How grading works

See how learners reason, not just what they submit

Cartedo assesses the evidence of learning — how students apply concepts, reason, decide, and defend their thinking.

You stay in control of the rubric, standards, and final grade.
Learner submission
Final score
3.5
Every score tied to a rubric & rationale
Grading rationale — why was this rated 3.5 stars
Applied reasoning
Links pricing directly to segment willingness-to-pay.
Evidence use
Cites the data, but omits the competitor benchmark.
Communication
Recommendation stated up front, clearly structured.
Graded in 6s
98.6%
Alignment with instructor grading
Step by step

How Cartedo grades a student submission

You stay in control of the rubric, standards, and final grade.

Get started for FREE
1
Student completes the simulation

They make decisions, submit work, or defend their reasoning in chat or voice.

2
Cartedo grades and gives feedback

Each learner gets a rubric-based score and personalized feedback in real time.

3
Instructor sees the rationale

Every score comes with a criterion-by-criterion explanation of why it was given.

4
Instructor stays in control

Review, edit, override, or export results anytime — you own the final grade.

What Cartedo grades

Cartedo grades the skills traditional assignments miss

Instead of scoring a polished final answer, Cartedo grades open-ended simulations using course-aligned rubrics, clear grading rationales, personalized feedback, and instructor-reviewable reports — so you can assess application, reasoning, and judgment without manually grading every submission.

Concept application

Did the student apply the course concept correctly?

Evidence use

Did they use data, facts, or resources to back the decision?

Reasoning quality

Can they explain trade-offs and assumptions?

Decision-making

Did they make a defensible recommendation?

Feedback response

Did they improve after feedback?

Defense

Can they explain and defend their work under questioning?

BUILT for the AI era

Students can use AI.
Cartedo still checks what they understand.

Cartedo is built for the AI era. Students can use AI where it helps — but they still have to apply concepts, explain decisions, and defend their reasoning. AI use becomes part of the assessment, not something to police.

01 · Collaborate
AI is allowed, and visible

In TEDO GPT simulations learners must use the built-in LLM — so you see how they prompt it, what understanding they add, and where they push back.

02 · Defend
A live defense they can't outsource

Seconds after submitting, they face stakeholders live — in text, voice, or group. Questions are personalized and rapid-fire, with no time to paste an answer.

03 · Reasoning
Judgment is what gets scored

Cartedo looks for evidence, trade-offs, and depth under challenge — rewarding students who can think, not just produce a polished answer.

The goal was never to catch AI use — it's to prepare learners for a workplace built on it. Cartedo helped me assess how well they work with AI, and the feedback helps them get better at exactly that.
Dr. Lisa Troy
Clinical Professor, Texas A&M University
Read Dr. Troy's story
Integrity

Trust, control, and academic integrity

Cartedo is designed for high-stakes academic use. Learners cannot succeed by submitting polished but shallow work — many formats require live defense, revision, AI-process evidence, or stakeholder interaction.

Integrity signals

Deep integrity analysis

  • Reads the right signals — reasoning, response latency, and how learners defend their choices
  • Confirms learners genuinely understand what they submitted
  • Flags shallow or AI-reliant work with a clear verdict for review
{{ igLabel }}Integrity check
You recommended Segment B. But the data shows Segment A has the higher margin — why discount it?
JA
Segment A's margin is higher, but its acquisition cost wipes out the gain within two quarters — so B is more defensible long-term.
Fair. What retention rate would flip your recommendation back to A?
JA
If A held above ~68% at 12 months, its LTV clears B's — then I'd switch.
Submission integrity scan
Integrity signals7 checks
Response latencyNatural
Typing cadenceHuman-like
Reasoning consistencyHigh
Handles pushbackYes
Revision after feedbackPartial
Owns the reasoningVerified
Depth beyond AI outputSome reliance
Verdict
Authentic understanding
Learner owns the reasoning, with light AI reliance flagged for review.
Cohort grading

Scales with judgment

  • Grade open-ended work for 20 or 2,000 learners
  • Consistent rubric application across the entire cohort
  • Instructor judgment stays in the loop — edge cases surfaced for you
Next: View feedback formats
Cohort gradingBUS 340 · Fall
750
submissions graded
Completed & graded Review needed Still working
2 flagged for your review
Edge cases the AI wants a human decision on.
Review
Grading rationale

Grading rationale you can trust

  • The rubric is followed to the letter, every time
  • Each score quotes the learner’s own work and explains what earned it and what was missed
  • Override any grade, anytime
Grading rationaleJ. Alvarez · Marketing
Framework usage
Does the learner rank products by contribution per constraint unit and allocate capacity systematically?

