Archive of Our Own ^beta

Don't panic.

 

 
Introduction  ^{Contents ↑}
 

TL;DR: there is no “magic window” of posting to maximize your reader traffic as such.  Maybe an initial boost can be had if posting right before or after a daily or weekly TV episode, or a month before or after some big new movie comes out, but in the long run it will all blend into the background (a week, a month, a year later, and the exact hour or day of your post won't really be more than a tiny additive factor, rather than multiplicative — though speaking of, here's a helpful hint: subscribe to yourself, so that you can see the e-mail notification come in, which at least for the past few years that I know of has arrived in the third quarter of the hour [so if you post at any time from a quarter-to through half-past, it won't matter, since the notification comes in at some point between ~XX:32 and ~XX:44, barring one or two exceptions over the past few years... as in the case of this morning with this meta-piece, which I didn't get notification of until 10:50:11, even though I posted at 10:24 — gotta love synchronicity 😂]).

I started looking at my traffic partly out of curiosity in that regard (since some small initial boost might still lead to an extra subscription here and there, sometimes) and partly because I like numbers and wanted to give accurate data in a weekly Sunday Stats thread in a Facebook writers group (which made me wonder when the least active reader traffic flow was, so that I could aim for that and be able to estimate the actual Hit count more accurately if I were to miss it by half an hour or something at some point).

I looked at anything that I could find, from FimFic to ToastyStats to AO3 Admin Posts.  You can see some of the more interesting source pages in the bibliography; some relate directly, some suggest interesting directions to look or ways to analyze the data.

AO3's 2020 graphs (source data here) showed a peak of traffic on Sundays and a lull on Thursday, though Tuesday through Saturday weren't much different from Thursday (scaling AO3's 2.604B Sunday Hits [out of 17.158+B total Hits] as 1, AO3's daily weights for 2020 were as follows: Mon 0.96-, Tue 0.94-, Wed 0.92+, Thu 0.91-, Fri 0.89+, Sat 0.93+); the AO3 days were in UTC, being AO3-sourced, and didn't show more fine-tuned timing than daily, and I couldn't find anything more-detailed (I had presumed there to an hourly breakdown available somewhere, but no dice).
  (And if you're wondering: yes, the annual flow of month-to month shows a usually-steady upward trend [though 2020 had a major spike due to practically everyone cowering in their mom's basement, hoarding a box of Twinkies, because someone told them that the air was made of lava].  Beyond the individual weeks' peak of Sunday and lull of Fri, UTC, the years overall show a distinct plunge in ~1Q, followed by a more-or-less steady rise through the rest of the year, ending higher than it had been the prior year, with the 9-ish upward months reminding me a bit of salmon leaping upstream.)
  By contrast, my own 18 Feb 2021 through 17 Feb 2023 daily Hit totals showed Thursday as the lull (since I take my numbers at the approximate lull traffic time of 0430, this use of “Thursday” here refers to how many Hits occur after Thu 0430 [exclusive] through Fri 0430 [inclusive], Central time — the first few hours' Hits of Central time Thursday still being the very slight ramp down of Wednesday's readers, and the ramp down Thursday night readers of Friday morning being bundled in with my Thursday Hit total), with Saturday doing only slightly better (scaling my own fics' 8,523 Sunday Hits [out of 41,948 total Hits] as 1, my daily weights for 2021-2023 were as follows: Mon 0.77-, Tue 0.65-, Wed 0.65+, Thu 0.58+, Fri 0.62-, Sat 0.61-; you'll see this one farther down, in the weekly ebb and flow section, but if the numbers here and there don't sound like such a great match for AO3's weekly shape, you might be surprised at how much better a match they look if you compare them in the light of 0431-0429 Central [and dealing with the stupidity of DST] w.r.t. 0000-2359 UTC sans DST).
  NB:  Since my records were focused on Central time rather than UTC, and were taken at 0430 (though you'll see shortly that the exact time might not change the results much, due to the variance being small), which comes to a total skew of 9.5 hours for two thirds of the year or 10.5 hours for one third of the year (depending upon Daylight Savings or Standard time, respectively) after UTC midnight, the lull in my week being set more squarely to Thursdays is in keeping with the fact that my Thursday timing is mostly AO3's UTC Thursday night and almost half of UTC Friday morning.

 

For the remainder of this meta piece, and for my own convenience, all times herein will be in Central Time, simply because that's where I live — CDT = UTC -5 summer, and CST = UTC -6 winter — unless otherwise stated; my best guess from my 20 Jan 2022 through 19 Jan 2023 [inclusive] data is that the traffic follows an Eastern [U.S. timezone] day shift-oriented sleep cycle pattern (reinforced anecdotally by an AO3 New Year 2020 tweet).  This might present some issue if we were trying to coordinate something, but you'll find that it's not at all as relevant to reader traffic flow as one might expect (though surely different authors will experience different results depending upon fandoms, perhaps timezone distribution of most of their readers, advertising on Facebook / Twitter / Tumblr / etc., choice of user name and pic / avatar, etc.).

I spent midnight 07/08 July 2021 through midnight 04/05 August (4 weeks, exactly) sleeping in snatches in order to get an initial rough idea of the traffic flow.  Stupid, and not terribly productive, but it gave me some sense of things.  My traffic flow wasn't much, collecting all of 782 Hits.  Alone, that wasn't enough to work out a lot, but I discovered two things:

• for my fics' traffic, there seemed to be a bit of an uptick in flow at the {1100, 1630-1800, 2230, 2300, & 0030} updates (indicating their respective preceding half-hours' traffic rates), with lulls around {0130, 0230, 0500, 0530, 1230, & 2130};
• AO3 updates the Hits' cached data during the second and fourth quarters of the hour (Kudos come from another server, and seem to be fairly real-time upon refreshing your screen), so it's best to check the data during the first and third quarters of the hour — i.e.: 1 minute (not more than 14) after the hour and 31 minutes (not more than 44) after the hour (there were only one or two instances where there might have been a new Hit shown before the 2^nd or 4^th quarters, but I'm fairly certain that these were my eyes playing tricks).

782 Hits isn't really enough to do a good longitudinal study of 48 half-hour spans: that's a mean average of 16.29(+) per span.  Put another way, that's a mean average of 55.86(-) per day over the course of 28 days.  Any way that you look at it, it's not even remotely enough for anything resembling a reliable bell curve, so that makes the first bullet point worse than useless (GIGO: bad data that seem OK can be grossly misleading).

Oh, and as an addendum to the above: on 08 Jan, a friend of mine read one of my fics at 0544 and kudosed it NLT 0549; the Kudos showed up on the 0600 refresh, but the Hit didn't register until the 0630 refresh, so keep in mind that if you're thinking of posting something based upon your traffic times... the traffic data might be mostly accurate, but might be a half hour (or more?) later in your Statistics page than whenever they were actually hit.

