Have I Found Out The Best Time To Publish
(It's not what you think )
Have you ever wondered if there's a perfect time to hit “publish” on Medium?
In his short and snappy article, Daniel Teurtrie confidently asserts that the best times to post are during the morning and evening in US time zones—when most readers are waking up or winding down. It sounds convincing, right? But here’s the catch: he didn’t offer any data to back it up.
As a writer always curious about the nuances of content engagement, I couldn't help but dig deeper. Could there be a magic window for publishing? Or is the quest for the "best time" far more complex than it seems? Spoiler alert: it’s the latter.
Armed with data from over 18,700 stories written by the top 30 Medium authors—whose collective influence includes more than 2.18 million followers, 28 million claps, and 330,000 responses—I set out to put Daniel’s theory to the test.
What I found was surprising and a testament to the intricacies of timing in content creation. The question of "when to publish" reveals layers of complexity that go beyond simple time slots.
Here’s what I discovered along the way.
The pursuit of the best time to publish
For writers, bloggers, and content creators alike, discovering the “best time to publish” can feel like uncovering a secret weapon. Timing, after all, could mean the difference between a story that quietly fizzles out and one that garners massive attention.
Every writer has probably asked themselves: When will my readers most likely engage?
Daniel Teurtrie confidently asserted that the magic hours are morning and evening, specifically US time. He didn’t mention which time zone, though.
He reasoned that these are moments when most readers have the time to scroll—first thing in the morning, coffee in hand, or at night, as they wind down. But while his theory makes intuitive sense, his article left me wanting more. Where was the evidence? What data supported this claim?
That’s when my curiosity kicked in. I wasn’t just looking for a rule of thumb—I wanted hard data. If there was an optimal time to publish, I wanted to know precisely when and why.
So, with my expansive dataset from Medium’s top writers, I set out to dig deeper.
I aimed to verify Daniel's claim and see if there was more to the story. Could the “best time” be influenced by factors that were overlooked? And could a single answer even exist for everyone?
I quickly learned this pursuit wasn’t as simple as I had thought. It was time to let the data speak.
Data-driven investigation
To test Daniel's claim, I created a database extract from 18769 stories, all written by the 30 authors on Medium. I narrowed my focus to the critical engagement factors:
Claps: The sign of reader engagement and Medium’s closest measure of a "like."
Responses: A reflection of how thought-provoking or discussion-worthy a story is.
Deep dive into the data
I created a Jupyter Notebook to analyze this dataset.
One of the first things I noticed was that the histogram of “claps” and “responses” shows that data doesn’t follow a normal (gaussian) distribution. There are only handful of stories with over 10,000 claps, and thousands with less than 1,000 claps.
I took a logarithm of these values to transform the distribution.
I removed outliers with a Z-score larger than three and fitted a Gaussian (normal) distribution using the log_claps data. The transformed log_claps histogram is relatively close to normal but skewed slightly to the right.
I did a similar transformation for log_responses, which didn’t look like a Gaussian (normal) distribution at all.
Given this insight, I continued with log_claps data as an engagement metric, as many statistical analysis packages like ANOVA assume a normal distribution in the underlying data.
Hypothesis Testing
To formalize this investigation, I created a Null Hypothesis and an Alternative Hypothesis:
Null Hypothesis (H₀):
The times of day (morning, evening, etc.) do not significantly affect the level of engagement (e.g., claps, responses) on posts. In other words, posting during the morning or evening does not lead to higher engagement than other times.
Formalized Null Hypothesis (H₀): There is no statistically significant difference in engagement (e.g., claps, responses) between posts made during the morning and evening versus other times of the day.
Where
left part is the average engagement (μ) during the morning or evening.
right part is the average engagement (μ) during other times (afternoon, night, etc.).
Alternative Hypothesis (H₁):
Posting during the morning and evening leads to higher engagement (e.g., claps, responses) compared to other times of day.
