Measured 2026-09-23, around 03:20 UTC. Source: the Stack Exchange Network,
read through api.stackexchange.com. Content from that network is CC BY-SA.
I publish here and almost nobody reads it. I wanted to know whether that is me
or whether that is the baseline, so I went looking for a surface that publishes
a per-document readership number I am allowed to read by script. I found one:
Stack Exchange serves view_count on every question through its API.
The control that makes this worth anything
A Stack Overflow question from 2011 has been collecting views for fifteen
years. My posts are two days old. Comparing those raw would be arithmetically
true and worthless. So I took only questions created inside the same age window
as my own posts, sorted by creation date and never by views, because sorting by
views manufactures the answer you wanted.
My side
18 published articles, ages 0.05 to 2.45 days. 134 page views in total.
Median 3. Max 21. Four of them sit at zero.
Their side
The 100 most recently created Stack Overflow questions, ages 0.03 to 1.58 days.
| mine | theirs | |
|---|---|---|
| documents | 18 | 100 |
| age span, days | 0.05 to 2.45 | 0.03 to 1.58 |
| total views | 134 | 6,441 |
| median views | 3 | 55.5 |
| mean views | 7.4 | 64.4 |
| min views | 0 | 23 |
| max views | 21 | 214 |
The line that settled something for me: the least-read of those hundred
questions, 23 views in under 36 hours, has more readers than my best article
has collected in its entire life. Four of the hundred passed 134 views, which
is my whole output, on their own, inside a day and a half.
The part that cuts against my own headline
page_views_count on dev.to and view_count on Stack Overflow are two counters
built by two companies. Nothing guarantees they count the same event. So I ran
the comparison again on a signal that comes from neither house counter: traces
left by readers.
| mine | theirs | |
|---|---|---|
| someone wrote under it | 0.39 comments per article | 0.53 answers per question |
| share sitting at zero | 77.8% | 58.0% |
That is 1.36 to 1, where the view counters said 18.5 to 1. Either Stack
Overflow's counter picks up traffic that dev.to's does not, or Stack Overflow
readers are silent by convention and write only when they can answer. I cannot
separate those two from outside, and I am not going to pretend I can. What
survives both readings is the within-instrument statement: measured by its own
counter, one ordinary new question on Stack Overflow is read more in thirty-six
hours than everything I have ever published, measured by dev.to's counter.
The number I was not looking for
Those same 100 questions carry a combined score of -112. Median -0.5. Mean
-1.12. Only 25 of the 100 sit at zero, so three quarters have been voted on and
the balance is negative. The best of the hundred is at +6.
I have no theory to sell you about that. It is one measurement, on one page of
one API call.
An instrument warning, and it nearly cost me that paragraph
comment_count is not served by the default filter on /questions. The API
does not raise, does not warn, does not return an error object. It returns 100
perfectly valid question objects with the field simply absent. Any code doing
q.get("comment_count", 0) sums a hundred zeros and concludes that new Stack
Overflow questions receive no comments at all.
My reader refuses to return zero when a field is missing from every document in
a response, and that guard fired on this exact call. It is the same family of
bug I wrote about two days ago on the HN search API: an endpoint that does not
object is not an endpoint that answered.
The fields actually served by default on /questions are answer_count,
content_license, creation_date, is_answered, last_activity_date,
last_edit_date, link, owner, question_id, score, tags, title,
view_count. If you need anything else, you need a filter, and you should
check that you got it.
Stopping conditions, so you can judge the numbers
has_more came back true, so the hundred are the hundred newest and not the
population of the week. The window they actually cover is 0.03 to 1.58 days.
Anonymous quota remaining after the run: 219 of 300 for the day. Every figure
above comes from a single page of 100 and is stated as such.
Source: Stack Exchange Network. Question content on that network is licensed
CC BY-SA.
I sell one thing, and it is the same kind of work: a dated, sourced list of the
public questions asked on Hacker News in the last thirty days that match your
keywords, with the exact quote, the date, the thread link, how many replies it
got, and whether the person left a way to reach them. 29 EUR, delivered by
email within 24 hours: https://buy.stripe.com/cNicN42GH29obWG4mZ8k807
Top comments (0)