 

 
Grave shift: a one year study  ^{Contents ↑}
 

Given all of that, I focused upon the range that I was mainly interested in: 0200-0700, since I was checking my data at 0430, for reasons.  That does seem to be almost the least active — hence least volatile — time of day, but we're not there yet.

You remember my mention of 20 Jan 2022 through 19 Jan 2023?  I spent that time checking my Hits every half-hour from 0200 through 0700 and logging them in a spreadsheet.  There are almost certainly better ways to do this with a userscript or Google analytics or something, but I felt more comfortable doing it this way.  As it stands, there were a dozen or so times when I missed a check-in (napping, or caught up in a chat, etc.), but not many, and these were easy enough to adjust for by telling the spreadsheet cells not to count those points in the totals and averages and such.

The data weren't terribly well behaved, so I can't say with confidence how CST (at only 4 months total) and CDT compare and contrast.  For 20 Feb 2022 through 19 Feb 2023 Hits of 0200-0700, the data started with 20980 Hits and ended with 45717 (this being the 0700 value, not the 0430), for a total of 24737 difference over time.
  For the record, the CST specifics were 598 + 669 = 1267 Hits in 0200-0700 out of 4174 + 4016 = 8190 total CST Hits regardless of time; similarly, the CDT specifics were 226 + 2386 = 2612 Hits in 0200-0700 out of 41578 - 25222 = 16356 total hits regardless of time (not a typo: I put out a Hermione/Harry smut fic's second half at this point, and traffic skewed some, but mostly it's just that there were 122 CST days counted vs. 237 CDT; the ratios are pretty close at total days' 122 / 237 = 0.51+ vs. total Hits' 8190 / 16356 = 0.50+ vs. mid-shift Hits' 1267 / 2612 = 0.49-, though as you'll see below in a moment, their times' peak and trough distributions weren't quite the same, though again, there wasn't much variance in the actual percentages).  Neither of these includes the CST→CDT and CDT→CST changeover days' Hits (and a couple of anomalies, for reasons), nor any of the usual individual cells getting skipped.

Some fics and new chapters did well, some not, or suddenly started seeing traffic after having been stable at a lower rate for months or more, and all of this was spread pretty randomly.  Some days' traffic was nearly non-existent or spiked to 200% for no apparent reason or due to the occasional binge reader.

Below is a heat map of those upticks for 0230-0700.  I'm well aware that everyone else in the world uses a standard red = hot / blue = cold, but I'm just not wired that way; I read blue = high freq / red = low freq.  My apologies, but it's a trivial effort to “correct” for either of us, so I've let it stand.  The same color “reversal” applies to the later heat maps.
  This is that part where I'd normally go off on an ADHD tangent to explain in detail how to set up conditional rules in a spreadsheet, but since this isn't a 101 on spreadsheets, I'm trying to stay focused here... plus, I suspect that while the general approach is probably about the same across the board, the specifics likely vary greatly from one spreadsheet program to another.  They're easy to do though, I promise.  Run an internet search for how to make them in your specific program (or simply open the Help file from the menu buttons within your spreadsheet program), and there will be a number of pages and videos with different explanations (if one of them doesn't make much sense, just try another instead).  In the end, it'll amount to opening some little pop-up editor, telling it something to the effect of IF cell = such-and-such {value range, contents, etc.} THEN {turn a certain color or something} (please pardon the pseudocode), and after closing the rules GUI you can then apply whichever conditional rule you wish to any given cell / range / array; in MS Excel, you can make a bunch of conditional rules, but with OpenOffice, you're limited to three through the GUI and must write a macro if you want more than three (macros are also really easy, I promise, but now we're really going out of bounds for this tutorial; OpenOffice has some pretty good help instructions, and you'll get a lot of help on their forum [hell, probably on Excel's forum, too, and I imagine LibreOffice's forum, though I don't use it, and so can't say either way]; naturally, there are helpful people on OO's / LO's / Excel's / and two Google subreddits [I can vouch for OO's and LO's subreddits], and presumably helpful groups in Facebook and similar sites — as well, you might want to check out OzGrid, where I used to find a ton of useful info for Excel macros back in 2008-2014, which can probably be tweaked a bit for use in OpenOffice or any other spreadsheet program).

Mid-shift reader traffic 2022-2023 (open image in new tab to zoom)

 

You can ignore the pale yellow cells along the bottom, since they're just there to let me compare the sum of the first 5 half-hours' percentage with the second 5's.

At 0.64% difference between the first 5 and the second 5, they're pretty evenly distributed.

So... what does that pic above mean, and why doesn't it show the 0200 that I mentioned?

Well, the numbers reflect only my own fics' reader traffic, so your specific values will almost certainly differ, and your relative rates from one time to another or one day to another might not be quite the same, but here's how to find them.

 

First question:  Count the time slots; 0230, 0300, [...].  There are 10 of them.  Since a 24-hour period has 48 half-hours in it, that means 10/48 = 20.83̅%.  The fact that my fics' 0230-0700 upticks account for 15.87% means that traffic is a bit slow in that range of time.

It also shows that while the mean average 0600 uptick (meaning how many new Hits occurred since 0530) is the least active at 0.98 per half hour, it's really not much less active than 0230's 1.182, which is the most active of this period.  Seriously: a range of 0.202 Hits (one fifth of a Hit) difference between the least and the most active half hour spans, for the midnight shift.  I can pretty much pick any of these times and not see much variation as such; if I needed to guess any uptick within this range of times, I suppose that I could semi-safely (it really does vary on the day-to-day scale) take any one of my known times' upticks and extrapolate from that to the unknown value at a rate of something like 1.099 per half hour.  It wouldn't be dead on accurate, given the non-integer value, and a given half hour can be distinctly higher or lower than it normally is, but it gives a ballpark sense of things.

Below the “Should be” cell, the other four just give the total of those 10 half hour averages (10.791), the average half hour being 1/10 of that (1.079), the average number of hits per day (taking that 10 hour sum and dividing it by the 15.87% cell, yielding 68.016 Hits per day for the whole 2022-2023 period), and the mean average Hits per half hour of an average day (68.016 / 48 = 1.417).  However, since my fics' late night numbers are a bit slower than what a flat rate percentage across the board would dictate, I subtracted the 5 hours' 10.791 from the day's 68.016, and divided that by 48-10=38 half hours, giving us a days-and-swings average half-hour expected value of around 1.506.

Keep in mind that the COMBO column shows the year's raw average value of Hits for a given half hour slot.  I did that because I was interested in the specific average Hits per half hour.  In terms of daily traffic rate ebb and flow though, that might be better served using percentage rates.  We've seen that the mid-shift accounts for only ~15.87% of my fics' 2022-2023, rather than the expected ~20.83%, so here (below) is that same COMBO result converted to percentage of the avg. Hit per day's 68.016:

Mid-shift rates in percentage (open image in new tab to zoom)

 

Second question:  0200 isn't shown here because it's just a starting point.  Every night at 0200, I marked down my total Hits in another part of the spreadsheet; then at 0230, 0300, [...] I marked down the subsequent numbers and had the amount of uptick per half-hour.  In order to know the 0200 uptick, I would have needed to know the 0130 number, and we get an infinite regression (well, 48 half-hour regression).

