Winning at Udemy: What the most successful courses have in common

min read

Do you want to learn a new IT skill on Udemy but don’t know which course is the best value in terms of cost and rating?

Resident Data Evangelist Michelangiolo Mazzeschi and Content Marketer Andrés Slaughter have analyzed the prices, discounts, ratings, duration, and reviews of over 8,500 courses

Udemy is an online learning and teaching marketplace with over 183,000 courses and 40 million students.

The 21st century is the first time in human history where the accessibility to learn anything we want is at our fingertips. From online universities, marketplaces, communities, and more, the choice of subject is endless.

With choice comes responsibility, and determining what’s going to be the best use of your time becomes a challenge. Even more so when faced with learning a new IT skill on an online marketplace like Skillshare or Udemy.

With 183,000 courses, it can be overwhelming even just determining which subject to learn, let alone the video.

With Udemy only allowing up-to-a-point viewing unless you purchase the course, you’re taking a shot in the dark about what value you’ll get.

The only data available is the average review of the course along with every review written and the price and duration. Now picture looking at all that. And do it again, and again. That’s the manual process of finding the right course for you. Tedious and time-consuming.

We did all the hard work for you and compiled actionable insights and a breakdown of what is the best and worst courses on Udemy so you can dive right into learning. Valuable learning.

If you would like to see the Udemy Course Review Subject Clusters in the dashboard while reading along, you can view the Clustering Dashboard here.

TL;DR Actionable Insights

  • Highest rated course subjects were focused on JavaScript, IOS Development, and PHP.
  • Lowest rated course subjects covered Ember.js, Cassandra & XML.
  • SQL, Swift, and Magento offer the largest selection of courses on the site
  • Powershell, Gitlab, and Machine Learning offer the smallest selection of courses on the site.
  • The most subscribed AND reviewed course is Complete Python Bootcamp: Go from zero to hero in Python 3, with 685,057 subscribers.
  • The course with the most lectures at 779 is: Back to School Web Development and Programming Bundle.
  • The longest course is 356 hours (over two weeks): The Last Complete Coding Course You Will Ever Need!
  • The longest average course duration was IOS Development, Unity, and Web Development.
  • The shortest average course duration was AJAX, PHP, and WordPress
  • The most reviewed course subjects were Python, React, and Web Development.
  • The least reviewed course subjects were Bot Framework, Machine Learning, and Java.

The Best Data Science Courses On Udemy

  • The highest-rated course subject is Javascript; being the only subject with a rating over 4.5 (4.55/5), it has a near median duration (7.2 hours) of all the courses, being in the top 38%, while only being in the top 47% of average number of reviews.
  • There are 49 courses available that teach Javascript, which is in the bottom 37%
    • Significant trends in course title were the inclusion of es6 and es5 as well as ECMAScript.
  • The course subject with the highest average reviews is Python, ranked 1st (top 99%) with an average review count per course of 7,496; it has one of the longest average duration (12 hours, 14th, top 5%) of all the courses. It has a respectable average rating of 4.32/5.
  • There are 35 courses available that teach Javascript, which is in the bottom 30%.
    • A significant trend in the course title was the inclusion of Heroku.
  • The longest course duration subject is IOS Development, ranked 1st (top 99%) with an average review count per course of 1,838 (ranked 8th, top 2%); it has the second-best average rating (4.44/5 2nd, top 99%) of all the courses.
  • There are 55 courses available that teach Javascript, which is in the top 55%.
    • A major trend in the course title was the inclusion of IOS11 and IOS 12 and Xcode11.

What works

66% of the most subscribed and reviewed courses used keywords such as ‘Complete,’ ‘Masterclass,’ and ‘Bootcamp.’

The top 15 courses also are outliers in terms of course duration as they ranged from 9.5 – 79 hours, with the majority falling within 40-50 hours.

The majority of courses have an average duration between 10 and 4 hours, with the most populous course subjects having an average duration between 4-6 hours. Only one-course subject (Python 10.1 hours average duration) goes over 8 hours average duration.

