Home « Forum « Author Hangout

Forum: Author Hangout

New view of story stats

Crumbly Writer

For the first time (that I recall), I just sorted my stories by total download counts, and noticed something odd. My most downloaded stories rarely have the highest scores, but they retain those high scores in spite of the low scores.

I can understand lower downloaded stories having higher scores, since they aren't read by as much of the general public, but I find it striking that the top stories aren't all highly rated. (Mine range from 7.7 to 8.1, so it's not much of a variance.)

My numbers on FS and Sci-Fi are much lower, but I found the same thing (even though the top five decrease from the highest to the lowest, although the three lowest are also the highest scores of them all).

I guess it shows that, if a story is decent, it'll drive readers, even if they intensely dislike certain aspects of the story (often as they're designed to do).

Ross at Play

I think you're looking at the wrong figures.
I think the closest approximation you can get to the most-read stories is total downloads divided by number of chapters, as opposed to the raw figure for total downloads.

Replies:   awnlee jawking
awnlee jawking

@Ross at Play

As the author, CW will have access to the number of readers who read the last chapters ie likely to have finished the stories. That's probably even more accurate, as the stories may have different rates of abandonment.

AJ

robberhands

I'd guess, for a multi-chapter story, the lowest download-count of an individual chapter represents the most realistic number of readers.

Ross at Play

@awnlee jawking

the number of readers who read the last chapters [is] probably even more accurate

Agreed. If only I'd written a multi-chapter ...

robberhands

The last chapter download-count is a bit bloated by people who didn't read the complete story but voted on it. That's the reason mostly the second last chapter has the lowest download-count.

Replies:   awnlee jawking  REP
Switch Blayde

@awnlee jawking

the number of readers who read the last chapters ie likely to have finished the stories.


Someone might give up on a story but jump to the last chapter to vote.

awnlee jawking

@robberhands

The last chapter download-count is a bit bloated by people who didn't read the complete story but voted on it. That's the reason mostly the second last chapter has the lowest download-count.


I'm not exactly a prolific author so my data pool is very small, but for all of my multi-part stories the final instalment has the lowest download count.

AJ

REP

@robberhands

Don't forget that if you decide to abandon a story after reading a few chapters, you have to go to the last posted chapter to remove it from your library.

Replies:   robberhands
robberhands

@awnlee jawking

I'm not exactly a prolific author so my data pool is very small, but for all of my multi-part stories the final instalment has the lowest download count.

I'm talking about long stories (CW's stories for example), which can't be read in a single sitting. How many readers will abandon such a long story at the second-last chapter? How many people will abandon it but still vote on it?

Replies:   Crumbly Writer
robberhands
Updated:

@REP


Don't forget that if you decide to abandon a story after reading a few chapters, you have to go to the last posted chapter to remove it from your library.


Did it look like I forgot that?

ETA: My bad, I misunderstood your statement. I don't load a chapter to remove a story from my library. I use the library feature directly. Do you mean, the last chapter gets loaded nevertheless?

Replies:   REP
Crumbly Writer

@awnlee jawking

I'm not exactly a prolific author so my data pool is very small, but for all of my multi-part stories the final instalment has the lowest download count.

That's natural, as sequels rarely do better than the original (i.e. you rarely draw in new readers), though each new book draws readers to your other books (i.e. a new sequel will cause more readers to read your first book first, and then the sequel only if they like your first book).

With each new book I sell, I typically sell more older books than I do new ones.

Replies:   awnlee jawking
Crumbly Writer

@robberhands

I'm talking about long stories (CW's stories for example), which can't be read in a single sitting. How many readers will abandon such a long story at the second-last chapter? How many people will abandon it but still vote on it?

Few abandon my stories on the first (or second) chapters, though some do when they discover the subject matter doesn't suit them. Most will wait until the fifth chapter, while those more dedicated will stick around to the 8th or 10th before quitting, even IF it doesn't suit them.