Every night I did that, so I had column after column like this (marking the head cells with the day and date, so that I could catch my mistakes — always add redundancy: two is one, one is none).  It got a little long, so I ended up doing it in batches of four-month-long rows, but that's mechanically irrelevant to the calculations, just a convenient workaround for me not to have to scroll way the hell to the right (though I did end up freezing the pane, so this precaution turned out to be unnecessary in this case).

When I goofed and missed one, I filled that cell with a bracketed uptick.  For example, I missed 22 Jan's 0530 uptick, so when at 0600 I found that there had been one single Hit, I didn't know if it had just come in at 0530-0600 or had occurred earlier at 0500-0530.  I simply filled that morning's 0530 cell with a question mark “?” (which the tallying cell won't count in its averaging), and filled the 0600 cell with a bracketed “[1]” (which the tallying cell also can't use, but which I can refer to visually, if needed) to indicate that there had been 1 Hit somewhere in those two half-hours.

That left-hand column in the pic, the column named “COMBO”, has 10 cells in it, one per half-hour span.  Each one of those looks at the 365 data points of how many Hits had come in during the relevant preceding half-hour (well, 365 minus data points that got skipped for one reason or another).

 

 
How was all of this calculated?  ^{Contents ↑}
 

Before I go into this in general, please note that I'm using OpenOffice (a freeware equivalent of MS Office; it comes in WIN, 'nix, OS X, 'droid, and ports to other operating systems), so there might be differences between that and whatever spreadsheet format you use (e.g.: LibreOffice [a fork of, and possibly superior to, OO], or MS Office) — for example, OO (and I think LO) uses semicolons rather than commas in order to separate values; oh, and there's also Google sheets.  (And no: I'm not affiliated with OO, LO, MS, Google, or any other software company.)
  Addendum:  Since posting this, I've considered adding a chapter on the basics of building a spreadsheet.  Some readers would need this if they want to follow in my footsteps with their own data.
  On the one hand, it should consider the operations of putting a row of dates in cells at the top of each column that will hold each day's half-hour upticks (you could either count the difference per half hour yourself and enter that raw difference, or enter the number of hits with each new half hour and tell the cell to subtract the previous cell's value), and typing in a column of times at the table's [leftmost] starting column (and perhaps again at the end column) so that you can track each row and not goof up (and if there are a bunch of rows, then you might want to color code the rows so that every 4-5 alternate between black and bright blue font, for example, for easier visual tracking), summing the cell values of each row's entries in another section entirely, and taking their averages out of the entire day's 24-hour span, converting raw numbers to percentages...
  On the other hand, this isn't exactly the place for a Spreadsheet 101 class.  There are also a few key differences between OpenOffice and LibreOffice and MS Office (and presumably every other spreadsheet format out there), and those would have to be considered too, since otherwise someone walking into this fresh off of the street wouldn't know why their formula wasn't working properly (written correctly in one format doesn't mean readable to another format).
  At least for now, I've decided to leave that for the reader.  It's beyond the scope of this tutorial, which is aimed only at how to obtain and juggle one's traffic data in order to get the peaks and troughs, not at the basics of spreadsheet data entry and operations.  Don't worry: there are plenty of tutorials out there for whichever format you use (for example: Microsoft covers things with videos, both OpenOffice help and LibreOffice help have instruction pages for these operations, YouTube has tons of videos on {OO calc, LO calc, Excel spreadsheet, and Google sheets}; likewise, for basic mathematics (and more advanced stuff), if you need to brush up, I'd suggest Khan Academy as a good starting point for study purposes, and 3Blue1Brown has some really excellent videos for quick [re-]familiarization with stat.s, calc., etc.); give it a search and follow their directions. 🙂
  Quick demo provided a few paragraphs down from here.

 

Before we go into the actual spreadsheet, some readers might wonder where the numbers even come from in the first place.  It's a fair question, since not everyone's familiar with their Statistics page, even if they've been on AO3 for a couple of years.  Look for the clickable button that says “Statistics”; if you're on a computer monitor, then it should be along the left side toward the top of the page, and if you're on a 'phone, then it should be in those buttons along the top edge.  Alternatively, here's the URL (sorta):

https://archiveofourown.info/users/YOUR_AUTHOR_NAME_HERE/stats?flat_view=true&sort_column=hits&sort_direction=DESC

And there you go.  Replace the “YOUR_AUTHOR_NAME_HERE” part with your author name (and you'll have to be logged in to do this), and badda-bing, you'll see your stats as of the most recent half hour cache update (though if you really really want to know the absolute most up-to-date info, each fic's data block at the top of their first chapter will show you their most recent data as of your most recent refresh of their page).  You go there, and all of your fics' basic data will be displayed for your perusal... and laborious manual transcription to your spreadsheet.

 

Now, about that spreadsheet and the calculations: look at the COMBO column.  We'll take the first entry as an example, the blue cell that reads “1.182”.  I set the cells to give me three decimal places, but you might prefer two or default, or none.

Inside of that cell, what you can't see is what I told that cell to do.  Ideally, there would be perfect data to work with, no missing bits, no other weirdness.  Strictly, even with values of zero, and text values (say that you wrote a word into a cell, instead of a number, for example), and so on, the =AVERAGE(... function will work perfectly well.  The long and unnecessarily complicated approach below is only useful as an example if you feel a need to micromanage certain aspects of your averages (clearly, I'm a little OCD, but [adding this a bit belatedly] there's reason for explicit declarations in math, rather than assume that a program operates in some specific fashion — and yes: I prefer stick shift over automatic transmission, thanks); otherwise, rest assured that simply using =AVERAGE(... will be perfectly fine for pretty much anyone's purposes.

In case you have some rough idea of how it works, but not quite enough to work it out easily, here's an example of how to do that (and there are always those spreadsheet tutorial links named earlier, if you need more in-depth study material) — these are only random numbers, slotted in for demonstration purposes; the takeaway points are the formulae in the cells and how they do all of the work for you (and my apologies for the perfectly clear offline-pic being turned all fuzzy and shrunken by Pinterest's hosting of it):

Quick snapshot of how to find averages and percentages in a spreadsheet (open image in new tab to zoom)

 

 

Fine, but what about all of the other columns in that big block, and those other two on their own off to the right?

