Reader analytics should tell an author which channels bring readers and how far those readers get, and it should do that without building a profile of any individual reader. The privacy-first version of that is counter-based: instead of logging a row for every visitor with a session identifier, an IP address, and a full referring URL, you increment a daily tally per story per channel and keep nothing that identifies the person. You give up unique-visitor counts and multi-step funnels, and in exchange your readers are not tracked and your author-facing numbers stay honest and legible, which for a reading site is the better half of the trade.
What should reader analytics actually tell an author?
It should answer three questions and stop: which channels are bringing readers, whether that is changing, and how far readers get into the story once they arrive. Everything a web fiction author decides in a given month comes down to some version of those, since the real choices are where to spend limited promotion time, whether the opening chapters are holding people, and whether the thing you tried last month did anything. I think almost every other metric that analytics vendors sell is downstream of an advertising business rather than a writing one, and an author who is not buying ads is being offered resolution they have no use for at a cost their readers pay.
The reason this matters is that analytics precision is not free, and the currency is reader privacy. To count unique visitors rather than visits, you have to be able to recognize the same person across sessions, which means storing an identifier of some kind and keeping it around. To build a funnel, you have to link a specific person's arrival to their later behavior. Each increment of precision requires retaining a little more about a human being who came to read a story, and the industry default is to retain it all and worry later. My feeling is that a reading site has a stronger obligation here than most software, because reading is close to thinking, and what somebody reads at two in the morning is nobody's business but theirs.
What counter-based analytics actually stores
A counter-based system stores a tally, not an event. On IlorisNovel, when a reader opens a story, the server classifies the referring source into one of a fixed list of channel buckets and adds one to the count for that story, that channel, and that calendar day. That row is the entire record. There is no session identifier, no IP address, no full referring URL, no user agent string, no reading duration, and no identity attached to the source at all. The full URL a reader came from is examined in memory long enough to work out whether it was Reddit or a search engine or another page on the site, and then it is discarded rather than written down, which matters because the paths and query strings of external pages routinely carry things nobody should be archiving.
The design has a pleasant side effect, which is that it makes the privacy promise structural rather than a policy. A promise not to look at data you have stored is only as good as the next person who has database access; a system that never wrote the data down cannot be induced to hand it over, and it cannot leak it in a breach. The channel list is deliberately short and closed, so an arbitrary tracking value cannot be smuggled into the store through a crafted link, and the day granularity means even the timing information is coarse. I would rather explain a limitation to an author than explain a data retention practice to a reader, and that preference is essentially the whole design.
Being precise about the boundary matters, so here is the one place it does not apply: an author can see a list of recent signed-in readers with their display name and the furthest chapter they reached, which exists so authors can recognize the regulars, and which only ever includes people who chose to make an account and read while logged in. That view is deliberately separate from the acquisition panel, and no reader's identity is ever attached to a traffic source. Anonymous readers are counted and never named. If you want the full mechanics of the source side, I walked through them in how to tell where your web fiction readers come from.
What you give up, stated plainly
You give up unique visitors, and that is the loss authors feel first. Because nothing links one visit to another, a reader who arrives from Reddit on three separate days counts three times, so the panel tells you how many reading sessions a channel produced rather than how many distinct humans it sent. For a channel that produces a small number of very loyal readers, that will look better than it is, and knowing the bias is the correction. You also give up funnels, meaning you cannot follow one reader from a social post through six chapters to a purchase, and you give up page-level referrer detail, since only the channel survives and the specific page does not. Anyone who wants post-level attribution has to build it manually by tagging their own links, which I covered in why your reader analytics say direct link.
What I do not think you give up is the ability to make good decisions. Compare a thirty day window against the previous one, look for a change big enough that you would have noticed it without a chart, and change one thing. That workflow does not need unique visitors, and honestly, most authors who have unique-visitor numbers use them as a vanity metric rather than as an input to anything. The failure mode of rich analytics in a small creative business is not that the numbers are wrong, it is that they are absorbing attention that should be going into chapters, and a tool that answers three questions in fifteen seconds respects that better than a dashboard with forty widgets.