Generally, those who DO quit on the first chapter are more likely to downgrade the story by voting it down from the get-go, since the subject matter doesn't appeal to them. Those who read more chapters before quitting are generally more forgiving (i.e. while the story might not go where they wanted, they see the general appeal of the story) and thus aren't as likely to 'vote it down').

Replies:   robberhands
robberhands

@Crumbly Writer

I've no idea how you can make such distinct observations. Regarding 'Few abandon my stories on the first (or second) chapters,...', this is from SoL's FAQ and it's certainly true for my stories' download counts:

... So whenever a reader clicks on the link for a serial story, the site usually sends the index page, and the index page contains a "next" link to the first file in the story, be it a foreword, chapter 1 or prologue. And thus the browser that prefetch pages will hit those first files right away, whether the reader click on that first part or jumps to the last part of the story. So the first parts of every multi-part stories will ALWAYS have abnormally high download counts. That high number doesn't mean anything.

It's impossible for me to know why or when someone stops reading my stories.

Crumbly Writer

@robberhands

It's impossible for me to know why or when someone stops reading my stories.

As I've mentioned previously in other discussions, I use a dirty trick. I typically post my cast list first, frequently updating it, only to delete it when the story ends before reposting it. This ERASES (nearly) all of those false-download counts, giving me a truer reading of how many abandon the story on which chapter, as there are fewer false 'read-ahead' chapters.

Replies:   Not_a_ID
REP

@robberhands

When I'm reading a multi-chapter story and decide to abandon it, I use the Story Index to access the last posted chapter and remove the story from my Library.

Replies:   robberhands
robberhands

@REP

I don't. I use my library reading queue. I remove a story and get asked whether I also want to remove it from the library.

Replies:   REP
awnlee jawking

@Crumbly Writer

By multi-part story I meant a single story posted in multiple instalments.

I'm not big on sequels :(

AJ

Replies:   REP  Crumbly Writer
REP

@robberhands

I've no idea how you can make such distinct observations.


CW and I have disagreed in the past about his identification of trends bases on analyzing stats.

One would expect reader attrition to result in a constantly declining number of downloads per chapter. However for one of my stories, one of the later chapters had a download count that was significantly higher than the 3 preceding chapters. I'm not into story stats, so I just noted the oddity and put it down to people accessing that chapter numerous times for whatever reason.

Before continuing, I will say that analyzing story stat data for personal use is fine by me. I only start spouting off when the data collection system seems flawed or the analysis techniques seem questionable, and then the person doing the analysis starts telling me that their analysis proves a trend.

Assume you ignore the download count of a story's early chapters, download count is still not an accurate representation of number of the number of readers. Individual readers may download a chapter 2, 3, 4, or more times before they finish the chapter. After they've finished the chapter, they may go back through multiple chapters looking for a passage they want to reread.

For the above and other reasons, I am highly skeptical of conclusions made from analyzing data acquired from SOL's story stats. My personal opinion is, if the person doing the analysis didn't collect or at least control the collection of the data being analyzed or if I am not familiar with the capabilities of the person doing the analysis, I question the results.

REP

@robberhands

Just two different ways to access the story deletion routine. I get the same questions.

REP

@awnlee jawking


I'm not big on sequels :(


Last time I checked, the second and subsequent installments were sequels.

the next installment (as of a speech or story); especially : a literary, cinematic, or televised work continuing the course of a story begun in a preceding one

https://www.merriam-webster.com/dictionary/sequel

Replies:   awnlee jawking
robberhands

@REP

For the above and other reasons, I am highly skeptical of conclusions made from analyzing data acquired from SOL's story stats.

Agreed.

Crumbly Writer
Updated:

@awnlee jawking

By multi-part story I meant a single story posted in multiple installments.

Understood. Multiple sequences is referenced as "books" to distinguish it from "multiple posts".

Crumbly Writer

@REP

However for one of my stories, one of the later chapters had a download count that was significantly higher than the 3 preceding chapters. I'm not into story stats, so I just noted the oddity and put it down to people accessing that chapter numerous times for whatever reason.