Rated 4 — the artifact ranks all seven items by margin per prep-minute with correct math (Cheddar Melt $5.30÷4 = $1.33/min), and prioritizes top items. Missing: explicit sequential allocation of the 90 minutes.

Score4/ 5
Change grade
98.6%
AI–instructor grading alignment across 30K+ learners
Calibrated by level

The same topic is not graded the same way for every level

Cartedo calibrates ambiguity, resource complexity, stakeholder pressure, and rubric expectations to the learner level — fair for undergraduates and rigorous for advanced learners.

UNDERGRADUATE

More instructions, clearer resources, structured prompts, lower ambiguity.

Rubric looks for

correct application, clear reasoning, use of provided evidence.

GRADUATE / MBA

Moderate ambiguity, multiple variables, cross-functional trade-offs.

Rubric looks for

prioritization logic, contextual judgment, evidence-backed recommendations.

MID-LEVEL PRO

Messy information, incomplete data, conflicting KPIs, stronger scrutiny.

Rubric looks for

strategic judgment, risk awareness, defensible reasoning under pressure.

Tailored rubric

A custom rubric for every simulation — course-aligned, career-relevant, and calibrated for your learners.

Cartedo builds each rubric from the simulation's topic, scenario, and learner level, so what's measured always maps to real workplace performance. Open any simulation to see its summary, then view the exact rubric it's graded against.

{{ card.company }}
{{ card.disc }}
{{ card.title }}
{{ card.levelShort }} View rubric
{{ simSel.eyebrow }}

{{ simSel.title }}

View simulation
{{ simSel.company }}
{{ simSel.company }}
Role: {{ simSel.role }}

{{ simSel.companyBlurb }}

Learner task

{{ simSel.task }}

Simulation details
Discipline{{ simSel.disc }}
Level{{ simSel.level }}
Format{{ simSel.format }}
Est. Duration{{ simSel.duration }}
Graded on a custom rubric

{{ simSel.rubricCount }} criteria, tuned to this topic and calibrated for {{ simSel.levelShort }} learners.

View rubric
View simulation
Generated rubric · {{ simSel.disc }} · {{ simSel.levelShort }}
{{ simSel.rubricCount }} criteria
Written task
Chat defence
Domain / Competencies
Skills
Skills Assessed
{{ rubAllLabel }}
{{ rubBannerTitle }}

{{ rubBannerDesc }}

{{ crit.name }}

Question: {{ crit.question }}

★

{{ band.text }}

Every criterion is written from this simulation's scenario and calibrated to {{ simSel.levelShort }} learners — not a generic skills list.
Feedback

Feedback that sounds like a real manager

Every learner gets specific, rubric-aligned feedback in the format that fits the task — delivered in seconds, at any class size. And when the simulation ends, each learner receives a personalized summary that explains the grading rationale and shows exactly how to improve, with specific examples drawn from their own submission.

Inline comments

Attached to the exact words the learner wrote — ideal for detailed written work.

View example
Quoted email

Quotes the learner's own words, then responds to each specific claim.

View example
Summative email

One concise, workplace-style email — a clear overall verdict after the task.

View example
Personalized summary

A tailored closing note for every learner — automatically, at any class size.

View example
How grading works — full walkthrough
cartedo — grading demo
Sampleillustrative preview
Grading Assistant
Regrade
Grading actions
Regrade with AI
Ask AI to evaluate this submission again.
Grade manually
Review the submission and assign your own grades.
Task 1: TedoGPT
Attempt score
75%
Grading Rationale Feedback Score Breakdown
AI Skills
Domain-guided AI reasoning & judgment
How effectively does the learner frame the task, clarify objectives, and provide the LLM with sufficient context to generate relevant and actionable outputs?

Rated 1 because the learner accepted AI's contribution-per-minute framework and menu recommendations without applying domain-specific judgment about demand distribution, capacity allocation, or ingredient spoilage. The learner did not redirect AI using the 35% Classic Cheddar Melt demand share or 90-minute constraint. Final direction appears model-led rather than learner-led.

{{ gaStar1of5 }}
Critical evaluation & verification
How effectively does the learner question, critique, and refine LLM outputs to identify inaccuracies, gaps, or weaknesses?

Rated 1 because the learner did not check AI's contribution-per-minute calculations, capacity allocation logic, or demand assumptions against the Menu Performance Analysis resource. No verification of whether AI correctly used the 90-minute griddle constraint or office-worker demand percentages. Learner accepted AI output without comparing it to provided materials.

