Most content libraries only grow. A speaker adds another recorded session. A company uploads another training module. A course platform adds another unit. The assumption behind all of it is simple: more material means more value, and more value means more use.
The data on digital libraries tells a different story. As libraries grow, the share of the material anyone actually opens tends to shrink. That gap between what is available and what gets used shows up across a wide range of research: consumer psychology, learning science, and interface design all touch some part of it. This piece walks through what several of those research areas actually say, and where the popular version of the story oversimplifies the real findings.
Access Alone Doesn’t Predict Use
A useful place to start is a large, unusually direct dataset: massive open online courses. Anyone who enrolls in a MOOC already has full access to the material. There is no paywall, no login problem, no findability issue standing between the learner and the content once they sign up. A 2015 analysis of 221 MOOCs by researcher Katy Jordan, published in The International Review of Research in Open and Distributed Learning, found a median completion rate of 12.6 percent, ranging from 0.7 percent to just over half depending on the course. Full, unrestricted access to the material was, on its own, a weak predictor of whether anyone finished it.
The same pattern shows up outside education. A study of nearly 800,000 employees’ 401(k) retirement accounts, led by Sheena Iyengar with Gur Huberman and Wei Jiang, found that plan enrollment fell as the number of fund options available to choose from rose. Employees had access to every fund on the list. Enrollment still fell as the list got longer.
What Happens As a Library Keeps Growing
The 401(k) finding points at something specific: it is not just that people fail to use large libraries. Growth in the library itself appears to suppress participation, even before anyone has looked at a single item on the list. Every additional fund option was associated with a small further drop in the odds that an employee would enroll at all.
That is a strange result if more content simply means more chances for something to be relevant to somebody. It makes more sense if the size of the menu itself is doing work on the decision, before the content of any individual item gets evaluated. That is the territory of choice overload research, and it is more contested than the popular version of the idea suggests.
Choice Overload, and Where the Research Gets Complicated
The canonical study here is Sheena Iyengar and Mark Lepper’s 2000 jam experiment. At a grocery store tasting table, a display of 24 jam flavors drew more browsers than a display of 6, but converted far fewer of them into buyers: about 3 percent purchased from the large display compared to roughly 30 percent from the small one. The study, along with related work on chocolates and optional class assignments, became one of the most cited findings in behavioral psychology.
What gets left out of most retellings is what happened next in the literature. A 2010 meta-analysis by Benjamin Scheibehenne, Rainer Greifeneder, and Peter Todd, covering 63 conditions across 50 experiments, found an average effect size close to zero once all the studies were combined, with wide variation between them. A follow-up review by Alexander Chernev, Ulf Böckenholt, and Joseph Goodman concluded the effect is real but conditional: it shows up most reliably when the options are hard to compare against each other, when the chooser is uncertain about their own preferences, or when getting the decision right matters a great deal. Choice overload is a documented pattern rather than a fixed law. It depends heavily on what the options are and who is choosing.
Why People Can’t Find What’s Already There
A separate strand of research looks at what happens after someone decides to look for something specific, rather than browse an open menu. The Nielsen Norman Group, a longtime usability research firm, draws a distinction between findability (locating something a person already expects to exist) and discoverability (encountering something they did not know was there). Their research on both traces low performance on either one back to two causes: a site’s information architecture, or its navigation design. Rarely is the problem that the content itself doesn’t exist.
A related finding complicates the common assumption that a search box solves this on its own. Nielsen Norman’s work on search behavior has repeatedly found that most people are poor at composing search queries that match how a site’s content is actually organized or labeled. A library that depends entirely on search, without categories or navigation to support it, puts the burden of retrieval on exactly the skill most users don’t reliably have.
Long Recordings Versus Modular Video
Format matters too, and this is where multimedia learning research has produced some of the most consistent results in the field. Richard Mayer’s segmenting principle, developed across dozens of controlled studies, predicts that people learn more deeply from a video or animation broken into short, learner-paced segments than from the same material presented as one continuous piece. In the original test comparing a segmented lightning-formation animation to a continuous version, learners in the segmented condition performed substantially better on transfer tests, and later reviews summarizing this line of research put the median effect size near the high end of what’s typically seen in education studies. The effect is strongest when the material is complex, fast-paced, and the viewer is relatively new to the subject: exactly the conditions under which a single unbroken hour-long recording is hardest to sit through.