Those other four columns in the main block are:

• the portion at the beginning, while it was still CST at the start of 2022,
• the portion that was at the start of CDT 2022, but before a smut fic that I feared might skew things (being Hermione/Harry-based, I figured that the readership might outweigh the general background traffic... turned out that I was right, but I'm not sure that its traffic times were really any different from everything else's),
• the portion that was CDT while I was uploading smut chapters, even though I didn't bother breaking that from the last month and a half of CDT after having finished with the smut chapters (at this point, I simply figured that any given fic might be boom or bust or anywhere in between, just like the regular punctuations of mysteriously high or low traffic, so this one's effects were as much a part of the whole as any other),
• and the first 2+ months of CST 2022-2023.

Not useful, as it turned out (not without several years of data to establish a decent baseline, anyway), but it was either tell you what they were or leave it unexplained (or waste the effort of re-creating the whole thing without the extra stuff just for the sake of an illustration).  I say “not useful”, but that's for me; maybe for your purposes, such a breakdown could be of value, or perhaps your traffic is of sufficiently high volume and low variance.

Those other two columns on the right combine the CST sections into a single CST column, and the CDT sections into a single CDT column.  Potentially more informative, but not terribly different from that final fully combined form in the “COMBO” column.

 

 
Day-shift's and swings' one year studies, and the eventual combined conclusion?  ^{Contents ↑}
 

Those will have to await Jan(+) 2024 and 2025, respectively.  Assuming all other things to be equal (e.g.: no EOTWAWKI, that I'm still alive, that AO3 is still running), I'll update this meta piece at those points.  I won't go over how to calculate it, since that won't have changed (though I might then move that section to below the combined conclusion), only what the days' and swings' data turn out to be.

Granted, the percentages that 2023-2024's days' and 2024-2025's swings' slots will reflect will be w.r.t. their respective years, so the grand total won't be perfect (it would be damned suspicious were it to turn out so).

Let's say that, by some crazy co-inky-dink, 2023-2024's day-shift study were to show 90% of the traffic to occur during day-shift hours.  That's hardly likely, but not the point.  Had we looked at all 3 shifts at once for the whole year, then we could see how they broke down in that one single solitary same year.  What we have in reality instead is broken into 3 shifts of traffic, with one year per shift, and almost guaranteeing that the percentages that the three take up won't quite sum to a perfect 100% — maybe a touch above or below, and probably not a whole lot more or less than 100, but virtually certainly not exactly 100.

Right this second, I'm only 28 days into this second survey, which is certainly sensitive to statistical variations.  Most cells contain values of 0-5, a few have 6 — but already, just 13 days ago I had a freak event of 25 Hits in a half-hour, and 10 days ago a similar event of 16 and 12 Hits in two back-to-back half-hours.  Needless to say, this will take a little while for the averages to smooth out from this (and maybe this isn't an uncommon event during day-shift?).

Because of that, it would be pointless to include a pic of the distribution so far: from the 0730 uptick cell through the 1630 uptick cell, 3 of those 19 cells are middling values (30-70%) and one is a high value (>70%), the remaining 15 cells are driven into the lower 30% range.  It's safe to say that this is an anomalous skew, and that the final result will surely be a very different picture (presumably reflecting the actual typical high and low traffic points of day-shift hours).

Minor update, 17 Apr 2023:  I've been tracking the day shift data as planned, but since 08 Mar have also been doing the unthinkably stupid [tracked separately, for the sake of comparing methods and subsets] — tracking as well as possible my own fics' half-hourly Hit-upticks and AO3's net fic-count (the final cached tally of all current visible fics minus however many deleted) over each day's full 24 hour span (yes: my sleep cycle is a shambles interrupted by many alarms, and I highly recommend that nobody ever attempt this, unless perhaps they're a peculiar flavor of masochist and/or truly dedicated and not terribly wise, with plenty of time on their hands and nothing else particularly pressing; right this second I have 1157/1195 spreadsheet cells of my fics' upticks and 1709/1780 of AO3's net counts, so I'm tracking 96-97% of the possible half-hour check-ins).  This means that I might have a full year's mostly complete data to present in next year's update, and not simply for dayshift-only span, yes, but moreover brings something interesting to mind: I used to get my Kudos notification e-mails somewhere around half past 4 or 5 (depending upon time of year), but have been receiving them at around half past 6 of late, which matches my fics' approximate least-active time, and “coincidentally(?)” matches perfectly with my observation of AO3's net overall fic increase's least active period (taking up 1.18% of the daily activity) so far being 0530-0600 Central (by following that about a half-hour afterward, giving the cache time to do its thing)...

 

 
Weekly day-to-day ebb and flow  ^{Contents ↑}
 

The number of Sunday Hits is the peak of traffic for me, which at least matches what I'd come to expect from having dug through the stacks, as it were.  If you haven't played with numbers much before, then you might be surprised to find that trying to get a nice, predictable bell curve is a funny sort of art.  Anything under 28 data points is garbage; 28 points often looks like something, but is pretty changeable; 29-ish it really starts to gel; 30+ and you have something that doesn't change a whole lot... when dealing with data points of small value and little likelihood of changing.

That doesn't mean that 30 is some magical number and everything will be stable then.  This example of 30 or so is an alright starting point if you're just looking at a binary situation with what turns out to be fairly similar rates from one set to another — say that you're in charge of training people at your workplace, and you can seat n-many for a session on any given day, and you find that sometimes they're not available (out sick or at an appointment or what have you), then you might find after a few weeks that you can generally count on some approximate percentage of unavailability any time that you go looking for trainees to attend.  With one's daily reader traffic, or AO3's traffic stats in general (Kudos, Comments, etc. per half hour, for example), things are a bit more variable, so 30 really isn't enough to get even a vague idea, but this isn't a STATS 101 class, so let's keep on target here.

You want a lot more than 30 points for a given trend, but consider that in terms of weeks (for the weekly pattern of each day's typical value, or for a given half-hour time slot value varying over the course of the average week).  30 weeks.  3/5 of a year.  You might like the thought of hundreds of weeks of data, but that's not terribly practical if it's just you and your own fics, and so forth.

Since I already had some numbers going in a spreadsheet for each week's Sunday Stats thread in that Facebook group, plus a few sporadic days here and there otherwise, I started keeping every day's 0430 Hit count as of 18 Feb 2021.  Maybe I'll update this sentence later, but right this second (0100 on 17 Feb 2023) that comes to 729 uninterrupted data points (105 full weeks, plus a day — or in other words: each single day's Hit uptick contributes 1/105 to the average for that day of the week, or 0.95(+)%; in other-other words, it's fairly resilient, and would take a day or two of 200-300% [or 0%] traffic just to change the array by a percent or two).

Weeks' average daily Hits and percentages (open image in new tab to zoom)
Sunday's through Saturday's values, plus the mean average in bright yellow at the end.
Thursday's readership is the lowest, when I upload weekly on Sundays.

What does it mean?  Meh, not a lot.

