Quality assurance (QA) data can be highly valuable for AI developers, especially when it includes clear expected outcomes or results. This guide helps referral partners identify businesses in the U.S., Canada, the EU, the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia (outside the U.S., Canada and the EU, at least 70% of the data must be in English) that generate such data, increasing their chances of a successful referral for data licensing opportunities.
What makes QA data valuable for AI?
AI models thrive on structured data that shows a clear relationship between inputs and outputs. For QA, this means data demonstrating how a product, service, or process was tested, what the expected outcome was, and what the actual outcome or observation turned out to be. This enables AI developers to train models to:
- Predict quality issues before they occur.
- Automate quality checks.
- Identify patterns in defects or deviations.
- Optimize processes for better quality control.
Key characteristics of valuable QA data from companies
When evaluating a potential company referral, look for the following characteristics in their quality assurance processes and data:
- Original and Proprietary: The company should have its own internally generated QA documentation, test cases, and results. This often comes from their own product development, service delivery, or operational workflows. Usually not a fit: scraped data, or material belonging to clients.
- Defined Expected Outcomes: For each QA activity, there should be a clear
pass/failcriterion, a target metric (e.g., latency under X milliseconds, defect rate below Y%), or a specific functional behavior that was anticipated. This 'expected outcome' is crucial for AI training. - Detailed Test Records: Beyond just a pass/fail, valuable data includes the steps taken during testing, environmental factors, inputs provided, and detailed observations of the actual outcome. This context helps AI understand why something passed or failed.
- Operational Maturity: Companies with robust QA data often have established processes, indicating an operating history of several years and perhaps 50+ full-time employees at peak. This suggests a consistent generation of relevant data. See who qualifies for more details.
- Rights to License: The company must own the data or have explicit rights to license it without breaching confidentiality, privacy, or third-party agreements. This is a non-negotiable requirement for data licensing.
Examples of QA data with expected outcomes
To better understand what to look for, consider these illustrative examples:
- Illustrative: A software development company's bug tracking system, where each bug report includes detailed reproduction steps, the expected software behavior, and the observed (incorrect) behavior. Alongside this, test case results that show a test scenario, the expected output, and the actual output.
- Illustrative: A manufacturing firm's production line quality control logs. These records detail specific product batches, the parameters of the quality check (e.g., dimensions, material strength), the tolerance range (expected outcome), and the actual measured values.
- Illustrative: A customer support department's internal QA of agent interactions. Call transcripts or chat logs are evaluated against a rubric (expected outcome for excellent service), with specific scores or comments on actual agent performance. This might include ratings on problem resolution, empathy, or adherence to protocol.
What to discuss with potential company referrals
When speaking with a potential referral, focus on their internal processes and how they ensure quality. You don't need to see the data itself. Ask questions like:
- "How do you ensure the quality of your products/services?"
- "Do you have internal documentation for your quality assurance processes?"
- "What kinds of records do you keep from your testing or quality checks?"
- "Do you track expected versus actual performance for any of your key operational metrics?"
SourceX evaluates opportunities for relevance to SourceX's current data needs and can help a referred company understand the value of their data. Encourage suitable companies to use the company fit checker for a preliminary assessment.
Remember, your role is to make the introduction to a business in the U.S., Canada, the EU, the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE or Saudi Arabia. The company controls what data scope it is willing to provide and its terms, and SourceX handles evaluation, contracting, and delivery. Learn more about how it works.
Rewards are calculated as the greater of $50,000 plus 5% and 15% of the compensation SourceX actually pays the company you refer, with no cap. A lead, meeting or signed agreement alone does not trigger payment. See the program terms.