The Cost of Cognitive Load
Underneath both the findability problem and the segmenting effect sits a more basic constraint on working memory. John Sweller’s 1988 paper introducing cognitive load theory argued that human short-term memory is severely limited, and that tasks requiring people to hold too many items in mind at once degrade learning, regardless of how good the material itself is. Later refinements split this load into three types: the difficulty inherent in the material itself, the difficulty added by how the material is presented, and the effort spent actually building understanding from it. Only the first and third are unavoidable. The second, load added by confusing menus, unclear categories, or a long recording with no way to jump to the relevant part, is the kind a library’s design either adds or removes.
Structure Changes What Gets Used
This is where the findability research and the cognitive load research point at the same conclusion from two different directions: the same body of content, reorganized, gets used differently. Nielsen Norman’s guidance on diagnosing low findability treats it explicitly as a design and architecture question, testable and fixable independent of the content itself. Nothing about the underlying material needs to change for usage to change. What changes is whether someone can locate the specific piece they need without holding the entire library in mind to do it.
Why People Want the Right Piece, Not More Pieces
Information science has studied how people actually search for things since well before the web existed, and the findings cut against the idea that people want to browse a comprehensive collection. Marcia Bates’ 1989 berrypicking model described real-world information seeking as piecemeal and evolving: people gather a bit here, a bit there, refining what they’re looking for as they go, rather than issuing one query and retrieving one complete, final answer. A separate framework used widely in workplace learning, Bob Mosher and Conrad Gottfredson’s five moments of need, describes learning demand as arriving in distinct situational bursts: when someone is doing something for the first time, applying something they already know, or troubleshooting when it breaks. In both pictures, the person isn’t looking for the whole library. They’re looking for the one piece that answers what they’re facing right now.
Return Has to Be Designed
None of the research above argues for adding more content, and none of it argues that good material will get used just because it exists. Each study, working in its own domain, points at some version of the same design problem: usage depends on how material is organized, broken into pieces, and made findable at the moment someone actually needs it, rather than on how much material sits behind the login.
That is the same conclusion LeaderPass’s architecture was built around. Return has to be designed into an environment on purpose. It doesn’t happen because the content is good, and it doesn’t happen because access was granted.
Frequently Asked Questions
Why don’t people use content libraries even after they get access?
Research across several domains, from online course completion data to retirement plan enrollment, finds that access predicts use poorly on its own. What predicts use better is how the material is organized, how many options a person has to sort through before finding something relevant, and whether they can locate the specific piece they need without searching the whole library.
Does adding more content to a library increase or decrease how much people use it?
The available evidence points toward a decrease, at least past a certain point. A study of nearly 800,000 retirement accounts found that plan enrollment fell as the number of fund options rose, even though every employee had equal access to every option. Growth in a menu of choices can suppress engagement with the menu itself, separate from whether any individual item on it is worthwhile.
What is choice overload, and does it really happen?
Choice overload is the finding that, under some conditions, offering more options reduces the odds someone chooses at all, or lowers satisfaction with the choice they make. The original jam-tasting study that popularized the idea is real, but a large meta-analysis later found the average effect across many replications was close to zero. Later research concluded the effect is genuine but conditional: strongest when options are hard to compare, the chooser is uncertain of their preferences, or the decision feels high-stakes.
Why do people abandon digital learning programs after signing up?
Research on massive open online courses offers one of the clearest pictures of this. A widely cited analysis of 221 MOOCs found a median completion rate of only 12.6 percent, despite every enrolled learner having full access to the material from day one. The drop-off happens for reasons that have little to do with access and much more to do with format, pacing, and whether the content is structured around how people actually consume it over time.
Does modular, segmented video work better than one long recording?
Multimedia learning research says yes, under most conditions. Richard Mayer’s segmenting principle, tested across many controlled studies, consistently finds that people learn more from video broken into short, self-paced segments than from the same material presented as one continuous recording. The effect is strongest when the subject matter is complex and the viewer is relatively new to it, which describes most recorded talks and training sessions.