Formalized Alternative Hypothesis (H₁): There is a statistically significant difference in engagement, with higher engagement during the morning and evening compared to other times.
This sets up a hypothesis test where you can compare engagement metrics across different time slots (morning, evening, afternoon, and night) using a statistical test.
Running the test
I converted the publishing datetime from the UTC epoch to the US Eastern time zone and extracted the hour of publication. I used the following 6-hour time slots in my evaluation
Morning: 6 AM - 12 PM ET
Afternoon: 12 PM - 6 PM ET
Evening: 6 PM - 12 AM ET
Night: 12 AM - 6 AM ET
The goal was to see if there was a noticeable correlation between when a story was published and how well it performed.
Early on, I started uncovering patterns—some expected, some surprising—but the deeper I dug, the more complicated the picture became.
Daniel’s theory was beginning to show cracks.
The volume of stories by the time
I plotted the number of stories per time slot. The number of stories published in each time slot is very different.
I plotted the number of stories published per hour and found this pattern below. Stories are published at all hours of the day, and the top 30 authors are based in the US, Europe, and Asia-Pacific regions. However, the volume of stories published per hour peaks between 7 AM and 3 PM ET.
The engagement by time
To accommodate this variation in story volume per hour, I wanted to normalize the log claps by calculating the average number per time.
I plotted the average log claps per time slot, and the result shows little difference in claps. Morning (6.538) is slightly lower than Afternoon (6.609) but higher than Evening (6.513) or Night (6.482). Are these minor differences statistically significant?
Finally, I created a boxplot showing the log clap quartiles by time slot. The median value (6.654796) for morning is slightly lower than for afternoon (6.716595) and slightly higher than evening (6.650279) or night (6.582025). But is this difference statistically significant?
The Kruskal-Wallis and ANOVA test results
I ran two separate tests to evaluate the statistical significance of these results.
ANOVA tests differences in means assuming normality and equal variances.
Kruskal-Wallis tests differences in medians without assuming normality or equal variances.
Here are the results:
So, did I bust Daniel's claims that morning US time is the best time to publish?
Yes.
According to this study of 18,434 stories, publishing in the afternoon between 12 PM and 6 PM Eastern Time is the best time to get the most claps.
Both ANOVA and Kruskal-Wallis tests yield p_value smaller than 0.05, meaning these differences are statistically significant.
Why timing isn't as simple as you think
What started as a simple question— when is the best time to publish? —quickly turned into a more complicated investigation.
At first glance, it seemed straightforward: post at a particular hour and watch the engagement roll in. But as I began analyzing the data, I realized there wasn’t a one-size-fits-all answer.
The role of variables
Timing alone is one of many factors at play. Several variables began to skew the results, revealing that the "best time" to publish may not be as universal as it seems:
Time Zones: Imagine sending a birthday invitation; if it arrives at different times in different cities, not everyone will celebrate together. Similarly, a story published at 9 AM in New York City reaches readers in San Francisco, London or Sydney at other times. Understanding your audience's geographic diversity is crucial. Daniel lives in France, but he claims the majority of the Medium readers/writers live in the US.
Audience Demographics: Consider your readers different coffee drinkers—some enjoy their brew in the morning, while others prefer a late-night espresso. Different audiences have different habits; night owls might engage more in the evening, while others check their feeds during work breaks. Knowing your specific audience’s behavior matters just as much as the time you publish.
Story Topic: Content is like a playlist; each genre appeals to different moods and moments. A personal development story may perform differently than a tech article, as each genre attracts readers with varying reading habits. This adds another layer of complexity to finding the ideal time to publish.
Publications: Publishing within a Medium publication can complicate timing. While publication potentially offers a broader distribution, the author's control over when to share their work is handed over to the editor.
So, is there a “best time”?
After all the analysis, I began to see patterns emerge—but not necessarily the ones I expected.
The average claps across all authors did show a small, but statistically significant difference - publishing in the afternoon between 12 PM and 6 PM Eastern Time is the best time to get most claps. The median (50th percentile) claps exhibited the same pattern.
