Skip to article content

How to create online assessments

A practical guide to planning, building, and improving online assessments.

Key Takeaways

  1. Start with decisions, not questions: Define purpose, audience, constraints, and how results will be used before you build anything.
  2. Use an assessment blueprint: Map learning objectives to item types and difficulty (Bloom's levels) so your test is valid and balanced.
  3. Design for fairness and access: Clear instructions, accessible formats, and sensible timing matter as much as question quality.
  4. Integrity is a design problem: Use question design, randomization, and allowed-resource rules before jumping to proctoring.
  5. Close the loop with analytics: Review item performance (missed questions, discrimination, time-on-item) and revise for the next cohort.
Papercraft illustration of a student taking an online quiz, with a small owl perched on the laptop monitor.
Paper-cut illustration -- learning through quizzes.

What an online assessment is (and what it is not)

An online assessment is any structured set of questions or tasks delivered digitally to measure knowledge, skill, or decision-making. It can be a low-stakes knowledge check, a graded exam, a compliance test, or a skills screening.

What it is not: a random set of questions pulled from memory, or a copy-paste paper test moved online without reconsidering timing, resources, and accessibility.

A useful rule

If you cannot explain what decision you will make from the score (pass/fail, placement, remediation, certification), the assessment is not finished -- even if the questions look good.

The end-to-end workflow (plan -- build -- deliver -- improve)

Most online assessments fail for predictable reasons: unclear objectives, ambiguous questions, unfair settings, technical surprises, and no post-run review. Use this workflow to avoid those problems.

  1. Plan the measurement

    Define purpose, audience, constraints, and scoring decisions. Write an assessment blueprint.

  2. Build the assessment

    Create items aligned to objectives, add instructions, set scoring rules, and create feedback.

  3. Configure delivery settings

    Timing, attempts, randomization, security, accessibility accommodations, and support plan.

  4. Pilot and publish

    Run a small test, fix issues, then share to learners with clear expectations.

  5. Analyze and improve

    Use item and cohort analytics to revise questions and settings for next time.

If you need a hands-on walkthrough for building the assessment in a quiz tool, see how to make an online quiz step by step.

Step 1: Define the goal, stakes, and constraints

Write down these decisions before you open any assessment builder:

  • warning
    Purpose: Diagnostic (placement), formative (practice), summative (grade/certification), or selection (hiring).
  • warning
    Population: Who is taking it (language level, accessibility needs, device constraints, time zones).
  • warning
    Decision rule: What score means pass/fail, mastery, or a next step (remediation, retake, escalation).
  • warning
    Conditions: Open-book vs closed-book, allowed tools (calculator, notes), and collaboration rules.
  • warning
    Logistics: Time limit, number of attempts, window of availability, and grading turnaround.

University teaching centers commonly recommend starting with the course goal and what evidence would demonstrate learning, then choosing the online format that best fits that evidence. See UPenn CETLI guidance on assessments and exams in online courses for a concise set of planning questions.

Step 2: Choose the right assessment format (not just question types)

Pick a format based on what you are measuring and the consequences of being wrong.

Common online assessment formats and when to use them
FormatBest forWatch out for
Low-stakes quiz (5-15 min)Retrieval practice, quick checks, spaced reviewOver-grading; keep feedback fast and simple
Timed test (30-90 min)Fluency and coverage across objectivesAccessibility and tech risk; avoid tricky wording
Open-book / take-home examApplication, analysis, authentic tasksPrompt design must reduce copy/paste answers; clarify allowed resources
Performance task / projectReal-world skill, synthesis, creationNeeds rubric and more grading time
Oral / video responseCommunication, reasoning, languageBandwidth, privacy, and accommodations

If your goal is learning retention (not just measurement), frequent low-stakes quizzes can help because of retrieval practice. A practical explanation is in our guide to the testing effect and retrieval practice benefits.

Need inspiration for structures and use cases? Browse online assessment examples you can copy and adapt to your objectives.

Step 3: Pick a tool that matches your delivery and reporting needs

Your tool choice should follow your plan. The biggest mismatch is using a lightweight form tool for a high-stakes exam (or using a full LMS exam module for a 6-question pulse check).

Two common paths:

  • LMS assessment modules work well when enrollment, gradebook, and course structure already live in an LMS.
  • Standalone quiz/assessment tools work well when you need fast creation, sharing links, embeddable assessments, flexible reporting, or use cases outside a course shell.

