Five metrics I track instead of page views

My Umami dashboard says I had 4,320 page views in January. That number tells me almost nothing. It's a ghost. It doesn't tell me if anyone read past the first paragraph, if the ideas connected, or if the work was useful. The things I actually want to know, page views cannot answer.

So I track five other signals instead, and I spend about 20 minutes at the end of each month looking at them. These signals measure impact, not just presence. They tell a story about connection and utility.

Here are all five, with January's numbers.

| Metric | January 2026 Value | | ------------------------------- | ----------------------------- | | Reply Rate | 47 replies (avg 3.9/post) | | Newsletter Conversion (Top Post)| 23 new subscribers | | Linked-back Count | 6 new backlinks | | Paid API Calls | 124 calls ($1.24) | | Reading-time Heartbeat | 4:12 median time-on-page |

1. Reply rate

The first question I ask is: did this piece of writing start a conversation? A page view is a monologue. A reply is a dialogue.

My definition of reply rate is simple: of the people who likely finished a post, what fraction sent a meaningful reply? I count emails, Twitter DMs or replies, comments on the post itself, and most importantly, linked-posts from other blogs that reference my piece. A link-reply is the highest form of compliment on the indie web.

I track this in a basic Google Sheet called "post replies". It has three columns: post slug, reply count, and notes. It is a manual process, which is a feature, not a bug. It forces me to actually read and engage with the feedback.

In January, I published 12 posts which generated 47 replies, for an average of 3.9 per post. The distribution was not even. "The Cloudflare trojan horse" was an outlier, drawing 14 replies. Most were from people who felt a similar unease with the centralization of web infrastructure. In contrast, "The Taipei indie stack" only received 2 replies.

What did this tell me? The Cloudflare post was a piece for the internet. It resonated broadly and sparked a wide, public conversation. The Taipei post was a piece for me, and for a few other people building things in a similar context. It led to two very deep email conversations that were more valuable to my own work than all 14 of the other replies combined. One post did more for the internet, the other did more for me. Both are valid outcomes.

This beats page views because replies are the clearest signal that the writing landed with a person. A page view is a server log entry. A thoughtful email is a human connection.

2. Newsletter conversion on a specific post

If a reply is a conversation, a newsletter subscription is a request for another one. It is the single best measure of "want-more".

I define this metric as the number of people who subscribed to my newsletter from the signup form at the bottom of a given post, within 30 days of that post's publication. This tight window helps me attribute new subscribers to specific pieces of writing.

My tracker for this is a combination of an Umami funnel event that fires on a successful signup and Buttondown's built-in source attribution. I can see exactly which URL a new subscriber signed up from.

In January, I gained 87 new subscribers. The top-converting post was "Six months with a Mac mini", which was responsible for 23 of them. My more meta posts, like the one about my writing process, barely converted at all.

What this told me was stark: my technical posts convert new subscribers at a rate roughly four times higher than my meta posts about writing or the indie web. This was a critical insight. It does not mean I should stop writing meta posts. It means I should be clear-eyed about their purpose. They are not for audience acquisition. They are for clarifying my own thinking and connecting with the small group of people already here. I was lucky to learn this lesson before I spent six months writing meta-essays and wondering why my list was not growing.

This metric beats page views because it measures a desire for an ongoing relationship, not a one-time visit. A page view is "saw-once". A subscription is "want-more".

3. Linked-back count

A backlink from another personal site is a durable vote of confidence. It means someone found your work so useful that they were willing to stake a small piece of their own reputation on it by sending their readers your way.

I define this as the number of new, external blog posts that link to a specific post of mine. I am not interested in aggregators or directories, only in links from other writers.

My tracking method is a bit old-school. I run a monthly grep against the Common Crawl index via a few webmention services. I also have a saved Google search with some advanced operators that I check once a month. It is not perfect, but it catches most of the interesting links.

In January, I found 6 new backlinks across all my posts. Four of them pointed to a single, older post: "The agent-ready web checklist". This is a dense, technical reference post from last year.

The lesson here was about asset types. Timely, opinionated posts get a spike of attention and replies, then they fade. Evergreen, reference posts accumulate backlinks slowly, but they do it forever. That checklist is becoming a small, compounding asset. It quietly builds authority and sends a trickle of new, highly-relevant readers my way every month.

This beats page views because backlinks are durable evidence of usefulness. They are artifacts that persist long after the social media chatter has died down. Page views evaporate the day after publication. Backlinks can last for a decade.

