Skip to content

Capstone Supervision

Instructor material

This page is for instructors and teaching assistants supervising capstone projects. Students should use the capstone brief and milestones and submission instead.

What supervision has to do here

The capstone assesses integration, and that creates a supervision problem the labs do not have: a project can be going well or badly for weeks before the artefacts show it. The instructor's leverage is at the checkpoints, not in the final marking, and the highest-value intervention is almost always the same one — making sure the student delivers a small working increment that someone else uses, early.

Checkpoint What to look for The intervention if it is missing
M0–M1 (scope and charter) Explicit exclusions; success criteria someone else could observe Push back on scope now. Every capstone that fails in week 8 chose an unfinishable scope in week 1
M2 (plan baselined) Schedule with float and a critical path; estimates with a basis and a range Ask one question: "which activity has float, and what happens to the date if the longest one doubles?"
M3 (first increment used) A real user, on realistic data, with a note of what they said This is the checkpoint to enforce hardest. Deferring M3 is the most common way a capstone drifts into documentation
M4 (reports against baseline) Reports written during delivery, containing measured variance A report with no variance is a finding: either nothing was measured, or the reporting is sanitised
M5 (change decided) Impact assessed against the untouched baseline; options costed Ask for the option the student rejected and why — it tests whether a comparison happened
M6 (close) Comparison with the original plan; lessons with owned changes Look for the not-delivered list. An empty one is almost always wrong

Using the rubric

The capstone rubric is the same instrument the student self-marks with, deliberately, so that a student can predict their result. Three consequences for supervision:

  1. Point students at what is not assessed before they start. Software ambition, hours spent, and formatting beyond professional adequacy earn nothing, and students routinely spend their effort there.
  2. Ask for the citation, not the claim. Each dimension is evidenced by an artefact; where a student cannot point to one, that is the feedback.
  3. Treat the self-assessment as evidence. An over-generous self-assessment with no citations is a finding in itself, and it belongs in the feedback rather than in the total.

The five patterns worth intervening on

Pattern How it presents Why it matters What to do
Documentation replaces delivery Artefacts improve; no usable increment appears The management record is the vehicle, not the subject; there is nothing to manage without a product Move the next checkpoint to M3 and require a user session before any further document work
Scope discovered rather than chosen The WBS grows; exclusions shrink The student is learning scope creep by experiencing it, which is useful — provided it is then handled Ask whether the change was recorded as a change, and whether the plan moved
Sanitised reporting Green reports, moved milestones Removes the option to act, and removes the evidence the monitoring dimension depends on Ask what the student would have done differently if the report had been Red in week 5
Late discovery of the data dependency Performance or quality work compressed at the end A recurring, teachable failure across the whole course Ask when realistic data became available, and what criterion would have surfaced it
Lessons as generalities "Communicate better"; no artefact cited Nothing transfers, which is the point of the exercise Require one lesson routed into a template or checklist the next project would use

The third and fourth patterns are worth discussing with the whole cohort, because they recur in the labs and in real projects alike. The advanced practice track covers them systematically — Lesson 4 for correlated risk and Lesson 6 for recovery.

Group capstones

Group work is permitted (option E in the capstone options), and it changes what can be assessed:

  • Assess individual contribution through ownership. Require each student to own named artefacts and sign them. The RACI matrix is the mechanism the course already teaches.
  • Require a conflict record. A team that reports no disagreement has either had none — unlikely — or has not recorded it. The decision log is where disagreement that was escalated belongs, and reviewing it is how you assess individual judgement rather than team output.
  • Watch for one student producing all the artefacts. Check the change log, decision log, and reports for authorship across the team.
  • Do not scale the scope with the team size. A four-person team attempting four times the scope will produce four times the documentation risk and no more evidence of management skill.

Academic integrity, given that everything is published

The answer keys and the rubric are public. That is a deliberate design choice — it makes the course usable for self-study — and it has three practical implications for supervision:

  1. Do not use the published quizzes for graded assessment. They are low-stakes self-checks; their answers are on the site. Write your own variants if you need graded questions.
  2. Grade the artefacts, not the answers. The capstone produces a project record, which is individual work in a way a quiz is not — and the consistency checks make a fabricated record hard to assemble, because eight documents must agree with one another.
  3. Say all of this to students. A student who knows that keys are public and that the rubric is shared will not waste effort trying to guess what the marker wants, and will spend it on the project instead.

Feedback that helps

  • Ask a question rather than issuing a verdict. "Who would use this report, and what would they do after reading it?" moves a student further than "too vague".
  • Reward early disclosure of bad news, explicitly, in the mark and in the conversation. It is the habit the whole course is trying to build, and it is the one that a grade can most easily punish by accident.
  • Point at the artefact and the section, not the overall impression.
  • Name one next action. Feedback that ends with a single concrete change is acted on; a list of five is not.