Requiem for an Online Publication
Part II: Audience
While each of the articles in this series stands on its own, readers may want to read Part I: Overture before proceeding. Note also you can click on any image below to view a high resolution version. — TCG
I would have thought a well-established publishing platform like Medium would be able to tell me something about my audience.
However, beyond some broad and vague generalisations, Medium can’t tell me much of anything — at least anything they are willing to share — about precisely how many distinct human beings have read any, some or all of what I have written. So, when employing Medium as the platform for a digital-first publication like the New RC Soaring Digest (New RCSD)— the defunct periodical which is the subject of this series — there were significant challenges in figuring just how big the audience there was for the precious words and pictures the contributor community laboured so long and hard to produce. Never mind being able to tell potential advertisers the same thing, at some point.
Before I continue, though, I can almost hear someone out there say, incredulously:
“You published this rag for two-and-a-half years and you never knew how many people were actually reading it?”
Nope, I’m afraid not.
Out-of-the-Box Medium Statistics
Actually, it’s not quite as bleak as that. Medium does provide three basic metrics to all authors using the platform: Views¹, Reads² and Fans.³ For these, they provide totals for all time unless you’re willing to read and transcribe numbers off a dynamic line chart which will provide these metrics broken down by day.
The picture is ever so slightly brighter for Medium Publications. In addition to the metrics above, for all publication stories they also provide Minutes Read,⁴ Views⁵ and Visitors⁶ as shown in Fig. 1.
This is not an exhaustive list of the metrics provided by the Medium’s stats screens. Rather, they are the ones which which were employed the most over the New RCSD’s run and the ones which are most directly relevant to this series of articles. I have footnoted, below, the Medium-provided definitions of each of these.
The Trouble with Medium’s Visitors⁶ Statistic
The most startling deficiency with Medium statistics is their inability to convey anything more than the “number of unique daily visitors” which they quickly qualify with “[each] visitor is counted once per day, even if they view multiple pages or the same page multiple times.”
To illustrate the depth of this weakness, the number of Visitors recorded for the New RCSD on January 31, 2021 — the first day for which statistics were recorded — was 1,232. On February 1, 2021 — the very next day — there were 764 Visitors. The critical statistic which is not available is how many of those 764 on the second day are the same — or different — from the first day’s Visitors.
If you assume all 764 on the second day were so impressed with what they read on the first day they came back for another look — that is, 100% audience overlap from the first day to the second — then the maximum size of the audience is still 1,232. On the other hand, if the good news of the resuscitated RCSD spread like wildfire and 764 brand new audience members showed up on the second day, then the maximum size of the audience for that two day period is 1,232 + 764 = 1,996.
That’s all I knew for sure. Over the first two days, the size of the audience was no less than 1,232 but no larger than 1,996.
Let me add a third day and it will become obvious how woefully inadequate these numbers really are. On February 2, 2021, the Visitors was 584. The minimum size of the New RCSD’s audience is yet again 1,232 whereas it could be as high as 1,232 + 764 + 584 = 2,580. Without knowing what the audience overlap from one day to the next, this wide range is the only ‘known known’.
On October 7, 2021, the New RCSD hit its high water mark for Visitors of 1,339. Consequently, this is the smallest the New RCSD’s audience could have been based on Medium’s definition of Visitors. At the other end of the spectrum the total is almost laughable — the sum of Visitors for every day of the the entire run of the New RCSD was — get this! — a whopping 170,716. In the highly-improbable-but-not-impossible event there was no overlap from one day to the next for the entire run of the publication, the size of the audience for the New RCSD would indeed be 170,716.
Jackpot!
Not so fast. When often asked the seemingly straightforward question “so how many people read the New RCSD?”, the only absolutely accurate response I could provide was “somewhere between 1,339 and 170,716”. Of course it’s a ridiculous answer and as such, one I never provided to anybody for any reason. Until now, of course.