The caveats
While these insights are valuable, they come with limitations. The “best time” isn’t a one-size-fits-all answer. It can depend on the type of audience, story topic, and whether or not the story is published under a publication.
Additionally, outliers and unpredictable, viral stories can skew results in unexpected ways. I removed all outliers with claps Z-score over three, to be able to run this analysis.
Ultimately, finding the perfect time to publish isn’t just about choosing a specific hour—it’s about understanding your audience, testing different times, and adjusting based on the data you collect.
The research complication: pursuing the truth
Data analysis has a way of transforming what appears to be a simple question into a web of nuances and contradictions.
At the outset, discovering the “best time to publish” seemed like a straightforward quest for an optimal hour. But as I dug deeper, the edges blurred, and what once seemed like a crisp answer gave way to a kaleidoscope of variables.
Every layer of analysis uncovered new factors—time zones, audience habits, topic preferences, authors —that quietly influence outcomes. And for every new insight, the list of variables that could skew results grew longer. What started as a clean data set became a world of messy, interconnected elements. The deeper you go, the more the picture shifts, and sometimes, instead of clarity, you find new questions waiting at the bottom.
It became clear that this pursuit isn’t just about finding one answer. The idea of a universally “best” time to publish may be a myth or an oversimplification.
What works for one writer may falter for another, and an audience responding to tech news at 10 AM might need to be more active regarding lifestyle stories.
It’s not about a golden hour but about understanding your and readers rhythm .
For each writer, for every story, the “best time” is a shape-shifter. It depends on who you are, your readers, and what you share with the world.
In the end, searching for the best time isn’t about a clock—it’s about how you resonate with the people waiting on the other side of your words.
Conclusions and Takeaways
After examining the data, one thing is clear: finding the “best time to publish” is far more nuanced than finding a quick fix or magic hour.
While there are trends—like lower claps at night and higher responses during late hours—no universal rule works for every writer or audience. The data reveals that engagement is influenced by countless factors, from audience habits to the nature of the content itself.
Advice for Writers
Instead of chasing a single answer, consider these actionable steps:
Know Your Audience: Know where your readers are and when they engage with your content. I created a free tool for this purpose.
Test and Learn: Try publishing at different times and tracking your engagement metrics. Look for patterns in your own data—don’t rely solely on general advice.
Consider Your Content: Different stories may resonate at different times. Short reads work better during commutes, while long-form pieces may get more traction in the evening.
Stay Flexible: What works today might not work tomorrow. Audience behavior can shift, so regularly revisit and adjust your strategy.
Whenever you're ready, there are 2 ways I can help you:
When want to boost your audience, engagement, and growth, here are 2 essential resources I’ve designed to help you succeed.
1. Top Authors Dashboard: Unlock insights into the strategies that successful writers use to grow their audience and keep readers engaged.
- Master Consistency: See how consistent posting builds a loyal following.
- Craft Engaging Articles: Learn the techniques top authors use to captivate readers.
- Explore Top Publications: Discover where influential writers are contributing.
- Optimize Titles & Tags: Uncover which titles and tags drive the most engagement.
Get FREE access to the Top Authors Dashboard and start optimizing your writing today!
2. Custom Chrome Extensions: Save time and effortlessly export your Medium data with these simple, efficient tools built for content creators.
- Medium Followers Exporter : Export your followers’ data directly into a CSV file.
- Medium Engagement Metrics: Export your Medium engagement metrics—claps, responses, and more.
- Medium Earnings Exporter: Export stories, links, publish dates, and monthly earnings in one click. Check out this guide to learn how to analyze your data.
Don’t miss out on these valuable tools—take the next step toward growing your writing career today!
This is called deep data backed research to invalidate a hypothesis. Well done again Finn hats off
This article is so insightful, and it did cost plenty of work. You not only provided insight, but explained the full path of logical reasoning leading to those insights, including the most difficult and time consuming - data preparation. Well done and thank you.