Skip to content
Wranx
Certainty-Based Marking

Find the people who are confidently wrong.

A score out of ten tells you who got it right. It cannot tell you who knew it from who guessed, and it cannot tell you who is about to act on something untrue. Certainty-based marking asks learners how sure they are as they answer, which turns two outcomes into four — and puts a name against the one that costs you money.

Why a pass mark is not enough

Two people score 8 out of 10. One of them is a liability.

They look identical on your dashboard. One knew eight answers. The other knew five, guessed three and got lucky — and is certain about two of the ones they got wrong.

A guess scores the same as knowledge

On four options, someone who knows nothing scores 25% by chance. A pass mark counts that as a quarter of the syllabus learned.

The gap it hides is the dangerous one

Not knowing something makes people ask. Being sure of something untrue makes them act. Only one of those reaches a customer, a patient or a site.

And it says nothing about what to do

A score of 60% tells you to run the training again. It does not tell you who needs what, so everybody gets the same remedial module and most of them did not need it.

How it works

One extra question, asked every single time

Alongside the answer, the learner sets how sure they are. The stake rises with the claim: certainty is worth the most when it is right and costs the most when it is not.

Admitting a guess is free

Answer at the lowest confidence and a wrong answer costs nothing at all. Honesty is never punished, because a system that punishes it stops being told the truth.

Claiming certainty is not

The asymmetry is the whole mechanism. If certainty and a guess scored the same when wrong, everybody would claim certainty and the extra signal would be worth nothing.

It changes how people answer

Deciding how sure you are means examining your own reasoning before you commit. That reflection is itself retrieval practice — the marking scheme teaches while it measures.

Certainty-based marking

Drag to set how sure you are, then look at what each answer is worth.

GuessingFairly sureCertain
If right
If wrong

Drag the slider. The two figures are the real ones the platform awards.

Four states, not two

Right and wrong is one bit of information. This is two.

Every answer lands in one of four places, and each one calls for something different from you. A pass mark collapses all four into a percentage and loses the instruction.

Correct, and sure

They know it

Knowledge they will act on without hesitating. This is the only state that belongs in a readiness figure, and the only one worth spacing further apart.

Correct, not sure

They think they know it

Right today, shaky under pressure. They will hedge in front of a customer or go and check. Fragile knowledge, and the cheapest of the four to fix.

Wrong, not sure

They know they do not know

A gap the learner can see. These people ask, escalate and look things up, so the risk is low — this is a training need, not an incident waiting to happen.

Wrong, and sure

They are confidently wrong

The one that costs you. No question gets asked, because as far as they are concerned there is nothing to ask about. Surfaced by name, on the day it shows up.

It is worth being blunt about which of these a conventional quiz can see. It sees the first two as a tick and the last two as a cross, and reports one number that mixes all four together. The distinction that matters most to you is the one it throws away first.

Where it comes from

Not our idea, and that is the point

Certainty-based marking was developed by Tony Gardner-Medwin, Professor of Physiology at University College London, and has been used in medical education since the 1990s — a field where confidently held misinformation has consequences nobody has to explain. It is one of the better-evidenced ideas in assessment: it improves the statistical reliability of the result, and it makes learners think harder about what they actually know. We did not invent it. We built a platform around it and put it in front of people whose confident mistakes reach a shop floor rather than an exam board.

What you do with it

A list of names, not a bar chart

The value of the fourth state is that it is actionable the moment it appears. You are not told a topic is weak; you are told who is sure about what, and wrong.

Who is confidently wrong, on what

Down to the individual and the concept. This is the report managers actually use, because it names a person and a subject rather than a percentage and a department.

The correction goes out on its own

A confident error queues the concept back at close spacing without anybody building a remediation plan. Being wrong is what schedules the next attempt.

Confidence as a trend, not a snapshot

Track how much of the team is in each state over a quarter. Rising certainty on stable accuracy is what readiness actually looks like on a graph.

Evidence for the conversations that need it

For compliance, an audit or an incident review, a record that separates who knew from who guessed is worth considerably more than a completion certificate.

Assessments

Chart progress with pre- and post-assessments

Most platforms can tell you a course was finished. The useful question is what people knew before it, what they know now, and whether that holds.

A baseline taken before the programme starts is what turns a final score into evidence. Without it, 82% is just a number; with it, you can say the team moved from 51% to 82% on the concepts you care about.

This is the measurement behind the results our clients quote. Tata Steel could show 97% of learners improved because the before was recorded as carefully as the after.

Customisable to what you actually need to know

Assessment here is not a gate at the end of a course. It is the measurement that tells you whether any of it worked.

Build assessments from your own material, target them at specific concepts, roles or sites, and run them at whatever interval makes sense — before a launch, after a rollout, or six months later to see what survived.

Results land in the same dashboard as the daily drills, so an assessment is a snapshot of a continuous record rather than an isolated event.

The number your board will ask for

Knowledge gain, per concept, per team, with a baseline behind it. That is a far harder thing to argue with than a completion rate.

The score is on the answer, never on the person

A negative number on a wrong, confident answer is how the arithmetic keeps the signal honest — it is not a penalty applied to somebody’s record. Mastery in Wranx only ever rises through demonstrated understanding, and fades with time; a bad day cannot take it away. Nobody is ranked on confidence, no one is reported for a wrong answer, and being certain and mistaken queues a drill rather than a note to a manager. The point of finding the confidently wrong is to fix it before it matters, which only works if people answer honestly.

It does not work on its own

Certainty-based marking finds the gap. Spaced repetition is what closes it — a confident error comes back at close spacing until it is right for the right reason, then stretches out again. The two are one mechanism, which is why neither is sold separately.

See it against your own material

A demo covers what the four states look like for your content and your teams — including, if you want, the part of your syllabus your people are most confidently wrong about.