5 Months on Substack: What the Algorithm Really Measures

For five months I ran almost every standard strategy. Result: Substack doesn’t evaluate “content” and doesn’t evaluate “subscribers.” It evaluates attention wiring: entry → retention → return → external reading → next hop.

This is a 5‑month report on Substack results: how Substack Analytics works and what the platform actually “weighs.”

Inside:

  • what the metrics ⟦PERCENTAGE DISCOVERIES⟧ / ⟦VIEWS⟧ / ⟦AVERAGE TIME READING⟧ / ⟦RETURNS⟧ / ⟦EXTERNAL DISCOVERIES⟧ mean;

  • why Notes, chats, and comments produce different kinds of signal (impulse vs weight);

  • why Mail and platform are different scenes, and how that breaks conclusions;

  • how “dead” subscriptions dilute percentages and statistically weaken a strong text;

  • where the boundary runs between a bridge and field maintenance.

I’m posting screenshots from analytics and email; you can read this two ways: analyze or take my word for it.

part 1. about chats.

You know, Cannot Name It
Findings — what actually moves Substack now. Viral Chat in Substack — how it really works
After three months of close observation, here’s what became clear…
Read more


Part 2 of the cycle. Substack evaluates not an audience, but wiring

Entry (compressed)

I read Substack as an attention system: where the signal passes, where it dies, where the numbers lie, and where they expose the mechanism.

Questions — in the comments. I’ll publish a separate detailed response later: all my observations, behavioral strategies, and error analysis — what pays back, what drains you, and where the platform turns you into a servicing node.

My difficulty is not analysis. My difficulty is packaging: turning it into an article or an analytical report, editing, graphs. So I’m simply posting the screenshots — up to you: analyze them if you want, or believe me if you want.

🔻What do you do every day “for growth” — and what does it do to your thinking?


Frame (fixing it upfront)

My Substack path started from zero: no English, no SEO, no marketing strategies. The theme is narrow, the texts are complex for a wide audience — in short, this is not content for fast growth.

Next — a dry breakdown, point by point:

For five months I ran almost every standard blogging strategy (except sweet diaries about coffee and cats). And I saw something else:

Substack doesn’t evaluate “content” and doesn’t evaluate “subscribers.” It evaluates wiring — the ability of a text to plug into attention flow and hold it.

One more fixation: I’m intentionally splitting this into a cycle.

Because as publishing went on, it became obvious: the most basic things (authorship, quoting, mentions, links) are technically unclear for a huge number of people.

Not at the level of “they don’t want to,” but at the level of:

This cycle is not teaching.
This is a record of how the system actually works, seen from inside and confirmed by metric behavior.


Method (so the text stands on something, not on mood)

I’m not pretending I have a “representative sample of all Substack authors.” I have what anyone inside the system has: my curves, my entries, my inbox, my behavioral experiments.

I measured:

Have you ever looked at the same post after 24 hours and after two weeks — as two different objects?


The pain I hear most often (and lived myself)

Almost immediately after the first publications, different authors started writing to me in DMs — independently, with the same question.

It sounds different, but the core is one:

“I read, I try to support others, I comment, I participate in chats, I push the field. But it’s insanely time‑consuming and energy‑consuming. Is it even effective? Where’s the line? When do I stop being a thinker and become the platform’s service staff?”

I recognized that question instantly, because I lived it in my body.

My texts 3–4 months ago were stronger and denser.
Not because I “burned out” or “forgot how to write.”
But because attention started to fragment:

It creates an illusion of movement — and at the same time grinds the thought down.

Here’s the key fixation that must be said hard:

Substack really does reward participation.
But it doesn’t distinguish who you are — a thinker or a servicing node.

The algorithm sees movement.
It doesn’t see the price of that movement for you.

So a conflict appears:

This boundary isn’t solved by universal advice.
It’s solved by architecture.

