Skip to content
TB
TeamBenchResources
Features

Using Custom Review Criteria

Quick Answer

Custom criteria let you define exactly what your reviewer evaluates. For each criterion, you provide a name, a weight (importance), and guidance text that tells the AI what to look for. This ensures reviews reflect your specific quality standards.

Custom review criteria are the foundation of effective content review in TeamBench. They transform generic "is this good?" evaluations into structured, repeatable assessments against your exact standards.

Each criterion in a reviewer represents one dimension of quality you care about. Common criteria include readability, accuracy, brand voice, SEO optimization, structure, engagement, and compliance. But you can define criteria for anything -- technical accuracy for engineering docs, empathy for customer support responses, persuasiveness for sales copy, or accessibility for public-facing content.

When creating a criterion, you provide three things: a name, a weight, and guidance text. The name is a label (e.g., "Brand Voice Consistency"). The weight is a percentage that reflects how important this criterion is relative to the others -- all weights must add up to 100%. The guidance text is the most important element: it tells the AI exactly what to evaluate and what constitutes a high or low score.

Writing effective guidance text is the single most impactful thing you can do to improve review quality. Be specific. Instead of "Check for good readability," write "Evaluate sentence length (target under 20 words average), use of active voice (at least 80% of sentences), vocabulary level (appropriate for a general business audience), and paragraph length (3-5 sentences maximum)." Specific guidance produces specific, actionable feedback.

Weights should reflect your actual priorities, not what sounds balanced. If brand voice is twice as important as SEO for your team, make the weights reflect that. A common mistake is distributing weights evenly across all criteria, which dilutes the signal from your most important quality dimensions.

Consider using three to seven criteria per reviewer. Fewer than three limits the granularity of your feedback. More than seven spreads the AI's attention too thin and produces less focused evaluations. If you find yourself with many criteria, consider whether some can be combined or whether you need separate reviewers for different aspects.

You can also use templates as a starting point and then customize the criteria. This gives you a proven foundation while allowing you to adapt the evaluation to your specific needs. Many teams start with a template, run a few reviews, and then adjust criteria and weights based on whether the scores align with their human judgment.

Review your criteria periodically. As your content strategy evolves, your quality standards should too. Criteria that were important six months ago might be less relevant now, and new priorities may have emerged. Keeping criteria current ensures your reviewers continue to provide relevant, useful feedback.

Related Questions

How many criteria should a reviewer have?

Three to seven criteria is the sweet spot. Fewer than three produces overly broad feedback. More than seven dilutes evaluation quality. Focus on the dimensions that genuinely matter for the specific content type being reviewed.

Can I reuse criteria across different reviewers?

Currently, criteria are defined per reviewer. However, you can duplicate a reviewer to copy its criteria, then modify the duplicate. Templates also provide pre-built criteria sets that you can customize.

What happens if weights do not add up to 100%?

TeamBench requires weights to sum to 100%. If they do not, you will be prompted to adjust them before saving. This ensures the weighted scoring calculation produces meaningful results.

Can the AI evaluate subjective criteria like creativity?

Yes, but the quality of evaluation depends on your guidance text. For subjective criteria, provide concrete indicators -- "creative means unexpected metaphors, vivid imagery, and non-obvious angles" rather than just "be creative." The more concrete your guidance, the more consistent the evaluation.

Still have questions?

Try TeamBench free and see how AI-powered content review works for your team.

Start Free Trial