In contrast, the highest-rated courses tend to follow the value of content in the course rather than duration as two of the top five highest-rated course subjects are over 10 hours (Unity 10.28 and IOS Development 32.49)

Furthermore, the course subjects with the highest average number of reviews all have an average duration of over 10 hours:

Hypothesis: The value of the course dictates whether or not the time investment will be reflected in the review. The determining factor in the conversion of the viewer to a subscriber on Udemy IT courses is the title which has been shown the best practice to include keywords relating to a total summation of learning so that the viewer feels the price justifies the means.

What doesn’t

Longer is NOT always better when it comes to making a Udemy course. Only 32% (8/25) of the longest courses (by duration) on the site had over 1,000 subscribers.

Value of diversity is critical, meaning adding content to a course matters. Only 20% (5/25) of the courses with the lowest number of lectures received 1,000.

The lowest 25 courses also are outliers in terms of course duration as they ranged from less than an hour to three-and-a-half.

23-course subjects have an average duration of fewer than 3.5 hours, with many of them having bad ratings:

  • Gitlab 3.55 (achieves optimal course duration time at 6 hours)
  • Odoo 3.54
  • Tensorflow 3.53
  • d3.js 3.46
  • Apache Spark 3.38
  • Bot Framework 3.25

The structure is just as important as what’s being conveyed through the title when it comes to the title. I would call those two the same. Many of the lowest-rated courses have titles such as these that follow no cohesive structure and lack intrigue or value proposition.

Furthermore, the course subjects with the highest average number of reviews all have an average duration of over 10 hours:

What to avoid

Simply put, Ember.js, Cassandra, and XML were the lowest-rated course subjects, with XML almost dipping into the 2-star rating (3.05).

Trends to avoid are themes previously discussed, such as lack of brevity and hook to the title:

Similarly, avoid course subjects that have a low number of reviews as they tend to follow the trend of low course duration, which has been shown to correlate with low value and effort from the instructor hence resulting in such low ratings:

Falling outside the range of 4-10 hours for a course unless the instructor has built a reputation will result in low effort when referring to the low course duration and lack of social awareness on the longer course duration. The courses with fewer lectures are also a course to avoid as it leans towards the hypothesis of low value through lack of effort.

Furthermore, the course subjects with the highest average number of reviews all have an average duration of over 10 hours:

Hypothesis: The bottom line is value plays a factor in determining whether or not a course will perform well over other variables such as title, thumbnail, price, etc. This is due to the fact that in some other cases of video content, the investment is time which isn’t perceived as more valuable than the combination of money + time + preparation. That results in a huge letdown if the value proposition of the course doesn’t meet the expectation.

Conclusion

The best courses are optimized for the modern age, meaning that time to those instructors is valuable hence why the courses range from 4-6.

As progression occurs, then the advancement of duration can be taken where the instructor releases a 20, 30, or 40+ hour video.

Why? Because they’ve proven to the viewer turned subscriber that they offer a fair value from the time + money + preparation it cost.

If you’re looking to start a course, lean towards subjects such as JavaScript, IOS Development. And PHP for best ratings and SQL, Swift, and Magento for the most diversity of courses within a subject.

And if you’re a data person, as I’m sure you are, if you’ve made it this far, then Python, React, and Web Development are perfect as they’re the most reviewed course subjects.

Dataset

Compiled from over 8,500 data points taken from the Udemy website. These were then exported into a CSV file and uploaded as a dataset in the Relevance AI platform, then vectorized & clustered.

The Udemy Course Reviews Dashboard was scraped from the Udemy website. To clean the dataset, I preprocessed the data, extracted insights using additional Machine Learning techniques (such as zero-shots, but sentiment analysis could also be used), and then selected the reviews for the encoding.

Technical Write up

Using the all-MiniLM-L6-v2 encoder, 4,500 different data points such as average ratings, reviews, duration, and price have been converted to 786-dimensional vectors. Afterward, a clustering algorithm (K-means) was applied to check which courses were most similar to each other with a total of 240 clusters.

Because there were other features in correspondence with every sample, we have been using this extra data to exploit the best from the clustering relevance ai application to show additional data.

Solution

The clustering app shows the representative samples for every cluster, including all the corresponding metrics used as additional insights.

Using textual data, we can extract zero-shot labels to build a WordCloud of the most powerful words for every cluster, which makes it much easier for us to identify the content of each cluster.

Winning at Udemy: What the most successful courses have in common
Andres Slaughter
March 18, 2022