The course the sponsor bought and the learners stopped attending

The diagnosis is an attribution error. 'Lazy and unmotivated' explains nothing and predicts nothing. The attendance drop is patterned, not random: it starts at week four and two learners gave a written reason. Adults attending an evening class after a working day have a high opportunity cost, and they are volunteering their own time even if the firm pays. A drop of five from fourteen is a course-fit signal.

There are also two different 'customers' whose needs may conflict: the sponsor (email and video calls), the learners (conversation, meaning spoken interaction without heavy grammar), and the teacher (cover the book). Nobody has checked a single one against actual workplace language.

Needs analysis inside the existing hours - triangulate:

1. Authentic texts, not opinions. Ask learners to bring or forward 5–8 real emails they have written and, with consent, record two short calls or take notes on one. Analyse them for recurring features: request forms, hedging, tense use, openings/closings, repair strategies. This produces the syllabus content.
2. Separate interviews with three learners and with the sponsor, by phone, ten minutes each. Do not put them in the same room; the sponsor's perception of need and the learners' felt need will differ, and both are data.
3. Self-assessment grid using can-do statements mapped to email and video-call scenarios - faster and more informative for this purpose than a placement test.
4. Observation of one real handover or team meeting if the firm permits.

Distinguish objective needs (what the job demands), subjective needs (what learners want), and lacks (what they can't yet do). Then feed the findings back to the group in a lesson. Being shown your own emails as the course material is itself a re-engagement device.

Reframe the apparent conflict. 'Conversation' in learner language usually means real-time spoken interaction and relevance, not the absence of writing. Video calls are conversation. So the two demands converge if the syllabus is organised around communicative events rather than language points.

Principle: one transaction, two modes. Take a recurring workplace event - a customer asking about a delayed consignment. The spoken task is the video call; the written artefact is the confirmation email that follows it. The same scenario generates both, so the sponsor sees written work and the learners see why they are writing it. Writing stops being homework and becomes part of the event.

A workable cycle: pre-task exposure to two authentic model texts (one call extract, one email) → noticing what the competent version does → short language focus drawn from those texts (request forms, softening, future certainty) → rehearsal of the call in pairs → production of the email → feedback on both, with the criteria shared in advance.

What to stop

• Stop 'covering units 1–10'. The coursebook is a resource, not the syllabus. The sponsor's original brief is the legitimate ground for departing from it, and the teacher should document the mapping so nobody thinks content was dropped.
• Stop skipping writing to make room for speaking. That produced the sponsor's complaint and taught the learners that writing is the expendable skill.
• Stop treating the grammar the learners object to as the organising principle. Grammar is retained but subordinated: it appears because a task needs it, which is exactly what learners mean when they say 'less grammar'.

Risk: if the learners' real complaint is that the class is too hard or too fast, a task-based redesign will not fix it. The interviews should test that.

Diagnose the spread first. 'A2 to low B1' is one CEFR band of difference, which is manageable within one class; the sharper issue is the three returners, whose problem is likely confidence and study skills, not level. Check with a short spoken diagnostic rather than assuming.

Differentiation by support, not by task. Keep one task per stage and vary the scaffolding:

• lower support: model text with sentence frames, closed role cards, a checklist of moves, planning time in pairs;
• higher support: open role cards, one constraint added (a dissatisfied customer, an interruption), no model text. This keeps the class together and avoids the lower group doing a different, easier lesson, which adults read as being written off.

Pairing. Pair strong with weaker for rehearsal, but rotate, and give the stronger learner a specific role (note-taker, challenger) so they are not simply acting as unpaid teacher. Avoid fixed 'helper' pairs; they create dependency and resentment.

The three returners. Homework will not happen, so build a 15-minute catch-up at the start of each session rather than assuming out-of-class work. Make the recycling visible: the same email frame reappears three weeks running with increasing demand. Reassure them explicitly about error correction - adults who have not studied for decades often read correction as judgement.

Assessment that satisfies the sponsor without distorting teaching. Use performance assessment: a recorded video call at week 5 and week 9, and a portfolio of three emails with a short self-commentary. Rate with an analytic rubric (communicative effectiveness, range, accuracy, fluency, appropriateness) mapped to CEFR descriptors, and share the rubric at the outset. A grammar test would be easier to administer and would tell the sponsor nothing about whether the emails improved - and would drive teaching straight back to the coursebook.