{{ gaStar1of5 }}
Human–AI collaboration
How effectively does the learner leverage the LLM for ideation, exploration, and iterative problem solving?

Rated 3 because the learner did assign AI a working role and fed it the griddle constraint, but leaned on it as an answer engine for the final menu rather than a thinking partner. There is one round of pushback, yet little exploration of alternatives or scenario testing before accepting the recommendation.

{{ gaStar3of5 }}
Domain Skills
Concept accuracy
Does the participant correctly calculate contribution margins using actual ingredient costs and identify prep time as the binding constraint?

Rated 5 because the artifact computes contribution margins for all seven sandwiches from exact ingredient costs ($3.20–$6.90) and selling prices ($8.50–$12.50), then correctly identifies the 90-minute peak-lunch griddle window — not total unit contribution — as the binding constraint, driving product mix on contribution per prep-minute ($0.70–$1.33).

{{ gaStar5of5 }}
Framework usage
Does the participant rank products by contribution per constraint unit and allocate the 90 minutes systematically?

Rated 4 because the memo ranks all seven items by margin per prep-minute with correct math and deprioritizes the lowest earners (Bacon Brie Bliss $0.74/min, Mushroom Truffle Melt $0.70/min). It stops short of showing the explicit sequential allocation of the 90 minutes across ranked quantities.

{{ gaStar4of5 }}
Error recognition
Does the participant identify unrealistic assumptions in AI-generated reasoning about demand forecasts, prep-time consistency, or capacity allocation?

Rated 1 because the artifact accepts its own menu recommendations without questioning assumptions about demand elasticity, prep-time consistency, or capacity allocation. It does not identify the ±1-minute prep-time variability during peak rush documented in operational notes, nor how it affects the 90-minute constraint.

{{ gaStar1of5 }}
Overall FeedbackShared with learner
Overall Feedback

Your memo correctly calculates contribution margins for all seven sandwiches and identifies the 90-minute griddle constraint as the binding capacity limit. However, the work relies entirely on AI-generated reasoning without verification, critical evaluation, or iteration.

Suggestions to Improve
  • Before accepting AI's menu strategy, verify its capacity allocation math against the resource. Check whether AI correctly incorporated the Classic Cheddar Melt demand share and 90-minute constraint when recommending the four-item lunch menu. Use a verification prompt like: "Walk me through how the 90 minutes would be allocated across these four sandwiches given their demand shares and prep times."
  • Identify flawed assumptions in AI reasoning before finalizing recommendations. The resource shows ±1 minute prep-time variability during peak rush, but AI used fixed prep times when calculating contribution per minute. Prompt AI to explore: "How would ±1 minute variability affect the capacity allocation and product rankings?" This builds error recognition into your process.
  • Treat AI as a thinking partner, not an answer engine. After AI produces initial rankings, ask it to critique its own logic: "What demand or operational assumptions in this strategy might be unrealistic?" or "How would the recommendation change if Mushroom Truffle Melt demand increased?" Iteration and role assignment strengthen collaboration and final artifact quality.
Career Relevance

In real consulting or operations work, managers expect you to verify AI-generated analysis against source data, identify unrealistic assumptions before presenting recommendations, and iterate on reasoning rather than accepting first-draft output. Building verification prompts, error-recognition habits, and collaborative AI workflows into your process makes you a more reliable analyst and decision partner.
{{ modalTitle }}
Generated rubric
Customer Segmentation · Undergraduate
MicroSim
Evidence useWeight 30% · Level 4 / 5

Can the learner compare the three customer segments using purchase frequency, acquisition cost, and brand-fit evidence before recommending a target?

Decision qualityWeight 25% · Level 3 / 5

Does the recommended target segment match the stated business goal and constraints?

Reasoning clarityWeight 20% · Level 4 / 5

Is the recommendation explained in a clear, logical sequence a manager could follow?

Live defenseWeight 25% · Level 3 / 5

Can the learner justify assumptions and respond when the manager challenges the recommendation?

Every criterion is generated from this simulation — instructors can review, edit weights, or override before publishing.
SAMPLE ONLY
Segmentation Recommendation
2026
Target Segment Memo

I would target the premium skincare segment because they buy most often, and I would build the launch plan around their repeat-purchase behaviour. They are clearly the best customers overall, so the brand should deprioritise the value buyers. The strongest version would tie this to acquisition cost and retention before recommending a target.

Alex Rivera
Alex Rivera
Today at 2:30 PM

Good — you cited purchase frequency. Strengthen it by tying to acquisition cost over a 12-month horizon.

