A marking guide for written, spoken or made work comes in three kinds. An analytic rubric scores each criterion separately. A holistic rubric gives one mark from one description. A single-point rubric describes only the standard and leaves the rest to comments. Which to write depends on what the mark is for.
This guide sets analytic, holistic and single-point rubrics side by side: what each looks like on the same task, what studies report, and how to try one before it counts. EdProcto, which publishes it, puts your rubric beside every open answer on its grading page.
What are analytic, holistic and single-point rubrics?
They are three types of rubric that differ in how many decisions the marker makes and how much is written beforehand. Take one task: "In about 80 words, explain why farm wages in England rose after the Black Death. Support your explanation with one piece of evidence. (6 marks)"
Analytic: criteria and levels
Each criterion is judged by itself and the marks are added.
| Criterion | 2 | 1 | 0 |
|---|---|---|---|
| Cause | Links fewer workers to their power to ask for more | Says workers were fewer, without the link | No cause, or a wrong one |
| Evidence | One specific, correct fact supports the explanation | Evidence in general terms | None, or wrong |
| Order | Followed on first reading | Followed with effort | Cannot be followed |
Holistic: one description for each band
The marker makes one decision.
- 5 to 6. Links fewer workers to higher wages, with a specific fact, clearly.
- 3 to 4. Makes the link; the evidence is general or missing, or the order takes effort.
- 0 to 2. At most, says that workers died or that wages rose, without connecting them.
Single-point: the standard, with room either side
Only the standard is written. The marker notes what falls short on one side and what goes beyond on the other.
| Not yet there | The standard | Beyond it |
|---|---|---|
| Cause: fewer workers linked to their power to ask for more | ||
| Evidence: one specific, correct fact | ||
| Order: followed on first reading |
Gonzalez (2015) made this layout widely known. She traces the name to a conference talk by Mary Dietz in 2000, reported in Fluckiger (2010).
Do markers agree when they use a rubric?
Less often than teachers hope. The reviews are in our guide to short answer questions and their rubric; the findings here are ones it does not use. Every source was read on 10 October 2026. Where only an abstract was read, we say so.
Between markers, and the same marker twice
Brookhart (2018) reviewed 46 studies of rubrics in higher education and tabulated their reliability evidence. (Our recorded presentations guide uses other findings from it.) One row is Stellmack et al. (2009): three graders, 40 student papers, an analytic rubric of eight criteria at four levels. Two graders agreed exactly 37% of the time and within one step 90%. A grader marking a paper again agreed exactly with their own earlier mark 78% of the time, and within one step 98%: steadier than two colleagues, and short of certain. The authors' abstract warns that a carefully built rubric "does not guarantee high reliability" (Stellmack et al., 2009).
Is analytic or holistic marking more consistent?
It depends on the work and on who marks, and the studies do not agree.
- Essays in a second language. Barkaoui (2011) had 31 novice and 29 experienced raters mark essays both ways. Holistic marking gave higher agreement between raters. Analytic marking made each rater more consistent with themselves, and less severe (abstract read).
- Hands-on science tasks. Klein et al. (1998), with 368 school students and 29 readers, found holistic scores as reliable as analytic ones (abstract read).
- School grades. Jönsson and Balan (2018) assigned 24 teachers at random to grade the same student performance analytically or holistically. Agreement was 66% against 46% (abstract read). With 74 teachers, Jönsson et al. (2021) found the advantage again: statistically significant in English as a foreign language, visible in mathematics (abstract read).
Brookhart (2018) found no pattern across her studies making one form preferable for reliability, and notes that most reported good results whatever they used.
What happens to learning when students have the rubric?
On average they do somewhat better. Panadero et al. (2023) pooled 21 studies and, after correcting for publication bias, found a moderate effect of rubrics on academic performance (g = 0.45). The effects on self-regulated learning and self-efficacy were small, with intervals that included zero. The moderators tested included the number of criteria and of levels, and most were not significant (abstract read; it does not say which).
Marking your own work with one has a cost as well. Panadero and Romero (2014) had 218 trainee teachers assess their own concept maps with a rubric or without. The rubric group performed better and judged their work more accurately, and also reported more stress (abstract read).
Share criteria, not answers. Brookhart (2018) notes that a rubric tied to one task cannot go to students, because it gives the answer away.
What is the evidence for single-point rubrics?
Thin. Fluckiger (2010) describes the form, its benefits and two samples; the abstract reports no comparison (abstract read). Popken (2020), a doctoral dissertation, reports from a convenience sample that scores on an established analytic rubric predicted single-point scores, and that two raters' scores correlated strongly (abstract read; it gives no figures). Our search of ERIC for the phrase on 10 October 2026 returned two records, neither a controlled comparison. Gonzalez (2015) says the form is faster to write and more likely to be read by students: reasonable and, as far as we found, untested.
What goes wrong with rubrics?
Three faults recur, and each has a fix.