For a deeper comparison, see LMS vs standalone quiz tools (and when each wins) and use our quiz tool checklist to choose an assessment tool based on security, accessibility, analytics, and workflow.

Quick decision matrix for assessment platforms
NeedOften best fitWhy
Course enrollment + gradebook syncLMS quiz moduleRoster, gradebook, and course permissions are built-in
Public link sharing or embeddingStandalone quiz toolEasy distribution without course accounts
High-stakes security controlsLMS + security add-ons or proctoringIdentity controls and policies are easier to enforce
Fast authoring + reusable banksStandalone quiz toolStreamlined item banks, duplication, variants, and exports
Advanced analytics to improve itemsTool with item analysisNeed per-question stats and exportable data

Step 4: Build an assessment blueprint (alignment + coverage)

An assessment blueprint is a simple table that maps learning objectives to the number of questions, difficulty, and item types. It is the fastest way to improve validity (measuring what you meant to measure) and reduce accidental bias.

A practical way to structure difficulty is Bloom's taxonomy:

  • Remember / Understand: identify, define, explain
  • Apply: use a method on a new example
  • Analyze / Evaluate / Create: justify, compare, troubleshoot, design
Assessment blueprint example (edit and reuse)
ObjectiveBloom levelItem typeCountPoints eachNotes
Explain the core conceptUnderstandMultiple choice61Include common misconceptions as distractors
Use the procedure correctlyApplyNumeric / short answer42Accept equivalent formats; define rounding rules
Diagnose an error in a scenarioAnalyzeScenario MCQ or multi-select32Require reasoning, not recall
Justify a decisionEvaluateShort essay / rubric16Use a 3-5 criterion rubric to mark consistently
Blueprint sanity checks

If more than ~80% of your points are Remember/Understand, the assessment may not reflect real performance. If more than ~50% is essay/project, plan the grading workload and rubrics first.

Step 5: Write questions that are clear, fair, and gradable

Question quality is the biggest driver of score quality. Before you add security features, fix ambiguity.

For a deeper library of do's and don'ts, use our assessment design tips and quiz best practices.

Multiple choice (MCQ) that actually measures understanding

  • Write the stem as a complete question. Avoid sentence fragments that force test-takers to decode grammar.
  • Make distractors plausible. Use common misconceptions, not silly throwaways.
  • Avoid clues. Keep option lengths similar; avoid "always"/"never" unless defensible.
  • One best answer. If two answers could be correct, rewrite the stem or tighten the options.

Short answer and numeric questions (great for reducing guessing)

  • Define acceptable formats. Example: "Enter to 2 decimal places" or "Use mm/dd/yyyy".
  • Plan for variants. Consider alternative spellings, units, or equivalent expressions.
  • Use tolerance ranges for numeric answers when rounding is not the skill being measured.

If you are drafting quickly and need prompts to adapt, start from ready-to-use question examples and then align them to your objectives and blueprint (do not use generic trivia as a scored requirement without alignment).

Rubric-scored items (essays, uploads, case responses)

Rubric items can increase authenticity, but they must be designed for consistent marking:

  • Use 3-5 criteria. More criteria often adds noise, not fairness.
  • Define levels with observable evidence. "Clear" is vague; "states claim, cites 2 relevant facts" is markable.
  • Run a calibration pass. Mark 5 sample responses first, then tighten rubric wording.

To streamline marking, see auto-grading vs manual grading workflows (and how to mark faster).

Feedback design (especially for formative quizzes)

Good feedback improves learning and reduces support tickets. A simple structure:

  • Correct answer: state it plainly
  • Why: 1-2 sentences
  • Next step: link to the concept or a short practice prompt

Step 6: Make the assessment accessible (so scores reflect skill, not barriers)

Accessibility is not only about compliance; it is about measurement. If a learner cannot perceive the question or operate the interface, you are measuring tool friction, not competence.