It means that my mean average Monday 0430 uptick is the largest one (meaning that Sunday's actual traffic is the busiest, since I tally Sunday's traffic on Monday morning, the darker blue cell), and that the quietest day of the week for my fics is usually Thursday (Friday morning's uptick in the pink cell).  Here's the thing, though: how much of this happens to follow from the fics being on AO3 (whose day-to-day weekly waveform is very similar to the one above), and how much from the fact that I post at 1030 Central on Sundays — that is, would it show this same pattern were I to instead post on Tuesdays at 0200 or Thursdays at 2000, or would it be shifted n-many days around plus or minus p-many hours, or would it reflect the general background AO3 waveform modified by the fact of my posting at these other times (I suspect this third one to be most likely, based upon data-monkey's write-up, linked a few paragraphs from here)?

That eighth cell pair, in deep yellow?  That's just for shits and giggles, reflecting the mean average of all of the Hits per day since starting this pointless daily record.  Of course, a rolling trend might be more useful, if this were meant to be used for anything, since that started at 600-700 Hits per month at the very beginning, and it's been ~2000±500 since Aug 2021's count.

NB, 27 Mar 2024:  With my year-long AO3-overall traffic study now done (after taking a 2 week break, the numbers on the spreadsheet are all collated, and I'm now writing up the findings and charting a couple of extras, so... ya gotta wait another week or so for that upload), I can give you a peek of what effect my not uploading weekly had on my works' hits (and hence what it might have on yours, or what the overall AO3 background reader traffic might look like).
  In a nutshell, sorta no real effect on the monthly totals, but a smoother distribution across the individual days' mean averages, and the weekly-upload-schedule's Thursday lull in reader traffic is now a background-entropy Friday lull.
  The pic of the daily hits is below, but as for the monthly totals of Wed 08 Mar 2023 (at 04:31 CDT) through Fri 08 Mar 2024 (same Bat-time), they got 25,328 in total; if simply dividing by 12, then that's a mean average of 2,110.6̅ per month; if you knock off the hits that the 3 new uploads got and all 128 hits that came in on one fic's update, then you get 24,548 / 12 = 2,045.6̅ (though if we assume that the updated fic might plausibly still have gotten ~28 hits otherwise, and so knock off only 100 of them, then we get 2,054), which are in the ballpark of the above-quoted ~2000±500 monthly of the preceding two years when I had been uploading weekly.  Note that it spreads the subtracted percentages evenly across the days; that's probably not perfectly true, but close enough, however, if we assume an extremum in which every single one of those extra hits (all 780 for a worst case scenario, not ~680) had been on one specific day of the week every time (and I know that they weren't, but I don't think that it's worth digging to get their actual distro.s, just for this), then that day's 13(+)%-15(+)% of the week would drop by only 2.68%-2.74% to 11(-)%-13(-)% in any event (with the other days gaining a compensatory 0.42%-0.49% each), so not really a huge impact to worry about re. the close-to-accurate assumption of equal distribution when subtracting them.
  Basically, unless I were to write something with some large and lasting effect (as with the fonts tutorial), my monthly/annual numbers aren't likely to suddenly change (which I take to mean that once anyone has a fair number of works done, they can pretty much count on their numbers not being greatly affected), and it doesn't much matter whether I upload weekly or not (so I, and presumably you too, don't need to sweat about meeting some self-imposed deadline, other than possibly frustrating readers awaiting some new chapter).

What does it all mean though, the whole enchilada?  Practically nothing, at the end of the day.  Really it just gives you some vague idea of when you might usually expect to see the highest or lowest traffic each week, and by approximately how much (and if you check the bibliography, you'll see AO3's records show very rhythmic monthly patterns per year).

This could be useful from a server perspective, or maybe marketing somehow, but from the perspective of leveraging increased flow... not so much on its own (unless you could find the “sweet spot” ratio of highest readership uptrend w.r.t. lowest upload rate by bumping it with the upload traffic rates, which I certainly don't have any access to — more power to you if you do, though).  There's an in-depth look at overall stats by data-monkey that's worth reading (and perhaps her code could be modified to target only your own fics' data, though it would still need to check every half hour and so forth — though I suppose that one could take a note from toastystats's method, check the total fics available on AO3 every half hour, and plug them into a spreadsheet, simply to work out the half-hourly upload rates), and one takeaway point is that upload traffic and the reader traffic per day of the week are pretty matched graphs (less related, but worth considering, is that one sees a spike in uploads on New Years day, Valentine's day, Halloween day [though only somewhat], a week before Christmas, and Christmas itself).

All that it really means is that maybe your curiosity is satisfied to some extent (and that's always fun), or that you might find yourself tracking your own numbers neurotically just to see how they change over time or to know not which fic has the highest current totals but what your fics' rates per month are.

 

 
Is there more of a pattern to be had?  ^{Contents ↑}
 

Probably subtle ones, but in this context that kinda rhymes with “noise is louder than signal”.  I'm sure that they could be found, but not with the tools of simply checking my fics' traffic rates.

I don't have a huge number of fics to aggregate the trends, and data mining would be painful at best when considering the different tags and changing trends and contexts of the month or year that any given fic was posted.

I don't post enough holiday-themed fics to compare their readership over time during ramp-up to and ramp-down from any given holiday.  I imagine that there's some sweet spot (too early and you're lost beneath all of the later fics, too late and nobody wants to read it anymore because they're already on to the next one), but that's just an assumption — and some of my Halloween fics see traffic all year long.

These might be available somewhere already, or perhaps could be deduced (or at least inferred) from very large data, such as all of AO3 in general, but that's way beyond the scope of my idle curiosity regarding each day's half-hourly traffic rates.

EDIT, 21 May 2024:  Something that I noticed over the years (and is so much a background fact to me that it never crossed my mind to mention it here) is the question of CTR or click-through rates (which the back of my head insists, for some unfathomable reason, upon thinking of as “cross-clicks”...).  In a tutorial sprinkled throughout with links, the rate is probably(?) higher than in a shits-and-giggles fic, since people are in the tutorial to learn something, and so have good reason to click a useful link to some information source (HTML, math, physics, whatever); in a fic, people are mostly just cruising for fun (and in fics, I link only in the header and sometimes footer, which people often don't read).
  In a lot of my fics I have links to songs that I quote (or paraphrase) a snippet from, and some few of those happen to be part of one of my [NON-monetized] YouTube playlists (only two of the vids themselves are mine, and my channel isn't monetized, nor do I wish it to become so).  When you have a YouTube channel, analytics are accessible for how many of your vids and playlists are viewed in a given span of time, how long they were watched, etc..  The number of hits that these fics receive vastly outweighs the number of views that my two vids or the fic-linked playlists receive, leading me to conclude that while some few readers might appreciate your leaving them a link to a song (or survival training vid, etc.), it's probably not going to be many.
  I can't speak to how many/few clickthroughs a tutorial's various links might see (my tutorials are loaded with them, but none of those go to my YouTube channel's videos or playlists, and that's the only place for which I have access to metrics), but I suspect that the tutorials' CTR is higher than that of my fics' linked song vids, and that the same likely applies to links leading to fics related to one that the reader just finished reading (this latter hunch-turned-conjecture is a little anecdotal, based upon a small increase in traffic after linking similar or related fics in other fics' end-notes).  This looks as if it might be slightly backed up by my playlist of [others'] survival videos when linked in a chapter footnote of some relevant fic, though that in turn might be contradicted by the lower rate of correlation with thematically similar/related reset tech and food preservation playlists (though I think that these have fewer relevant fics, which might contribute to lower view rates).

