‍

In Trust & Safety and training AI models, data quality is everything. Cinder’s Quality Assurance (QA) suite offers integrated tools to evaluate every decision—whether made by a human or AI—so you can finally understand how reliable your data is, create higher quality models and create a better user experience.

Unlike standalone QA tools, Cinder’s QA suite is built directly into our content moderation platform. That means:

1. Instant access to operational data—no extra setup or integration. No more exporting into another tool or using janky spreadsheets.
2. Real-time and asynchronous QA reviews adapted to all your workflows.
3. Seamless integration of QA results into AI training and measurement, reviewer allocation, and policy updates.

How Patreon Used QA to Increase Accuracy and Cut Handle Time 

Patreon uses Cinder’s QA suite to assess their BPO teams, improve training, and ultimately shift more reviews to their BPOs, saving time and money without sacrificing quality. The results were real:

"Cinder’s QA tool helped us identify process and policy gaps, improve accuracy, and confidently shift more work to our vendors—freeing up time for our internal team to focus on complex cases."

‍

How Cinder’s QA Suite Works

Cinder’s QA suite enables sampling, layered quality reviews, multi-review, and clear oversight tools that lead to direct action. 

‍

1. Simple sampling configuration 

Easily define and create QA samples with custom filters based on:

1. User, team, or policy
2. Review queue, automation, or any content attribute
3. Internally-developed or third-party classifiers

Examples:

1. Compare the accuracy of hate speech decisions made by classifiers and BPO reviewers in Germany
2. Compare BPO vs. FTE accuracy before outsourcing critical work
3. Assess third-party AI classifier performance to determine if internal development is needed

2. Layered QA Reviews

Set up multi-tiered QA processes, using both human and AI reviewers. This structure maximizes efficiency and allows BPOs to conduct multiple reviews while giving FTEs a third -layer for ultimate oversight:

1. Use first-tier QA reviews, second-tier grade-the-grader reviews, and third-tier benchmark reviews
2. Assign AI tools at any level of the QA process

3. Actionable Insights in Real Time

QA isn’t just about assessment—it’s about improving operations. Cinder’s real-time dashboards turn QA results into immediate action:

1. Adjust AI model thresholds instantly based on classifier precision and recall 
2. Reassign or train underperforming teams using QA findings

Multireview

It's important to get agreement first. Cinder’s multireview allows Admins to require agreement among multiple reviewers to finalize a data label. Disagreements can be automatically escalated to higher-tier reviewers. Our customers often use multireview on particularly nuanced labeling.

Cinder’s built-in Quality Assurance doesn’t just help you track accuracy—it is a foundational tool for optimizing your entire data operation. Save time, money, and effort while ensuring your data and decisions are rock-solid.

‍