On the Monday after a two-day consultative selling programme, a regional head listens to three recorded calls. All three reps open with a discount before the client has named a constraint. On Friday they had rehearsed two diagnostic questions instead. Nothing about the training was fake. By the first call review, it was gone.
Here is the claim worth arguing about: sales training decays faster than leadership training, not because salespeople are less motivated or the workshop weaker, but because sales behaviour is observed and corrected far more often. The same feedback loop that makes a sales team measurable erases a new skill before it has been practised enough to survive. That is not a reason to buy more workshops. It is a reason to move the reinforcement into the meeting that already does the correcting.
Why the call review pulls the behaviour back
Leadership behaviour is observed rarely. A manager runs a one-to-one badly in March and the evidence surfaces in a 360 or an appraisal months later, if at all. Feedback is slow, ambiguous, easy to argue with. A new leadership behaviour fades slowly and quietly: never actively punished, simply never tested.
Sales behaviour is observed weekly. Calls are recorded, logged in the CRM and opened in a pipeline or call review. Quota pressure supplies a hard, immediate consequence. In Baldwin and Ford’s transfer model the work environment is made of situational cues and consequences, including punishment and positive or negative feedback from managers and peers when someone tries a newly learned skill. A sales team meets more of those consequences in a week than a leadership cohort meets in a quarter.
Now put a new skill into that loop. The rep tries two diagnostic questions. The call runs longer, the client sounds unimpressed, and on the recording the attempt sounds hesitant because it is new. At the Monday review the manager asks why the deal has not moved and whether the rep is overcomplicating it. The old pitch closes faster. One awkward attempt teaches the rep that the new skill costs time and buys nothing, and within a week it has been actively corrected out.
That is the mechanism: high observation frequency plus a hard consequence produces fast active extinction; low frequency produces slow passive non-adoption. Both look like “the training did not stick” on a satisfaction form, but they need different designs. The general version, and the follow-up structure that works for slower cycles, is set out in why behavioural training fades.
What the forgetting curve actually says
Vendor decks open with a number: 70 per cent of training is forgotten within 24 hours, or 90 per cent within a week. Do not build a programme on it. It has no source.
Ebbinghaus published the original curve in 1885. He learned lists of nonsense syllables to two perfect recitations, then timed how long it took to relearn them after a delay. His measure was savings, the reduction in relearning effort in seconds: 33.7 per cent at one day, 25.4 per cent at six days and 21.1 per cent at 31 days. He was one participant, using meaningless syllables, and he measured effort, not whether anyone could perform a skill. He wrote that given the “special, individual, and uncertain character” of the results, no one would claim a law from them. A 2015 replication with four subjects found a similar six-day figure of about 23 per cent. None of this is a half-life for a skill. What survives is the shape: the steepest loss is early, then the curve flattens. That is why a single follow-up a month later misses the window, and why the folklore 90 per cent is not in his data at any interval.
The design answer is spacing. Cepeda and colleagues reviewed 839 assessments from 317 experiments in 184 articles and found that distributed practice beats massed practice, and that the gap between sessions that maximises retention grows as the retention interval grows. Revisit soon, then stretch the gaps. A workshop is a massed session. A weekly call review is a ready made set of spaced revisits, if you attach the new behaviour to it. Basing reinforcement on a number a finance review will recognise is the same discipline as the smallest honest version of four levels of evaluation.
Build reinforcement around the call review
Do not add a new meeting. Most reinforcement plans fail because they invent a calendar L&D has to chase. Sales already has a weekly ritual where calls are opened and managers give feedback; that is the loop extinguishing the behaviour, so put the new skill inside it. Write a one-page card, not a facilitator guide, with four fields: one behaviour, the cue in the call, a three-point anchor, and the owner. If it does not fit on one page it will not be used.
The method in the order the work happens
-
Pick one observable behaviour. Can a manager tick it on a 30-second clip without inferring intent? “Handles objections well” fails; “Asks two questions before stating a price” passes. This is the same test as a competency framework anchor: an attitude cannot be rated, and the case for a short set is the one we made in a competency framework nobody uses. One to three behaviours is the ceiling. Twelve will not happen.
-
Find the existing review. Name the meeting and the day. If the team has no weekly call or pipeline review, stop: there is no ritual to attach to, and the finding is that reinforcement has nowhere to live. That belongs in the training solutions conversation before a date is booked, not after.