  My point about click through from one fic to another is that while it probably won't cause your traffic to skyrocket (long-time users will usually, though not always, know already to look at fics' header data blocks [and the smallish series-footer] to see if a fic is in any series and/or collections, and to click the author's page to view their entire list of works), it might help a little (new users don't know about these options, and many years-long account-holding readers and even authors often don't), so it's still reasonably worthwhile (being an easy thing to <a href="https://some.url.com/somewhere" rel="nofollow">add a link going somewhere</a> or <b><i><a href="https://archiveofourown.info/works/88888888" rel="nofollow">link another AO3 fic</a></i></b> [or series, or collection] if it's related in some way) to do so for your hits and readers' enjoyment.

 

 
Closing notes  ^{Contents ↑}
 

These don't precisely relate directly to the number collecting and analysis, but they played their part in some ways that just don't fit anywhere else, so I've stuck them together here.

 

I use different browsers for different purposes.  For AO3, I've found that Chrome serves my purposes reasonably well (though yes, each browser has its pros and cons, and in Chrome's case, I had to write a style sheet [sort of like a userscript] specifically to force links to be blue-underscored, since Chrome [currently] lacks user settings for that).

I can't say for other browsers, but since Chrome is what I use for AO3, and because I keep several tabs open to the Statistics page (for reasons to do with comparing individual fics' upticks and such), I found that sometimes the tab basically goes to sleep (as with Opera and I think Firefox; I don't use MSIE, Edge, or TOR [or Safari] except to test things, so I can't say for those) and will auto-refresh upon my return if I don't keep that tab displayed while I sleep, and return to it periodically (maybe every hour or two?) while awake.

No thanks.

What's the solution?  Go to chrome://discards/ and toggle any relevant AO3 tabs to “X” (off), so that they don't refresh their caches.  Now if you go and refresh one of your Statistics tabs, the others won't refer to that one's latest data and refresh themselves (I'm really not sure if Firefox's about:performance has some equivalent of this).

Maybe.

At least, that was what was supposed to happen, and it had been functioning as advertised for the past 2 weeks.  That said however, it seems to have glitched this morning (I started this meta's initial 5.2K on the evening of Fri 16 Feb 2023, just before 2100, but didn't finish 'til almost 0700, then copied and pasted to AO3 and pre-formatted it 'til 0930 for the coming Sunday post... so yeah, as is so often the case: I literally wrote this instead of sleeping, now that I'm trying to adjust to staying awake during daylight hours for the 0700-1630 data checks), having self-detoggled a couple of my Statistics tabs back from “X” (off) to being checked on again (really gotta love synchronicity 🤣).  I guess that I'll just have to keep coming back to that one tab that I don't want to see refresh its cache.

Still though, at least it's been reliable for two weeks, so perhaps it will help somewhat to keep the trick in your back pocket even if it isn't perfect. 🙂

Whether it will also prevent the individual tabs refreshing from the most recent available cache when I next permit the browser (or the OS) to update, I  don't yet  now know: having updated the browser, I can safely say that yes, you do have to toggle things off again, they don't remain toggled off upon restarting Chrome.

A possible(?) alternative (or addition), might be to turn off Chrome's “memory saver” in chrome://settings/performance.  This memory saver feature seems to be fairly new though, since version 108 as of 29 Nov 2022, and the tabs' self-refreshing of volatile info from cache goes back well before that, so I don't know if turning this off would help.  As well, though it might be an option for those on a tower, this might not be so great for anyone on a laptop or other battery-oriented devices, since the memory saver is aimed at saving battery power by reducing memory usage.

 

The other thing, and this one's a big one, is another case of Chrome being a P.O.S..

How many times, if you're a computer user, have you accidentally closed the entire browser?  That might not be a problem if you could adjust a setting (as one can in every other browser on the planet)... but one can't [currently] do this in Chrome.

All that it takes to screw it up is clicking the Big Red X at the top right by accident... say, when you're using a wireless mouse (because thanks to the march of technology, you can't simply have it jacked into the port anymore, right?) and the battery is low, and the cursor suddenly jumps who knows where (or for that matter, if you're on a desktop tower and there's even a momentary power loss, and you're not running an uninterruptible power supply).

This might not matter if you're reading a fic, since all that you need to do is re-open your browser — but when you're doing things with tabs that read volatile data, that browser restart can ruin your day.

There are websites that take advantage of a Java loophole to prevent Chrome closing by accident.

When you try (intentionally or not) to close the browser and you have a tab open to one of these sites (and have indeed clicked any spot on that page in order to activate its userscript), the browser will immediately jump to that tab with an alert popup asking you if you want to leave the site.

Here's one of them: https://maki-chan.de/preventcloseads.htm
  NB:  having updated the browser, I immediately tried to close the browser without first clicking in that tab, and it turned out that yes, it closed, so you do need to click in that tab again upon restarting Chrome if you want it to function.  Upon re-reopening Chrome, I clicked around in that tab and then checked once more to see if I could “accidentally” close the browser, and the close-preventing tab worked fine this time, now that my clicking in it had reactivated the userscript.

If that page ever dies, try a search for something like {prevent chrome browser closing} or {prevent browser close tab}.

 

Since I track the total daily Hit uptick, I also do a monthly thing on each of my fics' individual data (Hits, Kudos, etc.).

In the monthly process, I don't check my series' Bookmark counts as such, but if you look at each of your fics' bookmarks counts on the Statistics page, and go into each of your fics directly, then you might notice that they're not quite the same — this is because some of them are private bookmarks.

Thanks to destinationtoast's digging, one can find out all sorts of things about one's fics' and series' [public] bookmarks easily — click that second link and take your pick of what you want.  For my purposes, I'm just curious about each day's latest [public] bookmarks made and checking to see about series, since the series' bookmarks don't show up on the stats page.  (It would be too much extra effort for me to bother tracking the monthly changes of each fic's private/total bookmark ratio, or to add each series's data to my tracking.)

For quick reference here, in case you're curious, the no-frills quick-check of all (whether recced / annotated or not) of the public bookmarks of your fics and series (or any other author's name) — as long as those fics aren't in any Unrevealed Collections — is simply a question of going to

https://archiveofourown.info/bookmarks/search?bookmark_search%5Bbookmarkable_query%5D=YOUR_AUTHOR_NAME_HERE&bookmark_search%5Bbookmarker%5D=-YOUR_AUTHOR_NAME_HERE&bookmark_search%5Bsort_column%5D=created_at&page=1

and replacing those two bits about “YOUR_AUTHOR_NAME_HERE” with whatever your AO3 author name is (or any other author's, for that matter).

