More stats thoughts · 3:09am Aug 8th, 2014
It's probably a symptom of my profession that I tend to go nuts with datasets. Either that, or I ended up where I did because of the existing predilection. Whatever - there's a chicken and an egg, and as long as they're both accounted for it doesn't matter which one killed Colonel Mustard in the library.
So, it's with increasing frustration that I look at the per-story stats exposed by the site currently. In particular I've checked in on Bluey Napoleon a couple times in the last few days, because it's a good story that I want to see do well. The first thing I notice is that the per-day views are for all chapters and not unique story hits. You guys probably knew that, but I didn't because I haven't spent a lot of time in chaptered stories, particularly not while looking at stats. As a side note, I've seen somewhere (don't remember where!) the belief expressed that the actual per-chapter counts are part of the stats, but they aren't. The data sent puts in labels for dates where a chaptery event happened. That could be an edit or an addition. Maybe some other stuff? I don't know. Here's what it looks like for Bluey:
[
[ new Date( 1407132000 * 1000 ), 324, "Footnotes" ]
, [ new Date( 1407218400 * 1000 ), 1002, "" ]
, [ new Date( 1407304800 * 1000 ), 676, "" ]
, [ new Date( 1407391200 * 1000 ), 134, "" ]
]
All the chapter additions happened on the same day, so it goes with the last one here. That gives the unfortunate impression that there are views associated with the footnotes, which I'm hoping isn't true. Per-chapter views alongside per-story views over time would be fantastic, but that's not what we have. Lacking any way to separate these views into meaningful buckets, this is more information that's less informative. Forced to choose only one metric by which to judge a story's success or popularity over time I would take unique story views, which I guess is the one we can't have for any time except right now.
I'm not just whining about this because it makes my graph thingy less useful. I'm also whining about it because I see a good stats api as one way to let this site cater more easily to a wider variety of readers. What follows is less about in-depth story stats and more about what types of data are available at a glance, site-wide. The connecting theme is that some attention to stat reporting would be a good investment, in my opinion. Here comes the ramble:
Anyone who has spent a few minutes around here can tell you that the feature box is terrible. Bluey got a good start on the site's popularity metric, which we can fairly confidently assert is some combination of view:like ratio and likes per time unit since publishing (with a skew towards the latter). It appeared on the feature box for several hours. Those hours were a small subset of the hours enjoyed by Twilight turns into a red and black alicorn, which is so blandly written that it may as well be a bulletted list of its very few (unimaginitive) plot points. The promise (unfulfilled) of riffs on a recurring fandom trope and the hastily recolored cover art seems to be enough to keep that sucker featured for days. This is what happens when you hand content curation to the masses. On average, people probably didn't like that story as much as the raw stats suggest - just by virtue of the views funneled its way by its feature box placement and the hook of its title and cover art, it kept collecting likes, and it's not really offensive except in mediocrity, so it doesn't attract a lot of dislikes.
A story like Bluey, unambiguously superior, has a few things working against it. It's longer - it takes a while for people to get through it (some won't at all), and many will wait til the end before they mash the thumb, even if they knew they were going to a hundred words in. It's also more divisive, because it's not bland. Some poor souls will react poorly to the writing style, or they don't like Blueblood on principle, or they have even bigger issues with footnotes than I do.
This isn't news to anybody. There are lots of efforts around to provide alternates to the featurebox, mostly in the form of individual reviewers and review groups. That's great as far as it goes, but none of them are anywhere near as accessible as the featurebox. Those site blog posts don't even really put a blip on the radar most of the time. I'm not even convinced that they put an alert on the feed button. I just saw a Seattle's Angels thing in looking through the aforementioned feature box alternatives and went back a few issues to find that my own story was written up like a month ago. Wow! That's cool - I had no idea, which is the problem. I know all about red and black Twilight, because it was waiting for me on the top of the page for like three days running.
I haven't been around here long enough to regale you with mostly-fabricated memories of the good old days, when only works of the highest quality graced the box. I have been internetting long enough to have seen way too many Eternal Septembers. When a community reaches the point where the noise completely obscures the signal, it starts to die. We could use some tools to cut through the noise.
What I envision is a feature by which users can choose a featurebox that is meaningful to them specifically. For that, we need a wider variety of feeds and stats than we currently have. A box composed of the most favorited stories of people I follow would be fantastic. That's an expensive query, so I understand why that specific idea might be less than appealing to the site operators, although there are workarounds. Much less expensive would be feeds from specific people or groups that you can mix and match. Maybe I want to see what PresentPerfect highly recommends, along with a sampling of recent additions to a curation group like 'Twilight's Library'. Someone else doubtless wants tentacle and oviposition clop delivered to them with even greater efficiency than the feature box already manages.
