I've Been Wrong About Happy Sheets the Whole Time!

A paper crossed my reading list earlier this week that stopped me mid-scroll. A 2025 longitudinal study, conducted across 14 organisations and tracking over 2,800 participants, found that learner satisfaction ratings collected immediately after training are a statistically significant predictor of on-the-job performance improvement at the six-month mark. The correlation between post-training satisfaction scores and independently assessed performance outcomes held at r = 0.71 after controlling for prior knowledge, role complexity, and manager support, which is remarkably strong for workplace research of this kind.

If you’ve spent any time defending end-of-course surveys against the accusation that they measure nothing useful, this will feel like overdue vindication. The researchers argue that satisfaction functions as a proxy for intrinsic motivation; motivated employees invest more cognitive effort in transfer activities after the training event, and those who engage in more self-directed practice perform better when it counts. The causal pathway proposed in the paper runs from enjoyment through motivation to effort to measurable performance improvement, and the data, gathered across multiple sectors and training formats, appears to support it.

There are nuances, naturally. The effect was strongest for skills-based programmes and noticeably weaker for mandatory compliance content, which the researchers attribute to different psychological dynamics around choice and autonomy. But the core finding is clear, and it suggests that the profession’s long-standing instinct to measure how learners feel about an experience may be considerably more defensible than its critics have been willing to admit.

If that’s all sounding a little too convenient, you’re right to be suspicious.

Happy April Fools’ Day.

Everything you’ve read so far is invented. The study doesn’t exist. The sample size, the correlation coefficient, the causal chain, the carefully placed nuance about compliance content; all of it was fabricated. But if you found yourself nodding along, or feeling a small flush of professional vindication, don’t feel bad. This happens to all of us at some point.

This kind of fabrication works because it tells us what we already want to hear. Most L&D teams collect post-training satisfaction data. A study that validates an existing habit attracts far less scrutiny than one that challenges it, and that asymmetry in how carefully we read is where misinformation finds space to operate. Confirmation bias can sneak up on any of us, and is the reason the first three paragraphs of this article may have felt comfortable rather than alarming.

Three patterns are worth being aware of in the content we consume as practitioners.

The first is honest error: people misread studies, over-generalise from small samples, or cite secondary summaries without checking the original research, and none of that requires bad intent to cause damage to practice at scale.

The second is agenda-driven content: vendors, consultancies, and platform providers present selective data in service of a sale with regularity, and the line between marketing and evidence is often blurred with considerable skill.

The third is the limit of individual expertise: I work on topics with clients that I don’t write about publicly, because writing with authority requires a depth of engagement that goes beyond surface familiarity, and that distinction matters more than our industry tends to acknowledge. Expertise in one area doesn’t confer credibility in all of them.

The profession deserves better than comfortable conclusions dressed in the language of evidence; we all do. Reading critically, verifying sources, and being willing to interrogate findings we’d prefer to be true is the minimum standard we should hold ourselves and each other to, and it applies to everything you read in this industry, including whatever you read next.