Knowing stats, that's unlikely to mean those readers skipped those chapters entirely, instead it indicates they read the three chapters at the same time, so it only registered the first and last chapters read that day (the last probably when they voted on it).

As for "proves a trend", the reason why I provide these here to get feedback, so determine whether anyone else's data backs up my suppositions, or whether they're just the responses by MY readers, or whether I'm reading too much into too little objective detail. I'm certainly NOT saying this is a definite trend, though my phrasing often sounds like that's my position. (Sorry, that's how I tend to talk. Everything is always definitive, rather than questioning.)

Finally, "trends" doesn't mean "proof-positive", it simply means "there might be a likely cause, barring future evidence which disproves the supposition". Just like when I read individual chapter scores to see how readers like particular chapters (say violence vs. romance), I like these "micro-analyses" because the identify reader preferences, and highlight how readers often see my stories in an entirely different way than I and my editors do.

If I'm aware of something, I can either modify it, or acknowledge it, buying me time to make amends in the story as I further flesh out the characters.

Ernest Bywater
Updated:

I've no intention of getting into the analysis of what the counts mean, or the full details of why they get viewed, but since this thread has moved into chapter count variations I think it may help to have an explanation of how the counters work.

The SoL system is aware of when a registered member visits any story page within the system, and it adds a count of one for every visit by a different registered member within the following limits.

1. There is a 24 hours 'day' which is reset to zero each day.

2. The reader history count is increased by 1 for every story accessed in that day.

3. The story download count is increased by 1 for every reader who access the story - but it counts 1 for every access on each day the reader accesses the story.

4. The chapter download count increases by 1 for every reader who accesses that chapter - but it counts 1 for every access on each day the reader accesses that chapter.

With 2, 3, and 4 if you access the same story or chapter multiple times within the SoL day it will only count 1 for the day. However, access the same story or chapter on multiple days it will count a 1 for each day you access the story or chapter.

What this means is if you read multiple chapters on the same day the story count goes up only 1 while each chapter goes up 1 regardless of how many times you access them that day. But if you re-read a chapter two days later the story count and that chapter count will both go up by 1.

.............

In regards to multi-chapter stories there are many reasons why a chapter may be accessed by the same reader a number of times while reading a story. However, the chapter count will only go up by 1 for each different day they access that chapter.

a. A reader goes back to a chapter to check something in it.

b. It's been a while since the last update a reader will go back to a chapter to re-read that bit before continuing.

c. Depending on how they have the story bookmarked when they go to read the new chapter posted it may take them back to the previous chapter first, and thus show a count for both chapters on that day. NB: This only applies where a bookmark link is used to get to the story.

...................

Sometimes false counts occur, but they are so small in number it isn't funny. By false I mean a reader accesses a chapter / story without reading it. The most common that happens to me is when I have a story left open in my browser on my computer while I do something else. The count won't go up until after I have the browser load the next page or reload the current page. However, if I leave the page up overnight and for some reason I close the browser when I re-open the browser it will reload the page, even if I don't start to read the story the fact it reloaded will see the system add 1 to the chapter and story count within the limits set in 3 and 4 above.

All of these factors lead to varied chapter counts.

In my own stories I notice the multi-chapter stories where I have a site or building plan in the story the chapter with the plan has the highest count. I suspect, but can't prove, it's because readers will often refer back to it when reading later chapters. This is especially true for the longer stories with a plan in them.

typo edit

Replies:   REP
REP

@Ernest Bywater

It is always nice to know how things like that actually work, EB.

So if I start reading a long chapter and am a slow reader or if I'm interrupted and it takes me 6 days to read the chapter with me accessing the chapter once a day, I will add 6 to that chapters down load count to just read the chapter. If on 4 subsequent occasions I refer back to that chapter on different days, I will have contributed 10 to its download count.

If others did similar things, that means the download count could be considerably higher than the number of readers that accessed the chapter.

Ernest Bywater

@REP

If others did similar things, that means the download count could be considerably higher than the number of readers that accessed the chapter.


correct, also, since you read the chapter on different days the story download count will increase by 1 on each day you access the chapter because it counts that as an access to the story.

