Improving E-Learning: My Reflections on Judging the iSpring CourseCreation Competition.

I was one of the judges for the iSpring course creation competition, alongside Melissa Millaway, Heidi Kirby, and Luke Hobson, and at the Grand Finale we shared some of what we had seen across the entries. Viewing other people’s work sharpens your eye for the things you know strong training should contain and shows you different ways of getting there. On the webinar we spent most of our time on the good work, of which there was plenty, so I want to use this space for the other half of the conversation, the patterns I think could improve across the board.

None of this is specific to the competition; these are the same gaps I see in almost every organisation I work with. The competition was about e-learning and most of what follows is framed that way, but the principles apply to any training experience you might create.

Narrative

There was narrative and storytelling in the entries, which is encouraging, so the question is one of quality rather than presence. A narrative needs an internal logic, a clear purpose, and as close an alignment to the domain you are training in as you can manage; once you’ve introduced it, you then have to use it. The pattern I saw repeatedly was a story that carried someone right up to the key learning point and was then dropped so the course could content dump the information on them, before the narrative reappeared later as though nothing had happened. Weaving the narrative through every learning moment is really important; if you introduce a model, present it inside the narrative and contextualise it to the situation the story has set up, instead of stepping outside the story to deliver it and stepping back in afterwards.

The clearest examples of this done well sit in books like The One Minute Manager, Who Moved My Cheese?, and Our Iceberg Is Melting, where the narrative has been carefully built to carry both the engagement and the content, even though it is often abstracted from any realistic domain; the story and the teaching are the same thing, and you are never asked to leave one to receive the other. Dropping in and out of a narrative robs it of its contextualising benefit, and there is a cognitive cost too, because separating the story from the point it is meant to frame is a form of temporal split attention, where attention and working memory are spent reconnecting things that should have arrived together (Chandler and Sweller, 1992).

Questions

There was formative assessment throughout the entries I judged, and most of it tested basic recall; the stronger work went past recall into more contextualised questions, though few entries reached full scenarios. Not everything has to be a scenario, but in a workplace context the value of a question comes from how closely it sits to the domain, because a recall question tells you whether someone can hold a fact in mind, whereas a domain-specific, highly contextualised question starts to probe whether they are equipped to apply it. That is no guarantee of transfer, and we still owe people all the usual work to make sure what they meet in training shows up as performance back in the role; contextualised, application-focused formative assessment is one of the steps that makes an effective training experience possible.

AI Images

AI showed up everywhere, and by far the most common use was image generation. I’ve written before about where each of us draws our own line on AI-generated images, and I won’t reopen that here; the consistency of the images is a separate, practical question, and it was the most common weakness I saw. When you generate images for a course, the style, the sizing, the way human characters are drawn, the way animals and objects appear, and the visual language of any diagrams all need to hold together across the whole experience. When every image looks like it came from a different place, as many did, you pull people out of the material and set them to work figuring out what is going on with the pictures, which is attention and effort spent on something other than the content (Chandler and Sweller, 1992).

In commercial content this matters more, because inconsistency reads as low budget, and that perception feeds straight into how usable and trustworthy the course feels; people tend to judge something that looks more considered as easier and better to use, even when the underlying quality is identical, which is the aesthetic usability effect at work (Kurosu and Kashimura, 1995). My view is that a lot of AI imagery currently harms the content it sits in by dragging down that apparent quality, so if you are going to use it, the consistency of the set matters at least as much as any single image.

Accessibility

Accessibility has improved a lot since previous years, and there is still a fair amount of ground to cover, much of it in the basics. Alternative text is the clearest example; the habit we need is not only to add alt text to images but to recognise that images do different jobs and need different treatment, so a decorative image takes an empty alt attribute so screen readers skip it, an informative image needs concise text that conveys what it carries, and a complex image such as a detailed infographic needs an extended description that does the same work the visual does for a sighted reader. That last case matters most in the context of AI, because if you are generating large infographics and dropping them into a course, you are taking on a substantial accessibility obligation; my first suggestion is don’t, and if you must, follow the guidance for complex images carefully.

Colour contrast has improved as well, though it still trips a lot of people up; the rules vary according to the size of the text and the purpose it serves. For time-based media, closed captions and a transcript are an essential in 2026, and the stronger position is to offer a complete alternative to the time-based media so that nobody has to watch or listen if that doesn’t work for them; people should also have control over any audio or video, with the ability to pause, move through it, and rewatch a section.

These expectations come from the Web Content Accessibility Guidelines, currently at version 2.2, and meeting them is part of the professional baseline. The guidelines are also changing; the next version, now being developed as the W3C Accessibility Guidelines (WCAG 3.0), is still a working draft and some way from completion, so the specifics here will move over time even as the underlying obligation does not (W3C, 2025a; 2025b).

Note: When it comes to accessibility, we really need to treat the Web Content Accessibility Guidelines as our baseline, the foundation from which we continue to improve. Thus, needing to get these basics right, we’re past the point now as an industry where we can say we’re working on it and maintain credibility. When we’re talking about things as simple as including alt text captions and getting colour contrast right, yes, we need to seek progress, not perfection, but we also need to hold ourselves to a standard that says inaccessible experiences simply aren’t good enough.

Titles

Melissa raised something on the webinar that deserves more attention than it usually gets, which is how we title and introduce a piece of e-learning. The title and landing page are the headline; a great opportunity to win someone’s attention and persuade them that this training is worth their time. We know people are busy, and we know that e-learning carries a poor reputation in most organisations, so we’re arguing from behind, and the title is where we make the case. A strong title is specific and concrete, it speaks to the value the person will get rather than the topic in the abstract, and it sets up an intent to engage before the first screen; a clear what’s-in-it-for-you statement that describes the situation the training helps with, in the words of the people who‘ll take it, does more to pull someone in than a description of the content ever will.

None of these are difficult to fix, and that’s the encouraging part of judging work like this; the gaps are consistent, which means the improvements are too, and a course that weaves its narrative properly, asks applied questions, holds its images together, meets its accessibility obligations, and earns attention with its title is doing many of the things that help people develop into competent performers.

To be clear, there are many other things we need to improve. I’ve got a whole separate article coming on the state of learning objectives, which were a consistent area requiring improvement in my judging. These five, to my mind, are the low-hanging fruit that would level up many of the entries I judged and a good deal of the e-learning out there in the workplace.

References

Chandler, P. and Sweller, J. (1992) ‘The split-attention effect as a factor in the design of instruction’, British Journal of Educational Psychology, 62(2), pp. 233-246.

Kurosu, M. and Kashimura, K. (1995) ‘Apparent usability vs. inherent usability: experimental analysis on the determinants of the apparent usability’, CHI ‘95 Conference Companion on Human Factors in Computing Systems, pp. 292-293.

W3C (2025a) WCAG 2 Overview. Web Accessibility Initiative.

W3C (2025b) W3C Accessibility Guidelines (WCAG) 3.0 [Working Draft].