs covering support tickets, de-escalation, troubleshooting, and sales conversations. Each pair is a realistic question with a full written answer, in the JSONL format used for supervised fine-tuning. No synthetic filler, no duplicated rows.

Example pairs

Real rows from the dataset, shown in full.

How do you de-escalate an angry customer whose order has been delayed by three weeks with no communication?

First, acknowledge their frustration sincerely without making excuses: 'I hear you, and I'm truly sorry for the lack of updates—that's not the experience we want you to have.' Then, take ownership of the issue: 'Let me personally look into your order right now.' Provide a concrete timeline for resolution—e.g., 'I'll get you a tracking number within 2 hours, and if it hasn't shipped by then, I'll escalate to our logistics manager.' Offer a goodwill gesture proportional to the delay, like a discount code or free expedited shipping on their next order. Avoid blaming carriers or internal teams; stay solution-focused. End by confirming their…

A customer is demanding a refund for a digital product they purchased 6 months ago, which is outside our 30-day policy. How should I handle this?

Start with empathy: 'I understand this product isn't meeting your needs, and I appreciate you reaching out.' Then, clearly restate the policy without being rigid: 'Our standard refund window is 30 days, which has passed, but I want to see what I can do for you.' If possible, offer a partial refund or store credit as a compromise—e.g., 'I can't process a full refund, but I can offer a 50% credit toward another product that might work better.' If the policy is firm, explain the rationale (e.g., digital access, licensing) and pivot to troubleshooting: 'Let me help you get the most out of what you already have—can you tell me what specific issue…

How do I explain a complex technical issue (like a database sync error) to a non-technical customer who just sees an error message?

Use a simple analogy: 'Think of your account like a library card. The error we're seeing means the library's computer is having trouble checking your card against our records—it's a small hiccup on our end, not something you did wrong.' Avoid jargon like 'API timeout' or 'cache invalidation.' Instead, say: 'Our systems are trying to talk to each other but got a little mixed up. I've already flagged this to our tech team, and they're fixing it now.' Provide a clear next step: 'You don't need to do anything; I'll send you an email once it's resolved, usually within 24 hours. If it pops up again, just reply to…

Who this is for

Helpdesk bots and sales assistants that handle upset customers.

Which models this works with

The dataset is plain JSONL, so it works with any instruction-tuned open model. On Rebiha you can train it directly on:

Training uses LoRA, or QLoRA on a 4-bit base for larger models. Your base model's weights are never modified — training produces a separate adapter.

How it works

  1. Pick this dataset and a base model.
  2. Rebiha runs the job on a managed GPU. You set up nothing.
  3. Download your model: a ready-to-run GGUF, plus developer assets (adapter, tokenizer, configs).

Fine-tuning or RAG?

Use RAG when the model needs facts that change — your prices, your documents, today's inventory. Use fine-tuning when you want consistent style and structure without prompting for it every time. This dataset teaches the shape of a good answer in this domain, not facts to look up.

Related datasets

Start fine-tuning