OneRange Vero

Final Assessments: types, rubrics, and when to use each

A complete tour of OneRange Vero's four assessment formats — quiz, graded lab, code project, and AI role-play — with guidance on picking the right one for your content

For administrators and course authors · 12 min read

  • What a final assessment is — The graded checkpoint at the end of a course
  • The four assessment types — Quiz, Graded Lab, Code Project, and AI Role-Play
  • Choosing the right format — A quick chooser by what learners need to do on the job
  • Designing a strong rubric — What makes AI grading reliable
  • Editing, versioning, and retakes — What changes when you update an assessment
  • Worked examples by content type — Real pairings of content and format
  • Common pitfalls — Mistakes to avoid when rolling out assessments

1. What a final assessment is

The graded checkpoint at the end of a course

Every Vero course ends in a final assessment. It's the moment learners prove they can actually do what the course taught — not just that they sat through it. The author defines a rubric, the learner submits work (or answers questions), and Vero grades server-side against that rubric. Results are pass / fail / retry, with a per-criterion breakdown the learner and their manager can both see.

Final assessments are pluggable. The same course can use a multiple-choice quiz, a hands-on graded lab, a code project, or an AI role-play — whichever best matches the skill you're certifying. You can swap formats on an existing course at any time; in-flight learners keep the version they started on, and new assignments pick up the new format.

Tip: Pick the format that mirrors the real task. If the job is a conversation, assess with a role-play. If the job is shipping code, assess with a code project. Quizzes are best when knowing it cold matters more than performing it.

2. The four assessment types

Quiz, Graded Lab, Code Project, and AI Role-Play

Quiz (multiple choice)

A short bank of multiple-choice questions, auto-scored against a passing threshold (default 70%). Fastest to author, fastest to grade, and the format learners are most familiar with.

  • Best for: policies, definitions, regulatory facts, safety steps, compliance attestation
  • Strengths: instant grading, easy to retake, scales to thousands of learners with zero admin overhead
  • Limits: tests recall and recognition, not performance. Easy to pass without true understanding if questions are weak

Graded Lab

A free-response task graded by AI against a rubric you write. The learner submits a written deliverable — a plan, an analysis, a memo, a customer reply, a triage decision — and Vero scores each criterion separately with feedback.

  • Best for: judgement-heavy work — drafting customer comms, writing an incident summary, prioritising a backlog, structuring a proposal
  • Strengths: assesses applied thinking, not just facts. Generates coachable feedback per criterion
  • Limits: only as good as your rubric. Vague criteria produce vague grading

Code Project

The learner edits code in an in-browser Monaco IDE and submits a working solution. Vero evaluates against rubric criteria such as correctness, structure, edge-case handling, and style.

  • Best for: engineering onboarding, SQL fluency, data-pipeline tasks, bug-fix drills, infra-as-code review
  • Strengths: tests real building, not pseudocode. Learners can iterate before submitting
  • Limits: needs a self-contained problem statement and a rubric that doesn't depend on running the code in your production stack

AI Role-Play

The learner has a live conversation with an AI playing a scenario character — an unhappy customer, a skeptical exec, a struggling report. Vero grades the full transcript against the rubric once the learner ends the session.

  • Best for: sales objection handling, manager 1:1s, support escalations, executive briefings, performance conversations
  • Strengths: only format that tests dialogue in real time. Captures tone, pacing, and recovery — not just the right answer
  • Limits: takes longer per learner. Best for high-stakes soft skills where a quiz would feel hollow

3. Choosing the right format

A quick chooser by what learners need to do on the job

They need to know it coldQuiz — fast recall, easy retake
They need to do it on the jobGraded Lab — written artefact graded by rubric
They need to build or fix somethingCode Project — IDE submission, scored on correctness
They need to handle a conversationAI Role-Play — live transcript graded end-to-end
Mixed (knowledge + judgement)Graded Lab. Don't stack two assessments on one course

Tip: If you're tempted to use a quiz for a skill that's really about doing — coding, talking to customers, writing a plan — pick the matching performance format instead. Pass rates will look lower at first, but the people who pass will actually be ready.

4. Designing a strong rubric

What makes AI grading reliable

Graded Lab, Code Project, and AI Role-Play all share the same rubric model: 2–5 criteria, each weighted 1–5, each described in one or two sentences of what "good" looks like. Vero scores every criterion independently, then aggregates against your pass threshold.

  • Keep it to 2–5 criteria. More than 5 and grading gets noisy
  • Make each criterion observable in the submission — not in the learner's head
  • Weight by what matters most. A safety criterion can outweigh a stylistic one 5:1
  • Write the "good" description in the voice of a reviewer, not a textbook
CorrectnessSolution returns the expected output for the sample and at least two edge cases
Customer empathyAcknowledges the customer's frustration before offering a fix; avoids defensive language
StructureOpens with the recommendation, supports it with 2–3 reasons, closes with a clear next step

Tip: Pilot a new rubric with 3–5 learners before rolling it out widely. Read the per-criterion feedback and tighten any criterion where Vero's reasoning surprises you

5. Editing, versioning, and retakes

What changes when you update an assessment

You can change the assessment format or rubric on a live course at any time from the Assessment format & rubric card on the course editor. Saving bumps the course's version. Learners already mid-course keep the pinned version they started; new assignments pick up the latest.

  • Past results stay interpretable — each result is stamped with the assessment version that graded it
  • Retakes use the version the learner is currently on. They don't jump forward mid-course
  • Pass threshold defaults to 70%. Override per course if your bar is higher (compliance) or lower (exploratory)

Tip: If you tighten a rubric significantly, consider re-assigning the course rather than editing in place — it gives existing learners a clean shot at the new bar instead of grading them against a rubric they never saw

6. Worked examples by content type

Real pairings of content and format

New-hire security trainingQuiz — knowing the policy cold is the goal
Incident write-up skills (SRE)Graded Lab — submit a written post-mortem, rubric scores clarity, blameless tone, action items
SQL onboarding for analystsCode Project — write 3 queries against a sample schema
Manager: giving difficult feedbackAI Role-Play — converse with a struggling report, graded on framing, specificity, and follow-through
Sales: discovery callAI Role-Play — graded on open questions, pain identification, and next-step proposal
Product spec writingGraded Lab — submit a one-page spec, rubric scores problem framing, scope, and success metrics
Bug-fix drill for engineersCode Project — broken test suite, learner makes it green
HIPAA refresherQuiz — short, annual, auditable

7. Common pitfalls

Mistakes to avoid when rolling out assessments

  • Using a quiz to certify a performance skill. If the job is talking to customers, a role-play is the assessment — full stop
  • Writing rubric criteria that overlap. Two criteria measuring the same thing inflate scores and confuse feedback
  • Setting the pass threshold too low to look generous. A 50% pass is a credential learners and managers stop trusting
  • Editing a rubric mid-cohort without telling learners. Bump the version and communicate the change
  • Stacking multiple assessments on one course. Vero supports one final assessment per course by design — split the content into two courses instead

Tip: When in doubt, ask: "If a learner passes this, would I be comfortable putting them on the live task tomorrow?" If not, the assessment isn't measuring the right thing yet