It's also kind of nice to see some of the notes that some readers leave in them.  If you're curious about only the fics that bookmarkers have both recommended and annotated:

https://archiveofourown.info/bookmarks/search?bookmark_search%5Bbookmarkable_query%5D=YOUR_AUTHOR_NAME_HERE&bookmark_search%5Bbookmarker%5D=-YOUR_AUTHOR_NAME_HERErec%5D=1&bookmark_search%5Bsort_column%5D=created_at&bookmark_search%5Bwith_notes%5D=1

Likewise, if you're just wondering about your fics' rec-bookmarks (regardless of being annotated or not):

https://archiveofourown.info/bookmarks/search?bookmark_search%5Bbookmarkable_query%5D=YOUR_AUTHOR_NAME_HERE&bookmark_search%5Bbookmarker%5D=-YOUR_AUTHOR_NAME_HERErec%5D=1&bookmark_search%5Bsort_column%5D=created_at

...or your fics' annotated-bookmarks (regardless of being rec'ed or not):

https://archiveofourown.info/bookmarks/search?bookmark_search%5Bbookmarkable_query%5D=YOUR_AUTHOR_NAME_HERE&bookmark_search%5Bbookmarker%5D=-YOUR_AUTHOR_NAME_HERE&bookmark_search%5Bsort_column%5D=created_at&bookmark_search%5Bwith_notes%5D=1

NB, 05 Sep 2024:  One odd thing though.  While these basically work, I was checking for any interesting recent bookmarkers' notes on my fonts tutorial and ran across one that said “author put more effort into writing this than nasa during the space race” (NB: if you need to link a bookmark, then you can find its address by viewing the page source of the bookmark search results page that has that bookmark on it, searching that page source page for the title of the bookmarked work, scrolling upward 5-10 rows to find a row that says li id="bookmark_0987654321" class="bookmark blurb group..., and using that bookmark number in the usual AO3 URL https://archiveofourown.info/bookmarks/0987654321; you can link to comments even more easily by hovering over the original comment's “Thread” or its replies' “Parent Thread” button to get a URL along the lines of https://archiveofourown.info/comments/999999999, or just check your e-mail notification if you have that turned on).  I loved that line, and then went about my business.  A little later I got to thinking and wondered if there had been any other recent fun / glowing / hating / etc. bookmarkers' notes on my other works (since I hadn't browsed them for quite a few months), so I ran a general search and decided to CTRL-F in order to skip past the fonts tutorial (which is most of the bookmarks) to speed the scrolling (since I had already checked those) and didn't see that quote anywhere (maybe the bookmarker had changed it, right?).
  After a little fiddling, I found that it showed up only on the first- and third-named searches, not the second and fourth ones (and since more bookmarks will accrue over time: that annotation was from 13 Aug 2024, if you need to click to an earlier page in the results by the time that you look, if you wish to see the glitch(?) for yourself).
  That's bizarre, to say the least, since although the first (all bookmarks, regardless of rec.s and/or annotations) and third (rec.s, with or without annotation) make some sense to include that one, the second (for bookmarks with both rec. and annotation on them, not just either alone) and fourth (annotated bookmarks, with or without rec.s) certainly should have included it, so... if you do check your bookmarks, then you might want to consider checking them across more than just one of these.

 

And I almost forgot one relatively minor point.  My computer's alarm clock started malfunctioning (it popped up visually, but silently, which isn't much help if you've dozed off).  Since I was using it to remind me every half hour to check the numbers, I rather needed it to function, and it seems to have difficulty when you have 15-20 alarms for different reasons (I do have one or two things going on outside of this study, after all).

I couldn't get it to work properly, even after checking updates and hard-restarting and so forth (everything but the registry), so I simply grabbed a freeware alarm.  I tried a couple, and they were crippleware or otherwise insufficient for my needs, and finally settled on a free alarm clock with the unfortunate name of... Free Alarm Clock (I can't cast the first stone: consider the topic of this meta-work, and then look at the title again).  Obviously, Captain Obvious came to my rescue with that one.

Is it safe?  I dunno, though I haven't yet seen any problems crop up from it.  And disclaimer time: no, I don't have any affiliation with whoever makes it, I'm just mentioning it because it happens to work for me; I'm on a WIN 10 OS (i.e.: 64 bit, and it's a tower), and this clock claims to be G2G for WIN 11 all the way back through XP, but they don't seem to have a port for any other OS.

Why bother mentioning this at all?  Because it's a small thing, and easily overlooked.  You wouldn't believe how quickly 30 minutes can pass unnoticed, and how tedious this process can become.  You might want to set a few alarms if you intend to follow suit with your own fics.

 

---

 

 
Bibliography  ^{Contents ↑}
 

NB: some of these have more recent updates, such as the annual AO3 data, but I used these only at the start of my research.  The updates should be easy to find with a search, for those seeking such.

ocalhoun's famous analysis of FimFic traffic times (misascribed everywhere as being a study of AO3 traffic...)
  https://www.fimfiction.net/blog/494106/when-is-the-best-time-to-post-a-story-answered-with-science

AO3's data
  https://ao3org.tumblr.com/tagged/stats

Daily AO3 Traffic in 2021 (data shown in PST)
  https://docs.google.com/spreadsheets/d/1VY3kJsl7i0O7ZBT9PEL6yWKKmPrVdpR0cIHap3IpDSI/edit#gid=851498242
  https://www.tumblr.com/ao3org/673074218496344064/image-monthly-ao3-traffic-for-2020-and-2021-in
  https://www.tumblr.com/ao3org/673074152446590976/images-daily-ao3-traffic-in-millions-of-page

Daily AO3 Traffic in 2020 (data shown in PST)
  https://archiveofourown.info/admin_posts/15931
  https://archiveofourown.info/admin_posts/19828
  https://docs.google.com/spreadsheets/d/11f_7CJiUaL_DZ5D3iO9hDekGQ_6kS55LPkxJm1vwezc/edit#gid=1946141121
  https://twitter.com/AO3_Status/status/1480836146028556288

Weekly AO3 Traffic in 2020 (data shown in PST)
  https://docs.google.com/spreadsheets/d/1fqYPeS7hS157cKjiC9x9PAjldtBjq2hq-SbQOp1yPUU/edit#gid=167712558

AO3 Status tweet New Year 2020
  https://twitter.com/ao3_status/status/1212394532089847811

AO3's data on country percents 2Q15
  https://ao3org.tumblr.com/post/128492863943/weve-been-keeping-track-of-international-traffic/amp

Destination: Toast! (tons of analyses here)
  https://archiveofourown.info/users/destinationtoast/works
  https://destinationtoast.tumblr.com/post/114598910014/see-bookmarks-of-your-works-for-ao3-creators
  https://destinationtoast.tumblr.com/post/655164850133139456/how-i-tracked-down-a-peculiar-problem-in-a-fanfic
  [User] Scripts for scraping Archive of Our Own (AO3), Tumblr, Fanfiction.net (FFN), and Wattpad to gather fandom data.