  • warning
    Readable layout: Use clear headings, short sentences, and consistent terminology (do not rename concepts mid-test).
  • warning
    Alt text and captions: If you use images, charts, or audio/video, provide text equivalents and captions.
  • warning
    Keyboard support: Ensure navigation and selection work without a mouse.
  • warning
    Color is not the only cue: Do not rely on color alone to indicate correctness or categories.
  • warning
    Timing accommodations: Plan how you will grant extended time (per learner or per group) without breaking fairness.
  • warning
    Plain-language instructions: Put allowed resources, time limit, attempts, and scoring rules at the top.
Accessibility quick test

Ask a colleague to complete the assessment using only a keyboard and zoomed text (200%). If they get stuck, a learner will too.

Step 7: Improve integrity with design choices (before proctoring)

Academic integrity and test security are real concerns, but proctoring is not the only tool. In many contexts, redesigning the assessment reduces cheating incentives and opportunity.

University best-practice guides emphasize clarifying expectations and choosing assessment types that match the online context (for example, open-book questions that require application). See UNF CIRT online assessment best practices for a practical checklist-oriented overview.

Integrity options beyond proctoring (and what each protects)
ApproachWhat it helps withDesign notes
Higher-order promptsReduces "search and paste" answersUse scenarios, justification, and trade-offs; require reasoning steps
Question pools + randomizationReduces answer sharingCreate item variants; randomize order; ensure equal difficulty across variants
Time limits (reasonable)Reduces external lookup timeTime should still allow reading and accessibility accommodations
Open-book policyShifts focus from recall to applicationTell learners what is allowed; write questions that cannot be answered by a definition alone
Honor statement + expectationsSets norms and reduces ambiguityBe specific: collaboration rules, AI/tool use, citations required
Two-stage assessmentsDiscourages last-minute cheatingShort timed quiz + follow-up explanation/reflection

If you do use take-home formats, research syntheses highlight the trade-offs: they can support deeper thinking but require careful prompt design and clear rules about resources and collaboration. See Monteiro et al. (2019) systematic review of take-home exams.

Step 8: Configure delivery settings that are fair and low-friction

Settings determine behavior. They also create the conditions under which your score is interpretable.

Core settings to decide (and document)

  • Availability window: When the assessment opens/closes (consider time zones).
  • Time limit: Base it on a pilot run; avoid guessing. Provide accommodation paths.
  • Attempts: One attempt for certification, multiple for practice (often with highest score or latest score rules).
  • Navigation: Free navigation vs one-question-at-a-time. Locking navigation can reduce answer changing but can harm accessibility and increase anxiety.
  • Randomization: Shuffle order and draw from pools where appropriate.
  • Feedback timing: Immediate feedback for practice; delayed feedback for high-stakes (to reduce item leakage).
Time limit reality check

If a question requires reading a long scenario, a tight time limit measures reading speed and stress tolerance as much as the target skill. Use shorter prompts or allocate more time.

Step 9: Pilot, then publish with clear instructions (reduce support tickets)

Run a small pilot before you send the link to a full cohort. Even a 10-minute pilot with 3-5 people catches most problems: ambiguous wording, broken media, mismatched scoring keys, and timing issues.

Pilot checklist

  • warning
    Try multiple devices: at least one phone, one laptop, one different browser.
  • warning
    Confirm scoring: check correct answers, point weights, and partial credit rules.
  • warning
    Validate timing: record time-to-complete for fast and slow test-takers.
  • warning
    Check accessibility items: keyboard navigation, captions, readable images.
  • warning
    Review instructions: can a new person explain what to do without asking you?

When you publish, include:

  • purpose and stakes
  • time limit, attempts, and deadline
  • allowed resources and collaboration rules (and AI/tool policy if relevant)
  • how to get help if something fails (and what info to provide)

If you need a ready set of setup fixes and common problems (scoring, sharing, access), use assessment troubleshooting and common quiz setup issues.

Step 10: Grade consistently and return feedback on time

Grading is part of assessment design. Make it predictable for learners and manageable for you.

Make grading decisions upfront

  • Auto-graded items: confirm acceptable answers, spelling rules, and partial credit (if used).
  • Rubric items: finalize rubric before launch; avoid changing criteria mid-stream.
  • Retakes: decide whether you keep the best score, average scores, or require remediation first.
  • Feedback timing: immediate for practice; delayed for high-stakes to protect item banks.

For faster workflows (and what to automate vs keep human), see how to mark quizzes faster with auto-grading and rubrics.

Step 11: Use reporting and item analytics to improve the next assessment

After delivery, treat the assessment as data. Your goal is not only a grade, but a better instrument next time.