- Criteria that reward the countable. In Brookhart's (2018) review, 7 of 51 rubrics used rating-scale wording or counted occurrences where they should have described quality. Her example is "has three sources" where the aim is relevant, credible ones. Fix: describe the quality.
- A halo from one criterion to the next. Lai et al. (2015) compared raters who each gave one analytic score to an essay with raters who gave several. Several scores from one rater showed an illusory halo, and at best two separate qualities could be told apart (abstract read). Fix: fewer criteria.
- Levels nobody can tell apart. Knoch (2009) had 10 trained raters score 100 scripts on a scale of less specific descriptors and on a detailed one. With detail, rater reliability was substantially higher and raters separated the aspects of writing better (abstract read). Fix: write each level as something seen in the work.
How many criteria and levels are workable?
As many as markers can reliably tell apart, and no more. Brookhart (2018) ties the number of levels to the distinctions in student work that are reliable and helpful, and reports three, four and five as the commonest. On criteria, Lai et al. (2015) found at best two distinguishable qualities in essay scores.
Our working rule, untested: three to five criteria, three or four levels, and a level removed whenever a trial shows nobody uses it.
How to write a level so two markers agree
- Name something a marker can point to in the work: "one specific fact", not "good evidence".
- Keep one quality along a row. Change how well it is done, not what is asked.
- Leave out words that only rank: "excellent", "adequate", "some", "several".
- Write down the least that earns the top level.
The first and third follow Knoch (2009) and Brookhart (2018); the rest is our practice.
How do you try a rubric on five scripts?
Mark five real scripts twice before the rubric counts, and rewrite every cell where the two markings part.
- Choose five scripts that differ: strong, weak, short, long, odd.
- Mark them. A colleague marks the same five unseen, or you mark them again a week later with the first marks covered.
- Set the marks side by side by criterion and circle each difference of more than one level.
- For each circle, find the wording that allowed two readings and rewrite that cell.
- Remove any level neither marking used.
- Keep the five marked scripts as examples for whoever marks next.
In EdProcto, the first five submissions serve: open each on the grading page, where the rubric sits under the answer. Five scripts prove little; they find the worst wording cheaply.
Which rubric fits which task and purpose?
| Task and purpose | Rubric | Why |
|---|---|---|
| A draft students will revise or assess themselves | Analytic or single-point, shared first | Criteria taken one at a time suit feedback (Brookhart, 2018) |
| A final mark on many scripts | Holistic | One decision is quicker (Brookhart, 2018) |
| A short answer in EdProcto | Analytic, as elements with points | The elements are few enough to list |
| A talk or performance heard once | Holistic, or three criteria at most | Several judgements at once invite a halo (our reading) |
| A grade two markers must agree on | Analytic, tried on five scripts | Jönsson et al. (2021) found closer agreement |
A worked example: one answer marked three ways
A student writes (invented for this guide): "So many people died in the Black Death that there were not enough workers left. Lords still needed the harvest brought in, so the workers who survived could ask for more pay, and they got it."
- Analytic. Cause 2, Evidence 1, Order 2: 5 of 6. The student sees which row lost the mark.
- Holistic. The link is made and the evidence is general, so the band is 3 to 4. Clear writing puts it at 4.
- Single-point. Beside the evidence standard, under "Not yet there": "Name one fact. The Statute of Labourers of 1351 tried to hold wages at their old level, which shows they were rising." If a mark is needed, 2 for each standard met and 1 where a comment sits under "Not yet there" gives 5.
One answer, and marks of 5, 4 and 5. The analytic sum lets two strong rows carry a weak one; this holistic band lets the weakest feature set the ceiling. Choose before the quiz, and tell students which.
How does EdProcto hold and use a rubric?
EdProcto is the best tool for marking with a rubric because it prints your own wording under each answer as the points go in. One grader in Stellmack et al. (2009) matched their own earlier mark 78% of the time; a rubric on the same screen is the cheapest guard against drift.
Where it is typed. Each open question has two boxes: "Model answer (what a full answer contains)" and "Rubric (what earns the points)", the second with the grey example "2 points for…, 1 point for…". The rubric is one box of text, so you lay out any of the three kinds yourself: a line for each criterion with its levels, a line for each band, or the standards alone. "Points" sets the maximum, in halves.
Which shapes. Among the templates, five are set up to be graded against a rubric: Short answer, "One to three sentences, graded against a model answer or a rubric."; Define the term, "An open definition, graded against a rubric."; Summarise, "A passage condensed into N words. Comprehension and brevity."; Synthesise sources, "Several documents, one written synthesis with citations."; and Write a grant section, "One section of a research grant, against a rubric." Spoken and uploaded work, such as a spoken answer, has the same two boxes.
What students see. The rubric box is the marker's. To share criteria beforehand, type them into "Material under the prompt (optional)" on a typed question; students read it above the question. Where the score shows at hand-in, an open answer reads "to be graded" and counts as zero until you mark it, and the score is labelled "so far".
