Forem Creators and Builders ๐ŸŒฑ

Cover image for ๐Ÿ’ฌ Forem Analytics - Visualize the Trends on ๏ธ๐Ÿฆ„๐Ÿ”–
Hung Vu
Hung Vu

Posted on • Originally published at

๐Ÿ’ฌ Forem Analytics - Visualize the Trends on ๏ธ๐Ÿฆ„๐Ÿ”–

In the past few years, has become a big player in the international tech blogging community, with a focus on software development regardless of the industries. With audiences from all around the world, I find it to be an inclusive community with diverse user backgrounds and is surely a great place to visit during your dev journey.

Despite having my own blog page, is where I have conversations with other members more often. Being a part of the community, I like to learn more about the trends to further improve the performance of my content. As is built on Forem, it has public APIs that can provide me data-backed guidance, and so, I created an open-source Forem Analytics web app for this very, well ..., analytic purpose.

This article only focuses on, but the web app also works for various other Forem communities like CodeNewbie, Wasm Builders, and more.

Table of contents

Prepare the data ๐Ÿงฎ๏ธ

To learn about current trends, it is better to not include old articles (e.g., published years ago). Unfortunately, there is no API to get articles based on a date range. That said, I can get published articles sorted by published date. As of April 2022, has roughly 5000 new articles per month, the two-month worth of data is about 10000 articles, and so on.

// Total article = Number of pages x Number of articles per page
Enter fullscreen mode Exit fullscreen mode uses heart (โค๏ธ), unicorn (๐Ÿฆ„), and save (๐Ÿ”–) as basic metrics to measure the popularity of an article. They are reactions, but let's not forget the comments (๐Ÿ’ฌ) too since it is essential to proof of user engagement. There is no downvote system, so for the mentioned metrics, the higher the counts the better. With that said, I grouped the metrics into three different categories.

  1. By the published time
  2. By article estimate reading time
  3. By tags

You can freely request as much data point as you want. At least up to now, I have not faced any request limit restriction.

Fetch articles from API public endpoint

When should an article be published? ๐Ÿ”‘

Finding out the most active time of the userbase is a good way to improve content performance. If I post an article during midnight, most people have already gone to sleep and by the morning, it is already flushed by several other articles.

That said, considering is an international platform, its users are from different time zones so just judging with my own local time is not a good generalization. To present the relationship between user engagement and article published time, heatmap is a great choice because it shows density at a glance.

Reactions summary by published time (summation)

Comments summary by published time (summation)

The graphs are displayed with my browser's local time zone, which is PST in this case. The date was manipulated with date-fns, so it should be in your local time too. With a sample size of 14850, I have roughly 3 months of data, and using a Z-score of 3, some top-performing articles are removed. Interestingly enough, it appears the articles perform better between midnight to noon from Monday to Friday PST, about midday and after that in GMT, or is it? ๐Ÿค”

The calculation as of now is using summation, which is heavily affected by sample size per time slot. In other words, if 1000 articles are being posted at noon on Monday, it is likely to have a better total engagement counts than only 10 articles being posted on Sunday at midnight.

Let's change the calculation method to mean and try to increase a minimal number of published articles per time slot to 30 to reduce the chance of one well-performed article affecting the average.

You can customize calculation method, Z-score, and minimal sample size per group in the panel below.

Changing the calculation approach

And the results are:

Reactions summary by published time (mean)

Comments summary by published time (mean)

It seems there is no clear trend with the mean calculation, but is solely based on the reactions summary, I guess Monday - Friday, 6 AM to noon (PST) seems to have consistent performance across the boards. Meanwhile, the weekend, especially Saturday, seems to have a lower engagement rate overall.

How long should an article be? ๐Ÿ”‘

This depends. If the audience is researching a solution to a complex problem, then a 15-minute article might not be long enough, but if they just want to know what JavaScript loop syntax is, maybe a less-than-1-minute article is more than sufficient. Let's see what the data says then.

Reactions summary by reading time (summation)

Comments summary by reading time (summation)

I'm not so surprised to see it is right-skewed. Long articles require much more effort so not many people are going to write one. Short listicles are more of a common thing on in my experience. Unless there is one specific topic to dive in, then it is unlikely to have an article longer than 7 minutes (>=1500 words). However, to know that the reading time goes up to 100+ minutes, well ..., that is something. ๐Ÿฅถ๏ธ

Anyway, let's calculate by mean and change minimal number of published articles per given reading time to 30.

Reactions summary by reading time (mean)

Comments summary by reading time

By increasing the minimal sample size per reading time group, I remove the statically insignificant data points. The reaction count seems to be increased along with reading time, which is to be expected. As said, long articles usually have their target audience in mind and are even SEO-optimized, so naturally, the engagement rate increases. This peaks at 13-minute with an average of 10.3 reactions, but I think 5-minute of reading time is a more reasonable length for most content creators out there.

On, you can give 3 reactions to any article, even your own. By writing a 13-minute article, I guess you are likely to have at least 3 readers enjoying your article. In contrast, there seem to be no definitive trends in the comments count.

Which tags should you use? Which topics should you write about? ๐Ÿ”‘

Everyone can use up to 4 tags in an article, but first and foremost, they must be relevant to the article. Sometimes, many of them fit so you may not know which to choose. If that's the case, how about using a data-driven choice instead?

Using the same data set, here is what we have judging by summation.

Reactions summary by tags (summation)

Comments summary by tags (summation)

These are mostly generic web development and computer science tags. Undoubtedly, among the 4500+ tags, these are the most popular on However, that creates the same problem as reading time (heavily skewed distribution). With that said, let's try finding the mean with at least 30 published articles per tag.

Reactions summary by tags (mean)

Comments summary by tags (mean)

Now we only have about 140 tags left. Interestingly the top three tags this time are:

  1. watercooler
  2. a11y
  3. discuss

Come to think of it, two out of three are specifically designed to start a discussion/conversation (the recipe for engagement), so I guess it's normal for them to be at the top. I did not expect a11y to be there, but hey, accessibility has always been an important aspect of software development, right?

Some other well-performed generic choices are:

  • todayilearned
  • career
  • tooling
  • writing

Wrap up ๐Ÿ“ˆ

There you have it, a short data-backed analysis to start as a content creator on However, should you listen to whatever I say here? Probably not, for a few reasons:

  1. The analysis can certainly be biased or fall into fallacious interpretation traps. For example, is a sample size of 30 per group a reasonable choice?
  2. Mistakes in implementation. Yes, I did not use any data analytics library so there is a chance an implementation is broken somewhere.

All in all, the analysis can only be your assistance at best. You need to determine what your audience actually wants, be it topics, article structures, writing styles, and more. High-quality content is the determining factor to bring you success. ๐Ÿ”ฅ

I wonder if this article can go to the moon using the analysis above. ๐Ÿฅบ๏ธ

You can always visit Forem Analytics to generate data at your convenience. It can provide data visualization not just for, but also for various other Forem communities like CodeNewbie, Wasm Builders, and more. Any contributions to the application are welcome!

Top comments (2)

ildi profile image

I love analytics so this is really fun to play with! Thank you for sharing with us @hunghvu. Itโ€™s cool to get a quick summary of whatโ€™s going on within a Forem community by using this tool. Great work!

hunghvu profile image
Hung Vu

Glad to know that it's useful to you ๐Ÿ˜Š