What to look for in results

Simple item analysis signals and what they usually mean
SignalWhat you might seeLikely causeWhat to do next
Very low correct rateMost learners miss itContent not taught, trick wording, or miskeyCheck key first, then rewrite for clarity or re-teach
Very high correct rateAlmost everyone gets itToo easy or over-practicedKeep for confidence checks, or increase difficulty
High time-on-itemSlow completion vs othersLong reading load, confusing interface, or multi-step skillShorten stem, add data table, or split into two items
Good students miss itTop scorers get it wrong tooAmbiguous stem or multiple plausible answersRewrite and re-pilot; consider removing from scoring
One distractor dominatesMany choose the same wrong optionStrong misconception (useful!)Add targeted feedback and practice; keep item if clear

In higher education contexts, studies of online assessments emphasize that settings and design parameters can influence how well the assessment predicts performance and how reliable the results are. For one example in a professional program context, see Mohanraj et al. (2024) on online assessment effectiveness and parameters.

Close the loop

  • Revise items with ambiguity, miskeys, or unexpected difficulty.
  • Update the blueprint if you under-sampled an objective.
  • Adjust settings if time limits or navigation caused friction.
  • Build a question bank with tagged objectives (topic, Bloom level, difficulty) for easier future assembly.

Special case: personality and scored outcomes (use with caution)

Some online assessments are not knowledge tests; they classify preferences or styles. These can be useful for engagement, onboarding, or reflection, but they require extra care around interpretation and consent.

  • Be clear about purpose: "development" is different from "selection" (hiring).
  • Avoid medical/diagnostic claims unless you have validated instruments and proper governance.
  • Explain outcomes: what the result means (and what it does not mean).

If you are building this type of experience, see personality assessments and when to use personality quizzes to choose scoring and result logic responsibly.

Pre-launch checklist (copy/paste)

  • warning
    Objective coverage: Blueprint matches what was taught and what you intend to measure.
  • warning
    Clarity: Instructions include time, attempts, allowed resources, and how scoring works.
  • warning
    Answer keys: Verified by a second person or by running the assessment yourself end-to-end.
  • warning
    Accessibility: Keyboard navigation works, media has text equivalents, and timing accommodations are possible.
  • warning
    Integrity plan: Randomization/pools where needed, clear policy, feedback timing appropriate for stakes.
  • warning
    Support plan: Learners know what to do if they lose access or hit a technical issue.
  • warning
    Reporting: You can export or view results in the format you need (per learner, per item, per objective).

References

Frequently Asked Questions

quiz How many questions should an online assessment have? expand_more

Use your blueprint rather than a fixed number. Start with: (1) the number of objectives you must sample, (2) the stakes, and (3) how long you can reasonably ask learners to focus. For low-stakes checks, 5-15 questions is often enough. For higher-stakes exams, add questions only when they improve coverage or reliability, not because you need a round number.

quiz Should I make my assessment timed? expand_more

Time limits can reduce answer-sharing and force fluency, but they also increase accessibility risk. If speed is not part of the construct (the skill) you are measuring, use generous timing, remove unnecessary reading load, and plan accommodations.

quiz Are open-book online assessments valid? expand_more

Yes, if the questions require application, analysis, or justification. Open-book does not mean easy; it means your prompts must be designed so learners cannot succeed by copying a definition. Clearly state what resources are allowed and whether collaboration is allowed.

quiz How do I reduce cheating without proctoring? expand_more

Use (1) question pools and randomization, (2) higher-order scenarios that require reasoning, (3) reasonable time limits, (4) delayed feedback for high-stakes assessments, and (5) explicit integrity expectations (including AI/tool rules). Proctoring can be added when stakes require it, but design choices usually deliver the first big gains.

quiz What accessibility accommodations should I plan for? expand_more

At minimum: extended time, alternative formats for media (captions/alt text), keyboard access, and a way to handle interruptions (resume rules). Plan these before launch so accommodations do not become last-minute manual exceptions that feel unfair to everyone.

quiz What metrics should I review after the assessment? expand_more

Review score distribution, per-question correct rate, time-on-item (if available), and which distractors were chosen. Look for signs of ambiguity (top scorers missing an item), miskeys, and objectives that were under-sampled. Use the results to revise the item bank and blueprint for the next run.