Copy-n-Paste Data Integration
Given that nailing down the total size of the audience was impossible, I turned my attention to Medium’s metrics and derivations from them which could provide further insight on the New RCSD’s readers.
Just before I get to the net output of this exercise and the resulting lessons learned, there’s another deficiency in Medium’s stats which — while certainly not as intractable as the problem above — still proved to be a significant make-work project for the duration of the publication’s run: Medium provides a maximum of 90 days history of any their stats, coupled with the glaring absence of any sort of statistical data export facility.
To collect the entire history of these stats for the New RCSD, I was obligated to periodically pick the daily numbers off Medium’s dynamic histograms, as shown in Fig. 1, and manually transcribe the data to an external, home-brew spreadsheet for subsequent analysis as shown in Fig. 2.
Incidentally, there were one or days in two-and-a-half years where I forgot to collect this data for greater than 90 days and seemingly lost a small amount of data in the process. These small gaps were plugged with some basic interpolation. Not perfect, of course, but likely good enough statistically.
The Immutable Release Cycle Power Law Curve
As the data collected as described above began to accumulate, the most obvious pattern which leapt off the page was how predictablyVisitors decayed over the course of the release cycle. Here is a plot of the raw data for each of the 30 releases of the publication:
Even a casual observer can quickly see in this noisy graph that the vast majority of Visitors show up on the first few days of the release cycle. The only notable exception above is April, 2023 — shown as a bold red dotted line — which had a relatively slow opening day followed by two secondary peaks around day six and day nine. This is seemingly the dramatic effect of social media ‘pushes’ where all the stops were pulled out to get the word spread as far and wide as possible via organic social posts on a variety of platforms. This will be covered in more detail in an upcoming Social Media article in this series.
Here is the basically the same chart except it’s ‘de-noised’ by averaging the daily data for each of the 30 releases into a single number for each incremental day of the release cycle.
It highlights two apparent anomalies: the first is that there is an uptick from the moment a new issue was released to peak Visitors. This can readily be explained as a reflection of the precise time a given issue was released over the course of the day. As was often the case, the workload associated with getting a new issue out meant the click — figuratively speaking — of the Publish button crept towards midnight on many occasions. This meant there was only a couple of hours — in some cases minutes — to collect Visitors before the clock struck midnight and rolled over to the ‘second day’ of the release. If that was the case, it was almost inevitable the numbers from the second day would exceed the first.
The other apparent anomaly is that somehow Visitors activity picks up sporadically at the end of the month. Again, this can be readily explained as a timing issue, similar to the above, related to the precise time when the subsequent issue was released coupled with Visitors being dumped into 24 hour buckets which inflexibly begin and end when Medium says they do. Sometimes the jump in Visitors crept into the day before the day deemed to be ‘release day’.
However the numbers bear out, qualitatively speaking, that the release cycle is a big bang at the beginning followed by an initial, precipitous decline in readership. This is followed by a smooth but flatter, steady decline to near zero until such time that the next issue comes out. This is a power law curve and is shown with the red trend line in Fig. 4 above.
Just one more thing — how is that a month can have have more than 31 days as these charts indicate? Simple — sometimes issues ran late in relation the preceding issue meaning the total period between issues was actually longer than a month.
Here’s the most important lesson learned from all of this: for the New RCSD there was a very short window to get the word out after a new issue published. As shown in the chart above — which is based on the averages shown in Fig. 5— half of all Visitors showed up in the first seven days of the release of a new issue. This threshold is indicated with the dotted red line in the diagram above. This behaviour repeated month-in, month-out almost regardless of how much effort was expended pushing the new issue on social or using other means.
Growing the Footprint
If Medium’s statistics can’t provide an absolute number for the size of the audience, almost as important was whether the New RCSD’s audience—whatever that might be — was growing, shrinking or treading water.
