A student can get the "right" number in a lab report and still have no idea why the experiment worked. Another student can botch the arithmetic and still nail the actual scientific reasoning. If your rubric only checks the final value, you're grading a calculator, not a scientist.
I used to grade lab reports by scanning for the expected numbers and docking points when they were off. It was fast, and it was also teaching my students the wrong lesson: that the goal was landing on the right number, not understanding what the number meant.
Separate "did the math work" from "did they understand what they measured"
These are two different skills and they deserve two different rows on your rubric:
- Procedure and data — did they follow the steps, record what actually happened (not what they expected to happen), and show their calculations?
- Reasoning — can they explain *why* the result came out the way it did, connect it to the underlying concept, and account for sources of error?
A student who gets a wrong number but writes "our thermometer probably wasn't fully submerged, which would explain why our temperature reading came in lower than expected" understands more chemistry than a student who got the "right" number by luck.
Drafting the rubric with AI, then editing hard
I don't let a model write my rubric from scratch and use it as-is — it tends to produce generic academic language that doesn't match how I actually talk to my students. What works is using it to generate a first draft against specific criteria, then rewriting the language myself:
1. Give it the actual lab (not just the topic): "This is a titration lab where students determine the concentration of an unknown acid. Draft a 4-level rubric (data collection, calculation accuracy, reasoning/error analysis, conclusion) for 9th grade chemistry."
2. Read every row out loud. If it doesn't sound like something you'd actually say to a student's face, rewrite it.
3. Add one example per level, in your own words, of what that level of reasoning actually looks like on this specific lab.
What a reasoning row should reward
Some concrete examples of the kind of writing that should score well on a "reasoning" row, regardless of whether the final number was exactly right:
- Naming a specific source of error and explaining its likely direction of effect (not just "human error")
- Comparing their result to a known or expected value and explaining the gap
- Connecting the result back to the concept the lab was supposed to demonstrate
Where this gets faster
Building the initial rubric structure and level descriptions from scratch every time you introduce a new lab eats an evening you don't have. Our [grading rubric generator](/grading-rubric) can produce that first-draft structure fast, and our [assessment blueprint tool](/assessment-blueprint) helps you map out ahead of time which labs across the semester are actually testing reasoning versus just procedure-following, so you're not accidentally grading the same skill five times and calling it variety.
The number in a lab report is the easiest thing to check and the least interesting thing about the experiment. Build your rubric so a student can't hide bad reasoning behind a lucky number, and can't lose credit for good reasoning because of one bad measurement.