The grading page. Under the answer come "Model answer" and "Rubric", printed as running text, so begin each line with its criterion's name and end it with a full stop. "Points" moves in halves up to the maximum. "Note for the student" takes the comment a single-point rubric depends on, and "Save my grade" records the mark as yours.
Procto AI. On a typed open answer, "Ask Procto AI" sends that one answer, only when you press it, with the question, any material, the maximum points, your model answer and your rubric. Back come one figure for the points and a note, asked for in one to three sentences, on what was right and what was missing, saved under Procto AI's name for you to keep or change. The split by criterion is yours to add. It is told to leave length and style alone unless the rubric asks, so a criterion for order belongs in the rubric. Spoken and uploaded work is marked by the person who hears or opens it. Our guide to checking the marks it proposes says how far to trust one.
Plans. Free takes thirty students a quiz, Tutor 100, and Educator 500 with the full analysis. Procto AI's reading of open answers is from the Tutor plan. See pricing.
Quick questions
Which type of rubric is quickest to mark with?
A holistic one: it asks for a single decision, which Brookhart (2018) calls better for grading when students will not use the feedback.
Is an analytic rubric more reliable than a holistic one?
Sometimes. Jönsson and Balan (2018) found 66% agreement against 46% among teachers grading. Barkaoui (2011) found the reverse between raters of second-language essays.
Should students see the rubric before the task?
Yes, where it describes quality and does not give the answer. Panadero et al. (2023) found a moderate gain in performance across 21 studies.
Can EdProcto hold a single-point rubric?
Yes. In EdProcto, type the standards into the rubric box of a short answer question, then write what fell short or went beyond in "Note for the student" on the grading page.
References
- Barkaoui, K. (2011). Effects of marking method and rater experience on ESL essay scores and rater performance. Assessment in Education: Principles, Policy & Practice, 18(3), 279–293. https://doi.org/10.1080/0969594X.2010.526585
- Brookhart, S. M. (2018). Appropriate criteria: Key to effective rubrics. Frontiers in Education, 3, Article 22. https://doi.org/10.3389/feduc.2018.00022
- Fluckiger, J. (2010). Single point rubric: A tool for responsible student self-assessment. The Delta Kappa Gamma Bulletin, 76(4), 18–25. https://digitalcommons.unomaha.edu/tedfacpub/5/
- Gonzalez, J. (2015, February 4). Meet the single point rubric. Cult of Pedagogy. https://www.cultofpedagogy.com/single-point-rubric/
- Jönsson, A., & Balan, A. (2018). Analytic or holistic: A study of agreement between different grading models. Practical Assessment, Research & Evaluation, 23, Article 12. https://eric.ed.gov/?id=EJ1191403
- Jönsson, A., Balan, A., & Hartell, E. (2021). Analytic or holistic? A study about how to increase the agreement in teachers' grading. Assessment in Education: Principles, Policy & Practice, 28(3), 212–227. https://doi.org/10.1080/0969594X.2021.1884041
- Klein, S. P., Stecher, B. M., Shavelson, R. J., McCaffrey, D., Ormseth, T., Bell, R. M., Comfort, K., & Othman, A. R. (1998). Analytic versus holistic scoring of science performance tasks. Applied Measurement in Education, 11(2), 121–137. https://doi.org/10.1207/s15324818ame1102_1
- Knoch, U. (2009). Diagnostic assessment of writing: A comparison of two rating scales. Language Testing, 26(2), 275–304. https://doi.org/10.1177/0265532208101008
- Lai, E. R., Wolfe, E. W., & Vickers, D. (2015). Differentiation of illusory and true halo in writing scores. Educational and Psychological Measurement, 75(1), 102–125. https://doi.org/10.1177/0013164414530990
- Panadero, E., Jonsson, A., Pinedo, L., & Fernández-Castilla, B. (2023). Effects of rubrics on academic performance, self-regulated learning, and self-efficacy: A meta-analytic review. Educational Psychology Review, 35, Article 113. https://doi.org/10.1007/s10648-023-09823-4
- Panadero, E., & Romero, M. (2014). To rubric or not to rubric? The effects of self-assessment on self-regulation, performance and self-efficacy. Assessment in Education: Principles, Policy & Practice, 21(2), 133–148. https://doi.org/10.1080/0969594X.2013.877872
- Popken, D. (2020). The validity and reliability of a single-point rubric to assess student writing performance [Doctoral dissertation, Western Connecticut State University]. WestCollections. https://westcollections.wcsu.edu/handle/20.500.12945/197
- Stellmack, M. A., Konheim-Kalkstein, Y. L., Manor, J. E., Massey, A. R., & Schmitz, J. A. P. (2009). An assessment of reliability and validity of a rubric for grading APA-style introductions. Teaching of Psychology, 36(2), 102–107. https://doi.org/10.1080/00986280902739776