Visualizing AO3 data
  https://visualizing-archive-data.tumblr.com/day/2021/06/23
  https://visualizing-archive-data.tumblr.com/post/624697539731570688/want-to-visualize-your-own-data-or-fandom-data

Do Longer Chapters Mean More Hits?
  https://ao3datafan.tumblr.com/post/178829197992/do-longer-chapters-mean-more-hits
  https://ao3datafan.tumblr.com/post/179058180837/im-not-sure-if-this-question-kind-of-counts-but-i

{Fandom Stats: Hits, Kudos, and the Kudos/Hits Ratio} and {Ship Stats in the Agents of S.H.I.E.L.D. fandom}
  https://sarahastro84.tumblr.com/post/137845085815/fandom-stats-hits-kudos-and-the-kudoshits
  https://sarahastro84.tumblr.com/post/159659352560/ship-stats-in-the-agents-of-shield-fandom

A Comparison of Fanfiction.net and AO3 Popularity, Engagement Levels, and Demographics
  https://razieltwelve.tumblr.com/post/97801888278/a-comparison-of-fanfictionnet-and-ao3-popularity

I only just now (22 Feb 2023) ran into Flamebyrd's bookmarklet approach to download a .csv of one's fics' stats; I'm thinking more of something that would update data with new info added (not overwritten, nor as a separate file every time), but this could be useful for some readers' purposes:
  AO3 Statistics CSV Bookmarklet

I don't speak Java, but this Greasy Fork userscript might be a starting point for simply auto-checking one's total Hits every 30 min:
  AO3: Tracking

I don't speak Python either, but this looks a little more relevant [than the Greasy Fork example] to automating the process of checking your fics' stats:
  A Python API for scraping AO3
  a slightly fresher version of this code on GitHub
Note that per alexwlchan.net: “It’s not actively maintained, but it does have a fix that isn’t in the code above – in particular, you need to pass an authenticity_token in the POST request, or the login fails. See users.py for details.”

How to Track AO3 Fanfiction Statistics Using Google Analytics
  https://web.archive.org/web/20210906162959/https://fanslashfic.com/2018/04/23/how-to-track-ao3-fanfiction-statistics-using-google-analytics/
  NB 1: cf. some alternative recommendations (due to privacy concerns) before proceeding with such (the blockquote below is an excerpt of the relevant material from a subreddit thread):

The 'how' is I run my own little analytics server (using Matomo) that gets data when a chapter is viewed, through an image file request (a 'tracking pixel'). I run my own server so that I can ensure no third parties have access to the data. (And to be clear, the data is not associated with AO3 user accounts or anything else for that matter, and it’s anonymized and the raw data discarded so the info I have in the end is ‘a person accessed these chapters at this time using this client'. I don’t need to or want to know anything beyond that.)

[...] it’d be a privacy-preserving option to the Google Analytics tutorial out there (don’t use GA, and I don’t mean this in a wink-wink-nudge-nudge way — don’t.)

[...]

It’s not the easiest way, but if you want to learn how to run a service on AWS, my suggestion is to look at RDS for your database and ECS or Fargate for running Matomo in a container (https://hub.docker.com/_/matomo).

An easier option would be to use Heroku, DigitalOcean, or other similar service that makes it easy to set up a DB and run the Matomo container.

Many webhosts also have Softaculous installed, and you can install Matomo through that without having to even touch DB setup.

The rest of the configuration (including anonymization, data retention) you can do through your Matomo admin interface, and then it’s just a matter of adding the tracker to your chapters.

There are alternatives to Matomo, but for my requirements of being able to run it myself and supporting tracking pixels (instead of cookies or scripting) it seemed like the best fit.

And yes, Google Analytics explicitly tracks users and combines their information with other activity. That’s a no-go for me. There are also hosted services that claim to preserve privacy, but running my own is clearer and cheaper.

  NB 2: if pursuing GA anyway, then you might need help interpreting the results.

Mining Fanfics on AO3 (data collection and analysis)
  https://medium.com/nerd-for-tech/mining-fanfics-on-ao3-part-1-data-collection-eac8b5d7a7fa
  https://medium.com/mlearning-ai/mining-fanfics-on-ao3-part-2-exploratory-data-analysis-7260cf35ee8a

NB, 27 Aug 2024:  For those interested in a broader scope of data, you might want to check out centreoftheselights' large scale data (chock full of things like age distribution, sexual orientation, fandoms, etc.), below.
  AO3 Census 2013, a 19 part study series of 10,005 users
  AO3 Demographics Survey 2024, a 5 part study series of 16,131 users

 

---

 

Also in this series:

• Part 1 — How to AO3
    (general info. for newb.s, but includes tips and information that years-long veterans often don't know)
• Part 2 — How to make and fix a series on AO3
    (plus unrelated tips re. Wayback, different spaces and dashes, TTS problems)
• Part 3 — Analyzing AO3 reader traffic flow
    (one's own fics' reader traffic)
• Part 4 — A year-long AO3 overall traffic analysis
    (when do people upload, read, comment, kudos, and bookmark?  Reader and new-upload traffic in general, across all of AO3⁠ ⁠)
• Part 5 — Fonts, and colors, and work skins, oh my!
    (change your letters and ink color, see others' work skin secrets, etc.)
• Part 6 — Chess puzzle extravaganza
    (256 randomly selected preconfigured chess puzzles, a new one with every refresh
    [secret message-reveal from BBEG, because of course];
    no JS: just HTML and some CSS)
• Part 7 — Targeting specific AO3 work sections (not site) with CSS effects
    (wanna put some color into your CSS-verboten summary?  Maybe drop an image below your series link, or style your comments section? 😉)
• Part 8 — Inside, Outside, Upside-down
    (a showpiece of fun that one can have with CSS — a challenge puzzle, rather than a tutorial;
    incl. CSS on a single paragraph within the work summary, among other things;
    try to navigate it, work out its little puzzles, and work out how it was all done;
    caution re. volume in ch. 1, and flashing lights on :hover in ch. 2)
• Part 9 — Green Rain font, from The Matrix
    (several ways to drop Green Rain into your background, simulating the digital rain effect)
• Part 10 — Building ConLangs, with a concrete example
    (going about constructing one's own invented language: words, grammar, writing, etc.)

NB:  If the work skin [for this piece or any other] was stripped by downloading, so that you can't see any red ink or yellow highlighting or any other special effect, then please see my fonts tutorial's sub-section on re-inserting rules.