Later in this cycle I’ll break down:

This isn’t a question of “being generous or selfish.”
It’s a question of preserving intelligence inside a system that rewards scattering.

Where do you feel the moment when “participation” starts eating your line?


Key chapter: the cost of servicing the field and the hidden asymmetry of return

0. Starting point (what people usually hide)

For the first 5 months I actively read, commented, and promoted other authors.

Not formally.
Not “nice post.”
Not three lines.

My comments:

In parallel I:

That was work.
By labor intensity, closer to an editor or a field curator than a “reader.”

1. What I got in fact

If you remove illusions, the result was:

The graphs showed clearly:

Most authors:

That’s the core asymmetry:

the system can count the signal,
but people are not obligated to recognize its value.

2. The mistake I saw too late

For a long time I was looking at Substack only through the site.

When I opened my email inbox and started cleaning it — the picture was different.

Facts:

Rule I fixed:

If within ,: 1 MONTH MONTHS a person didn’t find a way to actually introduce themselves — they no longer have access to my attention.

#5 here — email counter / Substack folder (with sensitive info blurred).

The fact that finished me: I unsubscribed from 600+ authors — and I got exactly one email back. The only question wasn’t “how are you?”, not “let’s meet,” but: “what didn’t you like?”

3. Email showed what Substack analytics doesn’t make visible

Through email it became visible:

Subscriptions, reading, and activity do not match.

And one more thing: reading from email reduces platform noise — attention doesn’t scatter across random notes, articles, and authors; you read what you chose, not what you’re being fed.

4. The main imbalance: servicing vs your own line

I saw that I started spending more energy on servicing the field than on my own thought.

That’s the risk point for authors‑thinkers.

5. Hard conclusion

I:

Do you know the number of people/authors who actually feed your thought — and the number of those who simply occupy the channel?


What Substack actually “evaluates”: the weight model (no myths)

If you remove the shop window, the mechanics look like this:

entry → retention → return → external re‑reading → next hop

Substack doesn’t love “emotion.” It loves weight.

Weight = when a text can hold across time and travel through transitions.

Do your texts live longer than a day — or do they burn on publication day?


Two timelines most people confuse (and burn resource because of it)

Substack reads texts in two different time lines.

Not “better/worse,” not “smart/dumb,” but earlier/later.

Timeline 1. Early impulse (hours — first day)

In this window the system sees:

If the text is:

— the early signal looks weak.

Not because the text is bad.
But because the metric is blunt and optimized for speed.

This is where the illusion is born:

“The algorithm doesn’t like deep texts.”

Not true.
It simply doesn’t see depth in the moment.

Timeline 2. Cumulative weight (days — weeks)

After the initial impulse, the second line turns on — and it matters more.

Here the system starts counting:

And here the flip happens:

Silent readers you don’t see
become the heaviest for the system.

Substack is optimized not for reaction,
but for retention over time.

The main distinction

The mistake is made by those who:

The system is not contradictory.
It’s simply layered in time.


Substack Analytics: what each number actually means (and where people fool themselves)

Official Substack metric guide (to align terms instead of arguing with words):
https://support.substack.com/hc/en-us/articles/5320347155860-A-guide-to-Substack-metrics

Substack shows numbers, but doesn’t explain what exactly they describe.

Most authors read analytics like a school report:
“good/bad,” “a lot/a little,” “up/down.”

But in reality these are behavior indicators, not quality grades.

Layer by layer.

1) (open rate)

What people think:
“How many people were interested in my text.”

What it actually is:
the percent of subscribers who entered contact with the email or notification.

Not interest.
Base purity.

The second confusion appears immediately: where the reading happens.

So open rate is not “newsletter success.”
It’s an indicator of whether your base contains people who even have a chance to enter.

High open rate usually means:

Low open rate usually means:

Key:

Open rate doesn’t say the text is good or bad.
It says what kind of environment that text landed in.

Is your open rate “reader love,” or a dirty base?

2)(views / reads)

What people think:
“How many people read it.”