Thus if you read a 20 chapter story all on the same day the story and each chapter get a count of one each. However, if you take four days to read the same story the chapters will each get a count of 1 while the story will get a count of 4 due to 1 for each day of access.

This is why the total count and the chapter counts will never match on the longer stories, it's rare for someone to read it all on the same day.

There's a lot of little quirks about the system I've learned from Lazeez over the years.

Replies:   robberhands  Not_a_ID  REP
robberhands
Updated:

@Ernest Bywater

I joined this discussion by stating my assumption that the lowest individual chapter download-count represents the most realistic number regarding readers completely reading a story. The correct number is impossible to deduce from those numbers. It's most probably lower but can't be higher. That's all I'm certain.

Ernest Bywater

@robberhands

I joined this discussion by stating my assumption that the lowest individual chapter download-count represents the most realistic number regarding readers completely reading a story. The correct number is impossible to deduce from those numbers. It's most probably lower but can't be higher. That's all I'm certain.


I agree with you, but, as you say, you can't be certain about any of it. Although it is a good indication.

awnlee jawking

@REP

I consider the next instalment of a serial to also be an instalment, whether it's a single chapter, multiple chapters, parts, sections, subsections or whatever delimitation the author chooses to use.

AJ

awnlee jawking

@REP

However for one of my stories, one of the later chapters had a download count that was significantly higher than the 3 preceding chapters.


That happened with one of my stories. However the following chapter was delayed, so I ascribed the extra readership to readers who wanted to re-read the chapter so recent story events were fresh in their memories when reading the new chapter.

AJ

Not_a_ID

@Crumbly Writer

As I've mentioned previously in other discussions, I use a dirty trick. I typically post my cast list first, frequently updating it, only to delete it when the story ends before reposting it. This ERASES (nearly) all of those false-download counts, giving me a truer reading of how many abandon the story on which chapter, as there are fewer false 'read-ahead' chapters.


Except pre-fetch and premier members can even mess with that, as SoL gives them the option to bookmark and return to last chapter read(or otherwise bookmarked).

So if they stop at chapter 10 for multiple updates(on the updated serials list) for example, they have potential to download multiple copies of that chapter.

Now Laz does have measures in place to prevent someone from being counted multiple times, IIRC it only works on a daily basis. So multiple successive story updates would presumably bypass that as it presumably plays out over days/weeks/months of even years in some cases. It also is device(browser)/IP specific, rather than account based. So there certainly are edge cases where even within that "same day" restriction, a single reader can end up being counted multiple times as they switch between devices/locations over the course of a day.

awnlee jawking

@robberhands

If you're on a story's index page and you click the link for chapter 27, do those browsers prefetch chapters 26 and 28?

AJ

Replies:   robberhands  Not_a_ID
robberhands

@awnlee jawking

I doubt it but I've no idea about the technical specifics.

Not_a_ID
Updated:

@Ernest Bywater


Thus if you read a 20 chapter story all on the same day the story and each chapter get a count of one each. However, if you take four days to read the same story the chapters will each get a count of 1 while the story will get a count of 4 due to 1 for each day of access.


Which isn't to mention the grey area of how SOL accounts for full story downloads from the paid membership. They may potentially never actually load even the index page, as they just pull it from the new/updated stories listings.

Not_a_ID
Updated:

@awnlee jawking


If you're on a story's index page and you click the link for chapter 27, do those browsers prefetch chapters 26 and 28?


Browser pre-fetch behavior varies from browser to broswer, and potentially from other meta-data provided with the page itself. But the "simple implementation" of pre-fetch for browsers is they would go about caching the contents of every link into the browsers memory before you have clicked on anything. At least, if your internet connection is fast enough. Presumably, it will wait for the first page to complete loading first.

