See the work behind
the grade.
Your students write the report, build the deck and argue about it here. So when you mark it, who did what is already on file.
You mark five names and hope.
One document, one mark, and no way to tell who wrote it. You find out in week eleven, if at all. The students who did the work already know, and they are the ones who complain.
You can't see contribution
A finished document tells you nothing about who wrote it, who fixed it, or who never opened it.
Problems surface too late
By the time a group falls apart, the deadline has gone. There was nothing to see until there was nothing to do.
Grading comes down to one signal
Peer ratings on their own drift: scores inflate the moment a grade is attached, and friendships move them. With nothing independent to read them against, a fair individual mark is guesswork.
Seven screens. All of these are real.
Not mockups. Click them.
The group writes its own rules and signs them.
Five questions. Meetings, deadlines, what finished work looks like, and what they do if someone stops turning up. Edit the wording later and every signature clears.
You can see who wrote what, as they write it.
Every sentence in the document is traced to whoever typed it. No form, no survey, nothing for anyone to remember to do.
Peer review runs more than once.
Only the last round counts towards a mark. Someone rated badly in week five who then picks up is marked on what they did after the warning, not before it.
Ruby sits well below the group. Round one stays on the record and carries no grade weight.
The default. Nobody can respond to a score they have not seen, so round two rates the work rather than the last round.
One rating each, and they have to say why.
Not five boxes. The criteria sit above it, including the ones the group added themselves. Nobody can submit without writing a sentence.
- Contributing to the team's work
- Interacting with teammates
- Keeping the team on track
- Expecting quality
- Bringing relevant knowledge and skills
- Turns up to the lab sessions, not just the write-up
- Answers in the group chat within a day
The group can escalate. It cannot punish.
Nothing counts against a student until you have read the record and spoken to them. Try issuing the card without ticking the box.
A member raises it, in their own words. Nothing happens automatically and no number can trigger this.
The record, then the student. This step is not optional and the system enforces it.
Marks at risk, and liftable the moment contribution resumes. It exists to change what happens next.
After a written investigation. Removal and a mark of zero, with the rest of the group not disadvantaged.
Two signals, side by side, and a mark you set.
What the document says, and what the group says. Dwixel suggests a factor. You decide the mark, and you can see every number behind it.
All of it. Including what it cannot see.
Their share, how far a peer rating can move their mark, and a plain list of everything worked out about them.
You are about 13 points below an even share. That is worth knowing while there is still time to change it, and it is not a judgement: this measure cannot see thinking, arguing, organising, or work you did offline.
At most +25% and at most −50% of the group's mark. Those two are deliberately different: there is more room to fall than to rise, which is worth knowing before you decide how much to worry about it.
Every number here came from a paper, and some of them changed the product.
Twenty-nine studies were read end to end and the findings were applied, including the ones that were inconvenient. Here are three places the research told us we had built the wrong thing.
We asked students to score five things. Now they score one.
A meta-analysis of peer assessment found that one overall judgement, made by a student who can see the criteria, tracks the tutor's mark far more closely than scoring separate dimensions does. CATME's own validation explains why: their five dimensions correlate at 0.76 with each other. Students were never making five judgements. They were making one and writing it down five times.
So the criteria are still printed on the form, because that is the condition attached to the finding. There is just one box to fill in.
Falchikov & Goldfinch, meta-analysis of 48 studies.
Counting comments rewarded the wrong thing
Twenty replies saying “looks good” beat one person who read the whole draft and said something hard. Comments are now weighted by what they are, and rated for usefulness by the people who received them.
Deleting text was worth almost nothing
About a fifth of edits to a collaborative document reduce its length, and those cuts are usually the work. Editing used to score a twentieth of what writing did. It is now worth four times more than it was.
Some of this cannot be fixed by measuring harder
A shared mark makes students share an outcome. It does not make them need each other, and the second is what the evidence on team learning actually supports. That sits in the assignment brief, not in the software, and the site says so.
Every claim on this site carries a citation you can follow, and the ones the evidence would not support were taken off it. If a student appeals, that is the difference between a number and a number you can stand behind. What we do not claim →
Built to run at the department and institution level.
Dwixel is governed at the institutional level: coordinators and IT keep control, faculty get oversight, and students just open a document. Privacy and roles are enforced in the database, not by convention.
Privacy by design
Instructors can review the group’s work and see who contributed what. Peer review stays confidential, work is never exposed to other groups or the public, and access is enforced by row-level security.
Role-based oversight
Organisation, department, faculty and student roles, separated at the database level. Coordinators oversee only their own classes.
Evidence & recovery
Locked submission snapshots, attributed version history, and automatic content backups — defensible records that survive accidental loss.
Programme analytics
Department-level views of engagement, group health, and students at risk of falling behind.
On the roadmap: SSO / SCIM provisioning and LTI 1.3 grade passback for Canvas, Moodle, Blackboard and D2L.
Bring accountability
to group work.
See a walkthrough with your own course structure. We'll show oversight, early-warning, and evidence in action — in 30 minutes.