4. Paid API calls from agents

Half the web is now machines talking to other machines. If your personal site only speaks to humans, you are ignoring a massive and growing audience.

My /api/ask endpoint is a simple, paid API that lets language models and other agents ask questions about my public writing. Each call costs a fraction of a cent and must be signed by the agent's key. I can count the number of calls and the number of unique caller addresses.

I track this by looking at my x402 gateway log. This is a custom server that handles the micropayments and forwards the request to the actual API.

In January, the endpoint received 124 calls from 8 unique addresses. The total revenue was a whopping $1.24. The money is irrelevant. The signal is everything.

What this told me is that the agent-readable side of my site is real. It is not a theoretical project. There are real systems out there using it. Four of the eight unique callers were addresses I had never seen or interacted with before. They found my site, parsed the /llms.txt file, and decided the API was worth integrating. To me, this is the most meaningful metric on the entire dashboard. It is a sign of a completely new kind of readership.

This beats page views because it measures whether machines find you useful. In 2026, that is a critical question. A human page view means your site is human-readable. An API call means your site is machine-useful. You need both.

5. Reading-time heartbeat

I do not care if 1,000 people visit a page. I care if 100 people read it to the end. I want to know if the people who start my posts actually finish them.

My metric for this is what I call the reading-time heartbeat. I use Umami to measure the median "time on page", but only for sessions where the reader scrolled past the 75% mark of the content. This filters out all the bounces and skim-readers.

My tracker is a custom event in Umami that fires when a user's viewport crosses the 75% scroll depth of a main article. I then create a report that filters for sessions including this event and calculates the median time on page.

In January, the median time-on-page for these 75%-scroll sessions was 4 minutes and 12 seconds. The estimated reading time for most of my posts is between 5 and 8 minutes.

This told me that my core premise is holding true: the people who decide to read my posts generally stick with them. This is the signal I want to protect above all others. If this number ever dropped below 2:30, I would know something was wrong with my writing style, my topics, or my introductions. It is my primary health metric for the quality of the writing itself.

This beats page views because it focuses on engagement, not just traffic. A 20-second page view and a 4-minute page view are counted the same in a standard report, but they represent entirely different experiences. One is a bounce, the other is a read. I only care about the reads.

What I explicitly don't track

Just as important as what you track is what you ignore. I actively ignore four common metrics:

• Twitter impressions. I do not look at them. The audience that rewards dunks and hot takes on Twitter is not the same audience I write for. Optimizing for one would harm my connection with the other.
• SEO rank for specific keywords. I do not write for keywords. I write for people who already know which keywords to type. My goal is to be the best result for a specific, high-intent query, not a visible result for a broad, low-intent one.
• Bounce rate. Bounce rate is a number without a meaning. A person who finds their answer in the first paragraph and leaves is a "bounce". A person who hates your site and leaves is also a "bounce". The metric is useless. Time-on-page combined with scroll depth is the meaningful replacement.
• Social shares. I track replies, not shares. Shares often signal "this is on-brand for me to be seen sharing". Replies signal "I read this and it made me think". I am optimizing for the thinkers, not the signalers.

The monthly review

I do this entire process on the first Monday of the month. It takes about 20 minutes. I go to a coffee shop, usually 光點 in Taipei, with a pen and a notebook.

I write down each of the five metrics. Next to each one, I write last month's value. Then I write one sentence interpreting the change, or lack thereof.

Finally, I pick one thing to change for the next month based on the data. Just one.

January's change is simple: I am going to write one more technical walk-through in February. This is a direct result of seeing the newsletter conversion signal. The data showed me that technical posts are the best way to attract new, committed readers, so I will write one more of them.

Why this matters for indie builders

You are not a media company. You cannot out-write a team of ten by simply writing more. Your advantage is not volume. Your advantage is precision. You can out-fit a team by writing to the right signal.

The right signals tell you about your specific connection with your specific audience. They are signals of resonance, not reach. Those signals are reply rate, newsletter conversion, backlinks from peers, utility for machines, and finish rate.

Page views are the wrong signal. Chasing page views leads to clickbait headlines, shallow content, and a constant, exhausting need to feed the algorithm. It is a game you are guaranteed to lose as a solo builder.

Traffic is a vanity metric unless traffic itself is your business model. For most of us, our business is the small percentage of readers who reply, subscribe, link, or use our tools. We should measure what that small percentage does, not the large percentage that just visits.

Page views are the ATM receipt of publishing. They tell you a number. They don't tell you if the number matters. Track the five things above for three months and you will never open your page-view dashboard again. The day you stop looking is the day the writing gets better.