Replies:   Ernest Bywater
Ernest Bywater

@Not_a_ID

pre-fetch for browsers is they would go about caching the contents of every link into the browsers memory before you have clicked on anything


Lazeez would be best to answer this, but I thought they came up with a way to allow the web page to tell the browser to not pre-fecth and SoL has that. I do know there was a change that has my browser pre-fetching less than it used to when the pre-fetch first started, but can't remember why that is.

Replies:   Not_a_ID
Not_a_ID

@Ernest Bywater

I thought they came up with a way to allow the web page to tell the browser to not pre-fecth


Which would be the comment about meta-data telling the browser that a given page isn't a good candidate for pre-fetch or caching. Still doesn't mean every browser is going to honor said meta-data or instructions.

REP

@Ernest Bywater

Something else occurred to me last night after I shutdown my computer. When I access a story, my computer opens the last story I read and I then access the chapter I want to read. From what I see, there is a pointer that is set to the last chapter I read and the chapter is always downloaded the next time I access the story. That means I increment 2 download counters to read 1 chapter. I put together the following scenario to see if that might lead to the chapter download counts leading to incorrect conclusions.

To make it easy, let's ignore the false download count problem with the first chapter. Assume an author writes a story with 10 chapters and the chapters are posted at 3-day intervals. 10 people read the story.

All readers read Chapter 1 on the day it is posted. The chapter's counter shows 10 downloads and all of the readers' pointers point to Chapter 1.

When Chapter 2 is posted, all or the readers read Chapter 2 on the day it is posted. When the story is opened, all of the readers uploand Chapter 1 and that increments its counter to 20. The readers select and read Chapter 2. Its counter shows 10 downloads. All of the readers' pointers now point to Chapter 2.

The same thing happens when Chapters 3, 4, and 5 are posted. The counters for Chapters 1-4 show 20 downloads and Chapter 5 shows 10.

Now assume 5 readers decide to wait until the story is completely posted before reading further.

When Chapter 6 is posted, the remaining 5 readers read the story. Chapters 1-4 counters indicate 20 downloads, Chapter 5 shows 15, and Chapter 6 shows 5. The 5 readers' pointers point to Chapter 6.

When Chapters 7-9 are posted, the same process occurs. Prior to Chapter 10 posting, the download counters show Chapters 1-4 at 20, Chapter 5 at 15, Chapters 6-8 at 10, and Chapter 9 at 10.

When Chapter 10 posts, 5 readers read the last chapter and the remaining 5 readers read chapters 6-10 in one day. After Chapter 10 posts, the download counters show Chapters 1-5 at 20, Chapters 6-9 at 15, and Chapter 10 at 10.

If the download count correlates to the number of readers that opened and read the chapter, the values imply that there were 20 readers. 5 readers dropped the story after opening Chapter 5 and a second group of 5 readers dropped the story after opening Chapter 9.

Ernest Bywater

@REP

Something else occurred to me last night after I shutdown my computer. When I access a story, my computer opens the last story I read and I then access the chapter I want to read. From what I see, there is a pointer that is set to the last chapter I read and the chapter is always downloaded the next time I access the story. That means I increment 2 download counters to read 1 chapter. I put together the following scenario to see if that might lead to the chapter download counts leading to incorrect conclusions.


I'll deal with this point first.

It will only count twice if the clock at the SoL server has clicked over the 24 hours into the next SoL day - if you do that and get back to it before the timer at SoL clicks over it won't increase because the system will see it as 2 accesses on the same day. - this where the different time zones have an effect.

Replies:   REP
Ernest Bywater

@REP

All readers read Chapter 1 on the day it is posted. The chapter's counter shows 10 downloads and all of the readers' pointers point to Chapter 1.

When Chapter 2 is posted, all or the readers read Chapter 2 on the day it is posted. When the story is opened, all of the readers uploand Chapter 1 and that increments its counter to 20. The readers select and read Chapter 2. Its counter shows 10 downloads. All of the readers' pointers now point to Chapter 2.


This will depend on their personal setting and how they access the story.

When you go through your Library bookmarks it usually takes you to the last chapter you were reading, but if you go to a story update from the 'Updated Story' list it usually takes you to the next chapter to read or the last chapter posted - - I'm not sure which is the default, because I've had both happen to me.

