How to Run a Training Session That Sticks

A training session sticks when it is built backwards from one behaviour people must perform differently next week, when they retrieve the material from memory several times during the session itself, and when two short follow-ups arrive in the fortnight afterwards. Everything else - the deck, the handouts, the trainer's expertise - is secondary to those three things.

The default fails for a structural reason rather than a personal one. A trainer who knows a subject well wants to cover it, so the session becomes a tour of the material at the pace of someone who already understands it. People nod, the feedback form comes back positive, and a fortnight later almost none of it is in use. Enjoyment on the day and retention a month later are close to unrelated.

This guide covers how to pick the one behaviour worth building the session around, how to shape a ninety-minute block, how to run retrieval checks that teach rather than test, how much content to cut and why, what real practice looks like, how to handle remote and hybrid rooms, and what to send afterwards so the session survives contact with everyone's actual job.

Why training fades within a fortnight

Memory for material you have only heard once decays steeply and quickly, and it decays fastest in the first few days. That is not a fact about your audience's commitment; it is how memory behaves without retrieval. A session that consists of clear explanation followed by nodding is the exact condition under which forgetting is fastest, because nothing in the room required anyone to pull the material back out of their own head.

The second mechanism is fluency. Watching a confident expert work through a problem feels like learning, and it produces a strong sense of understanding. That sense is a poor guide to whether you could do it yourself on Tuesday with a customer on the line. People leave rating the session highly and genuinely believing they have it, which is why happy-sheet scores are a bad instrument for anything except the catering.

The third is that nothing changes in the environment the learner returns to. If the new process takes longer than the old one for the first fortnight, and nobody's manager knows the training happened, the old habit wins by default. Design work can fix the first two problems; the third has to be handled by whoever owns the workflow, and it is worth saying so out loud before the session is booked.

Design backwards from the behaviour, not the topic

Start by finishing this sentence: after this session, people will be able to do X, in situation Y, without help. 'Understand the new pricing model' fails - understanding is not observable and nobody can tell whether it happened. 'Quote a two-year renewal with the new discount bands, unaided, in under five minutes' passes, and it immediately tells you what the session has to contain.

Then write the failure you are trying to prevent. Someone quotes the old bands out of habit; someone escalates a case that they could have handled; someone skips the safety check because the form is confusing. Naming the failure keeps the session anchored to reality and gives you a supply of practice scenarios that are worth more than any invented case study.

Now be brutal about scope. One behaviour per ninety-minute session, two at the absolute most. Trainers resist this because the sponsor asked for a list of eight topics, and eight topics genuinely do need covering - but they need covering across four short sessions, not compressed into one long one. If you cannot renegotiate the scope, at least sequence it so the first behaviour is fully practised before the second is introduced.

Write the assessment before you write the content. If you cannot describe the exercise that would prove someone can do the thing, you do not yet know what you are teaching. That exercise becomes the spine of the session, and everything you build is either preparation for it or practice at it.

The shape of a ninety-minute training session

Notice the ratio. Roughly twenty-five minutes of the ninety is the trainer talking, and it arrives in two chunks rather than one block. The rest is retrieval, practice and commitment. If your current session inverts that ratio, cutting content is not a compromise you make for time - it is the change that makes the session work.

The opening questions are the piece people skip because it feels backwards to test before teaching. Do it anyway. An unsuccessful attempt at an answer primes the explanation that follows, and it takes ninety seconds. It also gives you a genuine read on what this particular group already knows, which is usually not what the briefing said.

Retrieval beats re-explaining

Pulling information out of memory strengthens it far more than putting it in again. That is why a short quiz mid-session is worth more than a recap slide covering the same ground, and why asking 'what were the three conditions?' beats showing the three conditions a second time. The effort of the retrieval is doing the work, so the moment where people struggle slightly is the moment that counts.

Run the checks as low-stakes and visible. A scored quiz on the shared screen with everyone answering on their phones does three things at once: it forces every person to commit to an answer rather than nodding along, it shows the room where the group as a whole is confused, and it hands you a live list of what to reteach. Session Flo runs these from the same room code as the rest of the session, so the quiz is a tap rather than a new sign-in halfway through the afternoon.

Write the questions to teach rather than to catch people out. Plausible wrong options built from the actual misconceptions - the old discount band, the step people always skip - are worth far more than obviously silly distractors, because choosing one and being corrected is where the learning happens. Keep the pass mark out of it entirely; the number on the leaderboard is a by-product, not the point.

Reteach only what the results show. Three questions where half the room went wrong is a five-minute reteach; a question everyone got right needs no time at all, however carefully you prepared that slide.

Cut the content by a third

Working memory is small, and new material arriving faster than it can be processed does not queue politely - it displaces what came before. A session that covers twelve points at pace typically leaves people with two, chosen more or less at random. The same session covering six points with practice between them leaves five, and leaves them usable.

Make the cut mechanically rather than by judgement, because judgement always protects your favourite material. List everything you planned to cover. Mark each item as needed to perform the behaviour, useful background, or interesting. Delete every item in the third group and move the second group into a document you send afterwards. What remains is your session.

Reference material is not training material and it does not need to be delivered live. Nobody learns a policy table by watching someone read it. Put it in the follow-up pack, tell the room it exists and where it lives, and spend the recovered minutes on the two things a document cannot do: practice and correction.

The most common objection is that the sponsor is paying for coverage. Meet it directly with numbers: this session covers four things people will be able to do on Monday, and thirty-one things nobody would have remembered by Tuesday go into the reference pack. Framed that way, most sponsors take the trade.

What real practice looks like

The single biggest upgrade available to most training is swapping the invented case study for examples the participants bring. Ask for them as pre-work in one line: bring one real case from the last month where this came up. The material is instantly relevant, the ambiguity is real, and you are no longer defending a scenario that somebody in the room will correctly point out never happens.

Let people struggle a little. The instinct to rescue after four seconds of silence is strong and it removes exactly the effort that makes the practice worth doing. Count to ten, watch what people actually do wrong, and use those errors as your next teaching input.

Give feedback as one specific change rather than a verdict. 'Good effort' teaches nothing; 'you quoted before you confirmed the renewal date - do that check first next time' can be acted on immediately, and the next attempt proves whether it landed.

Spacing: the follow-ups that do the real work

Running it remotely or hybrid

The trap in hybrid training is that the in-room group gets the good version and the remote group gets a broadcast of it. Practice exercises are where this bites hardest, because pairing up is effortless in a room and requires deliberate setup online. Decide the pairings in advance, including cross-location pairs, so nobody is left as the odd one out at minute twenty.

Video calls also make free-flowing ideation harder than it is in person, so lean more heavily on structure when you are online: type first, share second, discuss third. That sequence costs nothing in a room and rescues a remote session.

Measure whether it stuck, not whether it was enjoyed

Satisfaction scores are worth collecting for one purpose only: spotting a session that was genuinely unpleasant, badly timed or technically broken. They are not evidence of learning, and treating them as such is how organisations end up with a well-liked programme that changes nothing.

What to send afterwards

A recap generated automatically from the session itself is worth more than one written from your own notes, because it contains what the room actually produced: the quiz results, the poll answers, the commitments people typed. It also takes minutes rather than an evening, which is the difference between it going out the next morning and not going out at all.

Keep the format identical every time you run the session. Regular attendees learn where to look, and consistency makes the follow-ups feel like part of the programme rather than a series of ad hoc requests from someone who wants their numbers to look good.