> For the complete documentation index, see [llms.txt](https://help.timelygrader.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.timelygrader.ai/user-guides-get-started-here/rubrics.md).
# Building rubrics that is clear, consistent, and free of error.
A more objective rubric helps improve the accuracy of grading suggestions which in turn improves the quality of first-pass feedback.
## Steps
{% stepper %}
{% step %}
### Rubrics are key to accuracy!
Whether you are importing an existing rubric from your LMS or using TimelyGrader to create a new one, remember that your rubric sets the flight path for grading suggestions. A clear rubric helps TimelyGrader understand what you value in students’ work. An unclear rubric can lead to inconsistent or inaccurate suggestions.
Creating a good rubric can be challenging—it is an art of its own! Start by checking that each criterion reflects the assignment’s learning outcomes and the instructions students received. Students should understand what they are being assessed on, and TimelyGrader should have clear guidance on what to look for.
Keep each criterion focused on a distinct aspect of performance. For example, separating “Quality of analysis” from “Writing mechanics” makes it easier to assess each fairly. Avoid assessing the same skill under multiple criteria unless that overlap is intentional.
Before using your rubric across a class, try it with a few submissions representing different levels of performance. Review the suggestions and their explanations. If the same misunderstanding appears repeatedly, consider whether the rubric needs clarification.
{% endstep %}
{% step %}
### Balancing too many or too little rating levels
More rating levels can allow finer distinctions between students’ work, but only when those distinctions are meaningful. Too many levels with overlapping descriptions can make it difficult for both instructors and TimelyGrader to choose the most appropriate rating.
We recommend 3 to 5 rating levels as a starting point. Choose the number that matches the distinctions you can clearly describe and reasonably expect to see in students’ work.
For example, a three-level rubric might distinguish between work that meets, partially meets, or does not yet meet expectations. A five-level rubric can provide more detail, provided each additional level describes a recognizable difference in performance.
Avoid adding levels simply to create more scoring options. If “Very good” and “Good” describe essentially the same work, an additional level creates ambiguity rather than precision.
Check that the descriptions progress logically from the lowest to the highest level and that the points assigned reflect those differences.
{% endstep %}
{% step %}
### Fill out rating descriptions
If your rubric only says “Good,” “Average,” and “Poor,” it gives TimelyGrader little information about what those labels mean. This is a common issue we see with rubrics.
Explain your expectations using clear, observable, and measurable descriptions wherever possible. Describe what a grader should be able to identify in the submission, rather than relying on a general impression.
For example, under a “Use of evidence” criterion:
Too vague: “Uses evidence well.”
More specific: “Supports the main claims with relevant evidence, explains how that evidence supports the argument, and addresses important limitations.”
Apply this detail to every rating level, not just the highest one. Lower levels should explain what is missing, underdeveloped, or inaccurate so that the differences are clear.
Numbers can help when they reflect a real requirement, but quantity alone does not establish quality. “Includes three sources” is easy to count; it does not explain whether those sources are relevant or used effectively.
{% hint style="success" %} **Hugo’s tip:** If you cannot tell two bands apart, neither can a grader, and neither can the TimelyGrader.
{% endhint %}
{% endstep %}
{% step %}
### Avoid conflicts and subjective language in your rubric
Each rating level within a criterion should assess the same underlying expectations. The level of achievement should change; the focus of the assessment should remain consistent.
For example, if the highest rating assesses accuracy, use of evidence, and reasoning, the next rating should describe weaker performance in those same areas. Introducing unrelated requirements, such as formatting or presentation, makes the levels difficult to compare.
If higher levels build on the requirements of lower levels, make that explicit. Avoid leaving graders to assume whether an expectation still applies.
Also consider submissions with mixed strengths. A student might provide accurate information but weak supporting evidence. Where this is likely, clarify which aspects matter most or whether they should be assessed as separate criteria.
Subjective phrases such as “strong submission,” “excellent understanding,” or “sufficient detail” need explanation. What would demonstrate those qualities? For example, “excellent understanding” could mean that the student accurately explains the concept, applies it to the scenario, and justifies their conclusion.
Finally, check that the rubric agrees with the assignment instructions and any additional grading guidance. Clear, consistent expectations give students a fairer target and help TimelyGrader produce suggestions that instructors can evaluate more confidently.
{% endstep %}
{% step %}
### For our advanced users...
Something we are learning is that the more you add details to the rubric, the more you are making your assignment paint by numbers. We believe the future lies with designing rubrics that are free of obvious errors and objective but leave some room for interpretation.
This may sound counter intutive but leaving that room gives the students an opportunity to question and think critically about what they are receiving.
{% endstep %}
{% endstepper %}
## Related articles
* [Telling the AI what to look for](/user-guides-get-started-here/grading-settings.md)
* [Designing a cognitive apprenticeship](/user-guides-get-started-here/apprenticeship.md)
* [Language, course details, and defaults](/user-guides-get-started-here/settings-language.md)
**Hugo’s tip:** If you cannot tell two bands apart, neither can a grader, and neither can the TimelyGrader.
{% endhint %}
{% endstep %}
{% step %}
### Avoid conflicts and subjective language in your rubric
Each rating level within a criterion should assess the same underlying expectations. The level of achievement should change; the focus of the assessment should remain consistent.
For example, if the highest rating assesses accuracy, use of evidence, and reasoning, the next rating should describe weaker performance in those same areas. Introducing unrelated requirements, such as formatting or presentation, makes the levels difficult to compare.
If higher levels build on the requirements of lower levels, make that explicit. Avoid leaving graders to assume whether an expectation still applies.
Also consider submissions with mixed strengths. A student might provide accurate information but weak supporting evidence. Where this is likely, clarify which aspects matter most or whether they should be assessed as separate criteria.
Subjective phrases such as “strong submission,” “excellent understanding,” or “sufficient detail” need explanation. What would demonstrate those qualities? For example, “excellent understanding” could mean that the student accurately explains the concept, applies it to the scenario, and justifies their conclusion.
Finally, check that the rubric agrees with the assignment instructions and any additional grading guidance. Clear, consistent expectations give students a fairer target and help TimelyGrader produce suggestions that instructors can evaluate more confidently.
{% endstep %}
{% step %}
### For our advanced users...
Something we are learning is that the more you add details to the rubric, the more you are making your assignment paint by numbers. We believe the future lies with designing rubrics that are free of obvious errors and objective but leave some room for interpretation.
This may sound counter intutive but leaving that room gives the students an opportunity to question and think critically about what they are receiving.
{% endstep %}
{% endstepper %}
## Related articles
* [Telling the AI what to look for](/user-guides-get-started-here/grading-settings.md)
* [Designing a cognitive apprenticeship](/user-guides-get-started-here/apprenticeship.md)
* [Language, course details, and defaults](/user-guides-get-started-here/settings-language.md)