I cannot recall who told this story back in the mid-1980s at an NSPI (now ISPI) Conference.
It seems that they had created some new sales training reflecting a new sales process.
The target audience Reaction Data (Level 1) measured during and after the training was satisfactory.
The post-training Mastery Data (Level 2) measured immediately after the training was satisfactory.
The post-training Transfer Data (Level 3) measured out in the field for a month or so after the training was satisfactory.
The post-training Results Data (Level 4) seen after a couple of months were not satisfactory.
Not at all.
It seems that the new sales process took twice as long as the previous one.
Which meant that revenues were halved.
That Sales Process and the associated Training was stopped cold and quickly redone.
The next one – the Sales Process – was fully tested before being released to all of Sales.
That Sales Process was the basis for the Instructional Content – Job Aids and Training – and because both weren’t fully tested – with what I like to frame as a FULL DESTRUCTIVE TEST – before making the content generally available – for access or deployment – it was a mini-disaster – for Sales and for the Training functions.
I cover “FULL DESTRUCTIVE” Pilot Testing in this book from 2020.
See my Amazon Authors Page for all of my books: https://www.amazon.com/-/e/B08JQC4C4V