Alex Rivera
Alex Rivera
Today at 2:31 PM

"Best overall" is unsupported — compare against the value segment's retention and margin before concluding this.

Comments attach to the exact words the learner wrote — ideal for detailed written work.
Inbox
Marketing ManagerMar 14
Feedback: Segmentation recommendation
Hi Jordan — a few specific notes against what you wrote…

Feedback: Your segmentation recommendation

Quoted feedback
Alex Rivera
Alex Rivera <a.rivera@brightwell.co>
To: me
Mar 14, 4:20 PM

Hi Jordan,

Good work on this — a few specific notes tied to exactly what you wrote:

"I would target the premium skincare segment because they buy most often"

Right instinct, and good use of purchase frequency. Tie it to acquisition cost so the case holds up over a full year, not just at first purchase.

"They are clearly the best customers overall"

This is stated, not shown. Compare against the value segment's retention and margin before calling premium the "best" — otherwise it reads as an assumption.

Revise those two points and we'll set up the defense.

Best,
Alex Rivera
Marketing Manager · Brightwell

Quotes the learner's own words, then responds to each — best when feedback needs to point at specific claims.
Inbox
Alex RiveraMar 14
Your segmentation recommendation
Hi Jordan, strong first pass on this one. You picked a…

Your segmentation recommendation

Summative feedback
Alex Rivera
Alex Rivera <a.rivera@brightwell.co>
To: me
Mar 14, 4:20 PM

Hi Jordan,

Strong first pass on this one. You picked a defensible segment and your reasoning was easy to follow — the instinct to lead with purchase frequency is exactly how we'd approach it here.

To get to the level we'd expect before a client conversation, back your comparative claims with acquisition-cost and retention data, and name the trade-off you're accepting by deprioritizing the value segment. That turns a good recommendation into one you can defend under questioning.

Make those revisions and we'll set up the defense. Nice work overall.

Best,
Alex Rivera
Marketing Manager · Brightwell

One concise, workplace-style email after the task — best for a clear overall verdict without line-by-line markup.
Grading Assistant
AFTER THE SIMULATION
Customer Segmentation · MicroSim
CAREER-READINESS
78%
Grading Rationale Feedback Score Breakdown
Overall Feedback

Your recommendation correctly identifies the premium skincare segment as the highest-frequency buyers and uses the provided purchase data to justify the target. You ranked the three segments by purchase frequency and proposed concentrating spend on premium buyers while deprioritizing value buyers. However, the analysis leans on a single metric and asserts that premium are "the best customers overall" without comparing acquisition cost, retention, or margin across the segments.

Suggestions to Improve

  • Before committing to premium, verify the claim against acquisition cost and 12-month retention. For example, compare CAC and repeat-purchase rate for premium vs value buyers before declaring a winner — premium can buy more often yet cost more to keep.

  • Name the trade-off you are accepting. Deprioritizing value buyers may forgo a larger, lower-cost growth pool. State explicitly what volume or margin you would give up, e.g. "choosing premium trades reach for loyalty."

  • Pressure-test "best overall." Ask what evidence would change your recommendation, and state the conditions under which the value or mid-tier segment becomes the better target. This turns an assertion into a defensible decision.

Career Relevance

In real marketing or strategy roles, managers expect a target recommendation backed by multiple metrics — frequency, acquisition cost, retention, and margin — not a single signal. Articulating the trade-offs and the evidence behind your choice is what lets a manager act on your recommendation and defend it to finance. Building multi-factor analysis and explicit trade-off reasoning into your process makes you a more reliable analyst and decision partner.

Generated for every learner the moment the simulation ends — a class of 2 or 2,000+ each gets an in-depth, personalized career-readiness report citing specifics from their own work. The in-flow feedback above lives inside the simulation; this report is the wrap-up after it.
The report

The final report shows more than a score

Instead of only "this learner scored 78%," Cartedo shows performance across domain knowledge, workplace skills, AI collaboration, revision, and defense.

"Applied the framework accurately in writing, but struggled to defend assumptions when challenged."
"Improved after feedback, indicating coachability and stronger revision judgment."
Career-readiness reportA−
Domain knowledge90%
Workplace skills82%
Revision quality86%
Live defense74%
CoachableDefense needs practice

Want to see how Cartedo would grade your assignment?

Send us a topic, assignment, quiz, case, or syllabus. We'll create a free simulation and show you the rubric, feedback, and grading report.

98.6%
AI–professor grading alignment
Measured against instructor-reviewed grading in tested simulations.
Get started