Why the ad model and the tracking model are the same decision
Reader tracking on free reading sites exists mostly to serve advertising, and once you see that, the whole landscape becomes legible. Targeted advertising requires knowing something about the person being targeted, which requires collecting and retaining that something, which requires third-party scripts on the reading page, which is why readers of the big free serial sites have complained for years about intrusive ads and trackers on the pages where they are trying to read a chapter. The tracking is not a regrettable side effect of the ad model; it is the ad model's operating requirement, and no amount of tasteful implementation changes that.
This is why we chose not to run reader-facing ads, and why the analytics look the way they do: having removed the only real commercial reason to profile readers, there was nothing left to trade privacy for. The money instead comes from readers choosing to unlock chapters, with the author keeping 93% of what they earn and cashing out directly, which is a model that works better the more a reader trusts the site rather than the more thoroughly it is measured. I have written about the broader economics of that choice in the platform tax on web fiction, and the short version is that a platform's revenue model determines almost everything else about how it treats the people using it. Running one, the thing that has surprised me most is how many product decisions turn out to be that decision wearing a different hat.
There is a version of this argument that overclaims, and I want to avoid it. Privacy-first analytics is not a moral achievement, it is an engineering trade with a real cost, and an author who genuinely needs conversion-level attribution because they are buying ads is right that these numbers will not give it to them. What I would argue is that the trade is correct for the overwhelming majority of web fiction authors, who are not buying ads, who need three answers rather than forty, and whose readers came to read a story rather than to be measured. If you want to see how the writing side feels before worrying about any of the measurement, you can open the editor with no account and start typing.
Common questions about privacy-first reader analytics
What is privacy-first reader analytics?
It is analytics built on aggregate counters rather than on records of individual visitors. Instead of storing an event per person with a session identifier, an IP address, and the exact page they came from, the system increments a daily tally for a story and a channel and keeps nothing that could identify the reader. The author still gets the shape of their traffic, and no profile of any reader is ever assembled, which is the difference between measuring a story and surveilling an audience.
What do authors lose with counter-based analytics?
Mainly unique visitors and funnels. Because nothing links one visit to another, you cannot say how many distinct people a channel sent, only how many reading sessions it produced, and you cannot follow a single reader from a social post through a chapter to a purchase. You also lose page-level referrer detail, since only the channel is kept and the full referring URL is discarded. What survives is the comparison that actually drives decisions, which is which channels are growing and which are not.
Do reading platforms track their readers?
Many do, and reader-facing ad tech is the usual reason, since serving targeted advertising requires knowing something about the person being served. Web fiction readers have complained for years about intrusive ads and trackers on the larger free reading sites, and the tracking is not incidental to the ad model but a requirement of it. A platform that does not run reader-facing ads has no comparable reason to build reader profiles, which is why the ad decision and the privacy decision are really one decision.
Can authors see which individual readers are reading their story?
On IlorisNovel an author can see a list of recent signed-in readers with their display name and the furthest chapter they have reached, which exists so authors can recognize the regulars who show up every week. That is limited to readers who chose to create an account and read while signed in, and it is deliberately separate from the acquisition panel, which stores no identity at all. Anonymous readers are never named anywhere, and no reader's identity is attached to any traffic source.
Is aggregate analytics accurate enough to make decisions with?
For the decisions a web fiction author actually faces, yes. The real questions are whether a promotion moved anything, whether on-platform discovery is doing more work than social posting, and whether readers get past the third chapter, and every one of those is answerable from daily counters compared across a thirty day window. Precision beyond that mostly serves paid advertising, and an author who is not buying ads is paying for a level of resolution they will never use.
by Jacob Tam · August 6, 2026
I run IlorisNovel, a platform where web fiction writers keep 93% of what they earn and readers are never tracked or advertised to. Both halves of that were the same decision.