Nobody Learns for Free: What Happens to a Course When the Price Drops to Zero?

Three price tiers, one institution, 15,048 learners

In our previous study we compared the same comprehensive course at two prices - full price in the thousands of shekels versus a symbolic 18 ILS (about $5) - and found a 17x viewing gap. But one question stayed open, maybe the most interesting one: what happens on the last step down, when the price drops to absolute zero?

To answer it, we analyzed the viewing data of 10,735 enrollees in the college's free intro course - a short, accessible digital course of 26 lessons that anyone can join for free - alongside the two purchase cohorts from the previous study. Together: 15,048 learners, three price tiers, one institution, and one learning platform measuring everyone the same way.

The result is a single sharp, unforgiving curve: on every metric we measured - opening a first lesson, median viewing, reaching half the course - actual learning rises steeply with price. Content differences between the courses cannot explain a curve this steep on their own; the people, and the size of the commitment they took on, can.

The price-commitment curve: every dollar buys persistence

The two core metrics tell the same story from two directions. The start metric - how many enrollees opened at least one lesson - climbs from 20.4% at free to 71.2% at 18 ILS to 94.8% at full price. Median viewing climbs from 0% at free to 1.4% at 18 ILS to 23.7% at full price.

Note that the curve is not linear: the biggest jump in the start metric happens between free and 18 ILS, a nearly symbolic payment, while the biggest jump in learning depth happens between 18 ILS and full price. In other words: paying anything makes people open the course; paying something substantial (and the deliberate process that comes with it) makes them learn it.

The shortest course with the lowest completion

You might expect a short course to be finished more often: the free intro course is only 26 lessons - more than five times shorter than the 147-lesson version and more than eight times shorter than the 224-lesson one. Finishing it is incomparably easier. And yet, only 2 out of 10,735 enrollees - 0.02% - completed it.

The stopping-point distribution, normalized to share of course length so courses of different lengths can be compared, shows how extreme the pattern is: 79.6% of free enrollees never started at all, and another 4.7% stopped within the first 5% of the course. Only 0.6% crossed the halfway mark - versus 21.2% of full-price buyers, whose course is more than eight times longer.

Even 18 ILS is not free: the power of a symbolic payment

The most practical finding of the study hides in the middle tier: a payment of 18 ILS, about the price of a sandwich, multiplied the odds that an enrollee would open the course at all by 3.5 (71.2% versus 20.4%). Something in the act of paying, however small, changes the relationship with the product: what cost money gets attention; what is free gets postponed to a "someday" that rarely comes.

But it is just as important to see what a symbolic payment did not do: it did not create deep learning. The median viewing of 18 ILS buyers stayed at 1.4% of the course, and none of them finished. The symbolic payment made people open the door - it did not make them walk in. Only the combination of a substantial price with a serious enrollment process did that, as the previous study showed.

The 'haven't gotten to it yet' myth, free edition

As in the previous study, we tested the obvious objection: maybe the enrollees just haven't gotten around to it? We split the free-course enrollees by enrollment month, from veterans who joined five months ago to the newest joiners. The median viewing is identical in every tenure group: zero. Even people who have had the course open to them for almost half a year never started.

This replicates exactly what we saw in the 18 ILS cohort, where the median stayed at 2 lessons even among buyers with nearly a year of access. Time does not solve the starting problem - it cements it: whoever does not open the course within the first few days almost never opens it at all.

What this means about free, about price, and about real learning

The conclusion is not "never learn for free" - quality free content is an excellent way to sample a field and check fit. The precise conclusion is: free access to content is not learning. Across 15,048 learners at three price tiers, the variable that best predicted whether a person would actually learn was not content quality, course length or available time - it was the size of the commitment they took on upfront.

These findings are consistent with the large-scale MOOC study published in Science, which found completion rates of about 3% in open online courses - but our data adds what edX cannot: one institution, one measurement system, and three price tiers side by side. For the practical takeaway, whether you are a learner choosing how to study or an educator deciding how to price: the price is part of the product. Commitment devices - real payment, fixed sessions, people expecting you - are not the enemy of access; they are what turns access into learning.

Methodology

Update history

• 11 Aug 2026: Precision pass after an external data audit: the full-price purchase window was corrected to late 2021 - 2023, and several phrasings were tightened. All statistics were independently re-derived from the raw exports and confirmed unchanged.
• 11 Aug 2026: English edition published alongside the Hebrew original. Free-course data extracted 11 Aug 2026; both paid cohorts were re-derived from the original export and verified against the previously published numbers (full match).

Frequently asked questions

Continue reading