Assessment

How to assess speaking without recording 30 students one by one

The traditional method works and nobody can afford it. Two lessons of one-to-one assessment per class, per term, marked from memory at the end of a long afternoon.

8 min read
Short answer

Stop trying to assess all thirty in the room. Collect speaking evidence continuously from practice students do on their own, use a small fixed set of criteria so marks are comparable, and reserve your live one-to-one time for the students whose evidence is unclear or who need a formal grade.

What one-by-one actually costs

Five minutes per student is a modest speaking assessment. Thirty students is 150 minutes — three lessons, before you account for settling the rest of the class while you work through them.

3 lessonsTo assess one class one-to-one at five minutes each
×4 classesWhat that becomes across a typical timetable
Once a termHow often it therefore realistically happens

And the marks that come out are less reliable than the effort suggests. You assess student one at full attention and student twenty-eight at the end of the afternoon. You have no recording to go back to. If a colleague marked the same students, the two of you would disagree more than either would like.

Assess continuously instead of occasionally

A single sitting is a snapshot taken on the day a student happened to have a cold or a bad morning. Evidence gathered across a term is both fairer and less work, because none of it is a separate event you have to schedule.

The shift is to treat speaking practice as the assessment. If students are speaking regularly and each attempt comes back with something recorded against it, you already have more evidence by half-term than a one-to-one sitting would ever produce.

DoCollect short samples often, across the whole term.
AvoidOne long assessment event that stands for everything.
DoKeep the recording so you or a colleague can check the judgement.
AvoidMarking live from memory with no way to revisit.
DoUse your live time on the students whose evidence is ambiguous.
AvoidSpending equal one-to-one time on students you are already sure about.
Read-aloud result

Thevolcanoreleasedthoroughcloudsofash.

One word to practice — the rest read clearly.

Accuracy88
Fluency92
Completeness100
A comparable result on the same task, for every student — not just the ones you reached before the bell.

Decide what you are grading

Most inconsistency in speaking marks is not carelessness. It is two teachers, or the same teacher in two moods, weighting different things. Fix the criteria and most of the drift goes away.

CriterionThe question it answersCommon trap
Task achievementDid they do what was asked?Rewarding a fluent answer to a different question
RangeDid they reach beyond safe language?Marking the student who risked nothing above the one who tried and slipped
AccuracyWas the language controlled?Letting a strong accent read as an error
RecoveryWhat happened when they got stuck?Not assessing it at all — it is the most real-world skill of the four

What this looks like in Dily

Speaking practice produces a record without you setting up an assessment. Read-aloud work comes back with how accurately and smoothly it was said. Open conversations come back with the full transcript, how long the student spoke, and what the AI coach noted.

  • You can play back what a student actually said, so a judgement is checkable rather than remembered.
  • The same criteria are applied to every student in every class, which is the part that makes marks comparable across a department.
  • Students see encouraging feedback; you see the detail behind it.

None of this removes your judgement, and it should not. For anything that carries a formal grade, listen yourself — you now have the recording and a starting position to argue with, rather than thirty conversations and a blank mark sheet.

Get started

Bring Dily to your classrooms

Tell us about your school and we'll set up a demo tailored to how your teachers work. Pilot programs available.

We usually respond within 1–2 school days.