B2B software support resolution tasks with rubric graders, one vendor

Summary

Agent tasks cut from seven years of support work at one US B2B software company. Each task replays a customer ticket up to the point an agent acted, with the help-center articles and release notes available that day, and asks the model to resolve, escalate or ask for more information. A rubric per task scores the reply against what actually closed the ticket, and tasks are tagged when the customer reopened it within 7 days. Customer, staff and company names, emails, URLs and account ids are replaced with consistent placeholders.

Collected to order. We collect it from US companies for your order. Scope, volume and timing are confirmed in writing before you sign.

Pricing on request

Request scope and pricing

At a glance

Planned scope. Volume, sources and timing are confirmed in writing before you sign.

Records: 19Kgraded tasks
Size: 360 MB
Tokens, estimated: 90M
Years: 2019 to 2026
Firms: 1

Owner profile: One US B2B software company, 300 to 500 staff, selling inventory and order management software to wholesale distributors
Industry: Software
Data type: Evaluation and task packs
Headcount: 300 to 500 staff
Region: United States
Languages: English
Formats: JSONL
Updated: 2026-09-29

Sources

System	Kind	Records
Zendesk	Support tickets	1.3M tickets
Jira	Projects and wiki	184K issues
Confluence	Projects and wiki	6.2K help-center and runbook pages
Slack public channels	Team chat	3.1M messages

Good for

evaluating support agents on the resolve, escalate or ask decision
rubric-graded reply quality against reference resolutions
training retrieval over help-center articles as they stood on the ticket date
measuring how often agent replies lead to a reopened ticket

Fields

Field	Type	Description
task_id	string	Stable task id
product_area	string	Product area the ticket was filed under
customer_plan	enum	starter, growth or enterprise
ticket_thread	array<string>	Customer and agent turns up to the decision point
resources	array<string>	Help-center articles and release notes available on the ticket date
expected_action	enum	resolve, escalate or request_info, from what the agent did
reference_resolution	string	What actually closed the ticket
rubric	array<string>	Checks the grader scores the reply against
handle_time_minutes	int	Agent handle time on the original ticket
reopened_within_7_days	boolean	Whether the customer reopened the ticket

Example record

{
  "task_id": "T-58361-004417",
  "product_area": "EDI and trading partners",
  "customer_plan": "growth",
  "ticket_thread": [
    "customer: Since Monday every advance ship notice we send to [COMPANY_9] comes back rejected with 'invalid hierarchical level'. Orders to our other 11 retail partners go through fine.",
    "agent: Thanks [PERSON_12]. Did anything change in how you pack or ship those orders this week?",
    "customer: We started packing by carton on Monday so we can put a label on each box."
  ],
  "resources": [
    "[KB_118] Set up advance ship notices for a trading partner",
    "[RELEASE_NOTE_42] Version 4.18 carton-level packing"
  ],
  "expected_action": "resolve",
  "reference_resolution": "Carton-level packing added a pack level to each notice, and this partner's profile still used the shipment, order, item layout. Agent switched the profile to shipment, order, pack, item and resent the 14 rejected notices, all accepted.",
  "rubric": [
    "Ties the rejections to the switch to carton packing",
    "Checks the partner's hierarchy setting instead of the order data",
    "Resends the rejected notices or tells the customer how",
    "Does not tell the customer to stop packing by carton"
  ],
  "handle_time_minutes": 31,
  "reopened_within_7_days": false
}

One record, de-identified. Placeholders such as [PERSON_12] and [COMPANY_4] stay the same across the whole dataset, so threads still read correctly.

Licenses available

Exclusive
One buyer. Nobody else can license it during the exclusive term, usually 2 years.
Non-exclusive
Licensed to more than one buyer.
First look
One buyer gets 2 to 4 weeks to make an offer before anyone else sees the dataset.

What ships with it

Datasheet: what it is, what it is good for, verified size, sources, years, format and limits.
Provenance file: the owner shown by code, staff notice and consent status, exclusions, mapped to the EU AI Act Article 53 training-data summary and California AB 2013.
Manifest with a hash for every file.
QC report and de-identification report.
Your own copy, with unique canary records.

How it is de-identified

Default exclusions removed, files flattened to text, identifiers found in three layers and replaced with consistent placeholders, and the result measured against a 99% bar for each identifier type in each source.

The full method