I was able to shed some light on this by plotting Visitors over time as illustrated in the following diagram:
Given the power law behaviour explained above, it’s easy to spot and understand the monthly peaks in Visitors in Fig. 6. These have been smoothed out a little by taking a seven day moving average of these values. The trend line, shown in red, shows steady growth in Visitors at a rate of about 5.7% per year.
Going back to the original source data, the New RCSD received an average of 193.6 Visitors each and every day over the course of the entire, two-and-half year run of the New RCSD.
Having created something from nothing, it’s a number of which I’m still proud.
In Praise of the Periodical
Although it will be covered in more detail later in this series, it was a carefully considered decision to retain the monthly format of the New RCSD. The legions of readers of the legacy publication — which ran 12 issues a year continuously from 1984 to 2018 — would find the once-a-month ‘heartbeat’ of the New RCSD familiar and comfortable as and when they discovered the new version of their favourite journal.
It seems when legacy publications go online, often the first thing they toss over the transom is their periodical nature and begin pushing out stories and other content as-and-when it’s ready. Sometimes even sooner than that. If you’re a 24 hour news operation that actually has 24 hours worth of news to report, that’s important. For almost everyone and everything else, a periodical may well work better.
The power curve analysis which is the central theme of this article also speaks to the benefits of retaining the New RCSD as a monthly. Not publishing anything between issues built expectation, delayed gratification and made new issues a real event which inevitably got talked up on social platforms. It was predictable, free publicity!
I encourage future publishers of online journals to consider retaining periodical releases — whether they be weekly, monthly, quarterly or whatever. At the very least you will know exactly when that critical seven day power curve window opens and closes, and around which you can plan the other activities of your release cycle. Continuous publication means every day is release day — sounds great except that you have just taken away a critical focal point for the readership. Not to mention creating a workflow which is effectively unsustainable for a publisher of modest means. Like me.
Hopefully Medium recognises this reality and puts some tools in place to help manage down the huge amount of work resulting from working around this really fundamental deficiency in their platform. The world really does need more carefully considered, prepared and edited stories as opposed to uncaptioned, unedited image dumps ten seconds after something happens.
Next Time
One thing noticeably absent to this point is anything other than references to generic ‘readers’ and ‘audience’. Also of significant interest — particularly to potential future advertisers — is some high level, GPDR-compliant⁷ detail about the readership’s demographics — who they are are where they are would be a good start. In the next article in this series — Analytics — I’ll be taking a look at some of the additional tools I employed in this regard, as well as how readers ‘consumed’ each issue of the New RCSD.
If you have any comments or questions about any part of this series — or on the subject of online publishing in general — please consider leaving them in the Responses section below so that all readers can benefit of both your question and my answer. You can find it by clicking the little 💬 below.
Thank you so much for reading and see you next time.
©2023 Terence C. Gannon
Footnotes
¹ Views — “Number of visitors who clicked on a story’s page.”
² Reads — “Number of visitors who have read the entire story (an estimate).”
³ Fans — “Number of unique readers who clapped for this story. Note that each user can clap up to 50 times for a story, but they will be counted as one fan.”
⁴ Minutes Read — “The total amount of time spent reading your publication.”
⁵ Views — “The total number of views your publication has received on all posts and pages.”
⁶ Visitors — “The average number of unique daily visitors who have visited your publication. Each visitor is counted once per day, even if they view multiple pages or the same page multiple times.”
⁷ General Data Protection Regulation — “the toughest privacy and security law in the world. Though it was drafted and passed by the European Union (EU), it imposes obligations onto organizations anywhere, so long as they target or collect data related to people in the EU”
Resources
- The New RC Soaring Digest — It’s beyond the scope of this series to discuss in detail the subject matter covered of the New RC Soaring Digest. However readers with an interest should simply visit the site. We were very proud of what we accomplished with this humble journal!
Do you want to launch an online publication and feel you may be able to benefit from some of my hard won experience? If so, please do not hesitate to contact me to discuss whether I might be of assistance.
Subscribe to the series mailing list to be notified when new parts are available and other updates. A PDF version of this article is available upon request.