.............

As I've often said, the stats you get are only indicative, there's no way you can say they're absolute figures. And the points you raise are valid, which is why I have a huge packet of salt on hand when people start talking about what the stats mean.

I know readers who will read the first chapter when it's posted, and from that they'll decide if they'll read the full story when the last chapter is posted or if they'll totally ignore the story. They also refuse to even look at the first chapter if they know the author is the type who adds codes as the story goes up, because they want to know all the codes before they start.

If ten people really like a story and re-read it every month the system will show lots of reader access and the counts go up, but they may be the only ten people who ever read the story. There's no way an author can know.

Replies:   docholladay  REP
docholladay

@Ernest Bywater

When you go through your Library bookmarks it usually takes you to the last chapter you were reading, but if you go to a story update from the 'Updated Story' list it usually takes you to the next chapter to read or the last chapter posted - - I'm not sure which is the default, because I've had both happen to me.


What happens if you are a premium user and the story is bookmarked. Clicking the story from the updated story list will take you to the last chapter read.

If you are a free user all multi-chapter stories open the story index page. For premium users the result will be the same for any story which has not been bookmarked.

To be honest I actually would prefer the index as a default.

Crumbly Writer

@REP

It is always nice to know how things like that actually work, EB.

So if I start reading a long chapter and am a slow reader or if I'm interrupted and it takes me 6 days to read the chapter with me accessing the chapter once a day, I will add 6 to that chapters down load count to just read the chapter. If on 4 subsequent occasions I refer back to that chapter on different days, I will have contributed 10 to its download count.

The other thing I do, is that rather than bookmarking specific pages, I'll simply leave the page open in my browser. If I do that with 20 or 50 different pages, every time I open my browser (say one each day), it'll update each of those chapter counts by one each time.

I'm unsure whether the browsers actually register the page unless i click on it (Firefox addressed this some time ago), so the problem isn't as bad as it once was, but it's still something to be aware of.

It's better to bookmark than to leave pages open indefinitely.

Crumbly Writer

@robberhands

I joined this discussion by stating my assumption that the lowest individual chapter download-count represents the most realistic number regarding readers completely reading a story. The correct number is impossible to deduce from those numbers. It's most probably lower but can't be higher. That's all I'm certain.

The first chapter (the diagram in Ernest's case, or the character list in mine) is ALWAYS higher, not because it's read that often, but because, when readers access the chapter index to jump to the latest chapter, it 'reads ahead', counting the chapter even though you've never opened it. That's the biggest single 'contamination' in the stats, unread pages counting as actual downloads. Eliminating those 'false' counts provides more realistic, but still not perfect' chapter counts.

REP

@Ernest Bywater

It will only count twice if the clock at the SoL server has clicked over the 24 hours


As stated the posting interval is every 3 days, so the clock will have ticked over.

REP

@Ernest Bywater

This will depend on their personal setting and how they access the story.


Yes that is true. The different variables in how readers access a story and what chapters they download once the story is accessed adds at least one layer of complexity to the generation of the chapter download counts.

However, my scenario was intentionally simplistic and it limited the way in which the 10 readers downloaded chapters. It was based on all readers having the same setting for opening the story; namely, open the last chapter read.

There's no way an author can know.


I agree wholeheartedly and that is the reason I believe an author's conclusions based on information that contains factors the author doesn't know should not be trusted. I also have a big bag of salt.

I created the scenario to demonstrate that analyzing a dataset (e.g. chapter download counts) can lead to false conclusions. Anyone who wants to use a dataset in an analysis needs to understand how the data points were generated and how the data collection system gathered the data points.

What my very limited and controlled scenario demonstrated was 10 readers read the entire story generating chapter download counts that differed from what one might expect to have been generated. If someone analyzed the resulting chapter download counts without knowing there were only 10 readers, the conclusion may have been a few readers had dropped out of the story (i.e., the total number of readers was greater than the 10 that finished the story).

Back to Top