These are actually things I can already do (even the aggregate-favorite of followed people thing) by bludgeoning the server with lots and lots of automated requests. Leaving aside the fact that that's not really a nice thing to do, there are much more elegant solutions available at the server level. I'm not saying that the site maintainers owe us anything like this. I am saying that, from where I sit, it would be good for the community.
Thank you for the good words (and the tracking) of Brewmare! Yeah, I think it's just hitting too niche a market, or I scotched the sales pitch/description, or people just don't want to read about Blueblood, or something.
An Equestria Daily feature hit as it was falling out of the Featurebox (about 24h in actually; it did better if you turn Mature off), and that influx of traffic extended its foothold, but even so it never did better than about 5th. I'm pretty sure a non-trivial part of its traffic has come from word of mouth — you're its third blog-post signal boost. So it hit a collective "meh" and a super-strong reaction from a small core. I'm actually more surprised over its relatively outsized downvote collection; but I attribute that to sexual themes in the description and no actual porn to draw people into the story, which gets the wrath of the moral brigade without actually pulling many more eyeballs in.
I don't think there's any story-unique-views-over-time stat, but I do take snapshots of that for my stories every month or two, to track the long tails.
As for feeds, joins groups! Then you get notified of story adds. Present Perfect adds stories to "I'm Gonna Rec It!" every time something scores an HR, so even if you don't read his reviews you can get notified when he loves something. Twilight's Library _is_ a group, and Seattle's Angels and the RCL both drop their featured stories in their group folders as they post. Mix and match those on your site feed and you can get pretty darn close to a curated version of the site.
What Horizon said about the groups. If you're in a proper set for your interests, lots of interesting stories go by as feed page notifications.
I AGREE WHOLEHEARTEDLY AND WE SHOULD STOP INVESTIGATING HIS DEMISE AT ONCE
2353205
2353337
I'm aware of groups, honest. The problem is again one of accessibility. We can tell at a glance that the feature box drives more traffic than any group or blog post or whatever else. Even EQD is powerless before it, apparently. This has a warping effect on the entire site. People write for the feature box, instead of for quality. That's not a problem if you're already aware of its failings, but it's kind of a shame that the most prominent billboard on the site is mostly useless, both because of its own failings and because of the behavior it encourages.
Like I said, I already know where to find all the information I'm looking for - the problem is how it's exposed. The feed is a good example of potentially useful information that's presented poorly. If I want to join curation groups and get notifications of story additions, that's great, and it's great that the feature is available. But those notifications are going to be mixed in with all the other stuff on the feed, on a page that's not actually designed to present stories. The format of the featurebox is actually spot on: the top X stories by a given metric, displayed compactly, easily navigable and with summary information available at a click (with no extra page load). It's just the content that's lacking. I feel like a way to change the metric (and a way to invent new metrics we can easily switch between) would go a very long way to addressing this.
2353492
Ah, that makes sense.
And speaking of Brewmare's stats, here's an interesting thing: the chapter views. Bad Horse has looked into the proportion of readers who will drop away between the first and second chapters; usually you get a fairly precipitous drop between chapters 1 and 2, and smaller drops between succeeding chapters as you go on, and the final chapter represents the number of readers who have actually finished the work. Brewmare is a special case, because there's a suggestion up front to open the third chapter concurrently for easy footnote access (and because people who click on the links between chapters are driving up the view count as they go back and forth). This is the only story I've published where Chapter 1 is not the one which sets the overall view count (Chapter 3 actually has more).
2355378
Oh man, that's actually where I saw the thing about someone thinking you can get that info from the data, now that you remind me. But it's not actually there - you can make educated guesses, and say that most of the views after a chapter addition (and before the next one) are for that chapter, but it's just a guess. If someone happens along, sees your second chapter and decides to read the first too, it all gets counted the same way. We only have a timeline, no chapter association. And even the timeline is of limited use, particularly on chaptered stories - with only 60 days of view data, if your release schedule is not fairly quick you're not really going to have a good idea of how the attention around chapter three compares to that for chapter one, because that data is lost in the mists of time by then.
I'm actually super tempted to write a little utility site that snapshots data for stories it's told to so that we can reconstruct a lot of the missing bits (unique story views in particular). This is the kind of project that seems appealing when the whole writing thing isn't working out, I guess; can't do what I do for fun, better pretend I'm at work.