-
Write the three-point anchor. 0 is absent, 1 is attempted but not to standard, 2 is done to standard. Add one sample clip per point so two managers score the same call the same way. Without the samples the score drifts and the card becomes an opinion.
-
Set the cadence with expanding intervals. Weeks 1 and 2: score one call per rep in every review. Weeks 3 and 4: every second review. Weeks 5 to 8: one rep’s call per review on a rotation. After week 8: one sampled call a month. This follows the spacing finding, frequent early and stretched later. The rep’s own CRM note must name the behaviour in one line.
-
Give the manager the debrief script. “Did you use the new questions?” gets a yes; “How did it feel?” gets an opinion. The question that works points at the tape: “Play the 30 seconds where the client named a constraint. What did you ask next?” Then one instruction, not a lecture. This is coaching done by the person who owns the number, so it sits with the manager rather than an external trainer.
-
Read the score before the number. In the review, the behaviour score is read before the deal value or the close. If the only number given airtime is revenue, the card is theatre within a fortnight. The consequence has to point at the new behaviour at least once a week for eight weeks.
Put the cadence into the programme design before the session, with a name against each checkpoint. If it is not on a calendar before the workshop, it will not exist after it.
A worked example
Illustrative example. A business-to-business sales team of 18 reps ran a two-day consultative selling programme. The one behaviour chosen was “two diagnostic questions before any pricing discussion”. Before the programme, 2 of 18 sampled first calls contained it, and 61 per cent of first proposals opened with a discount.
Nothing about the content changed. Four things were added: a one-page card, a three-point anchor with sample clips, a weekly call review score read before the pipeline number, and the expanding cadence above. The regional head, not a trainer, ran the review. At week 8, 14 of 18 sampled calls contained the two questions, and discount-first first proposals fell from 61 per cent to 24 per cent. The card took about ten minutes per rep per week of management time. These figures are illustrative, not a client result or a benchmark. They show the shape of the test: one behaviour, one existing meeting, one score, eight weeks. Run it on one team before you roll it out.
The honest limit
This method needs three things and fails honestly without them.
It needs recorded or reviewable calls. With no call recording and no weekly review, there is nothing to attach reinforcement to; creating that observation is a management change, slower and more political than buying a workshop. Say so in the proposal.
It needs a manager who will reward the new behaviour over the short-term number. If quota pressure punishes a slower discovery call, reps will game the card, and when the regional head is replaced mid-programme the cadence usually stops.
It needs one to three behaviours, not twelve. And the comparison has a boundary: leadership behaviour is observed too rarely for a weekly card, so this cadence does not transfer to a leadership programme. That is the point, not a footnote. Writing one observable behaviour and one existing number is the same discipline as converting a course request into a training needs gap statement. The feedback loops are not the same, and neither are the remedies.
What to do on Monday
Open the agenda for next week’s call review. It exists in almost every sales team, usually on a Monday. Add one line before the pipeline numbers: score one behaviour on one call per rep, using a three-point anchor, and read the score aloud. Pick that behaviour from the last sales programme you ran. If you cannot name one observable action, the programme had no target and the review cannot reinforce it.
If there is no weekly call review on the calendar, that is the finding: the reinforcement has nowhere to live, and the honest recommendation is to create the review before you book another workshop. Twenty minutes of a meeting that already happens will do more than a second day of training.
Send the manager the debrief question, not the slide deck. “Play the 30 seconds where the client named a constraint. What did you ask next?” If that gets asked every week for eight weeks, the skill has a chance. If it is asked once, the workshop was an event.
Sources
- https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm (Ebbinghaus 1885/1913, Chapter VII: savings table)
- https://verbamor.com/blog/ebbinghaus-90-percent (the 70 and 90 per cent folklore has no primary source; Murre and Dros 2015 replication)
- https://www.yorku.ca/ncepeda/publications/CPVWR2006.html (Cepeda et al. 2006: 839 assessments, 317 experiments, 184 articles)
- https://jasonhuangatwork.com/papers/Blume%20Ford%20Baldwin%20Huang%202010%20JOM%20-%20Transfer%20Meta.pdf (Blume et al. 2010: 89 studies, cues and consequences, decay over time)
- https://www.structural-learning.com/post/ebbinghaus-forgetting-curve (savings versus recall, single participant)