What it actually is:
how many entries happened on the page.

It’s the most surface metric. It’s sensitive to:

High views can mean:

And say nothing about what happened next.

3) ⟦AVERAGE READING TIME⟧ (average read time)

This is where mechanics start.

What it is:
the average time a person stayed inside the text.

Retention, not entry.

Long time means:

Short time means:

Important:

Retention signal is stronger for the algorithm than views.

(Dashboard → Posts → choose post → Link clicks). If you meant actual time-on-page: Substack’s native analytics typically doesn’t show it — then insert a screenshot from external stats (e.g., Google Analytics time on page).

4) ⟦FULL READS⟧ (completion), if shown

This indicates whether people reached the end.

But the trap:

For the system, returns are stronger than linear completion.

5) ⟦REFUNDS⟧ (returns)

One of the most underrated signals.

A return = a person opens the same text again later.

For the algorithm it’s near‑perfect:

Returns are often created by silent readers.
That’s why they’re so heavy.

6) ⟦EXTERNAL DISCOVERIES⟧ (external opens)

Substack counts:

If reading happens after that — the signal is counted as full.

Meaning:

7) (likes, comments)

The most overrated metric.

It’s social noise, not depth.

Useful for:

But:

likes and comments without retention and returns almost weigh nothing.

8) Why numbers “drop” when the text gets stronger

When you:

— views may drop,
— likes may shrink,

but:

The system reads this as signal strengthening, not failure.

9) The main distinction

Substack is optimized for weight.

That’s why:

Read analytics not as a grade,
but as a behavior map — and it stops scaring you and starts working.


Subscribers ≠ weight (and how dead subs weaken a post)

Subscribers are potential entry, not strength.
If they:

— for the algorithm they are almost zero.

How inactive subscribers weaken a post (not just “don’t help”)

The algorithm counts percentages and density, not absolute numbers.

Roughly:

In the second case the text is five times stronger as a signal.

When you have many “dead” subscribers:

The algorithm doesn’t know why they didn’t read.
It only knows: most entered people did not enter presence.

Therefore:

That’s why:

Substack rewards not audience growth,
but attention concentration.

Are you ready to look at “subscriber growth” as a possible signal deterioration?


Silent readers: who actually moves metrics (no romance)

A silent reader isn’t “passive.” It’s a person who:

They produce little social noise and a lot of behavior weight.

Why they’re most valuable to the system:

The route:

  1. A reader arrives via direct link from email/messenger.
  2. Reads slowly.

  3. Returns in a day/week.

  4. Forwards to one more person.

The algorithm doesn’t see “wow.”
It sees:

And concludes:

this text holds on its own and spreads on its own.

Who is your “silent heavy”? Do you allow for the fact that your most valuable people never write to you?


Strategies authors usually try — and what I experienced

I ran this not as an “author,” but as someone who measures behavior and doesn’t believe words.
Not theory. My experience and my stance.

1) “Regular schedule” (frequency as prayer)

What I experienced: regularity alone doesn’t lift you. It only makes you predictable.
What I think: predictability without movement is tidy silence.

2) “Long essays” (hoping depth is a magnet)

What I experienced: long texts produce better presence metrics (time, returns), but don’t guarantee new entry.
What I think: depth keeps those already inside. Entry is created by connections.

3) “Short notes” (pace instead of meaning)

What I experienced: short posts pick up reactions easily, but they fall apart fast. You keep tossing matches in, just to make it look like it’s burning.
What I think: pace without wiring is a bonfire fed by matches: lots of motion, very little heat.

4) “Optimizing headlines/topics” (window dressing)

What I experienced: a headline affects entry, but doesn’t hold. Spikes exist — no tail.
What I think: a storefront doesn’t replace routes. The system loves not storefronts, but transitions.

5) “Recommendations & reciprocity subscriptions” (mutual dust)

What I experienced: the feed fills with similar voices, emails flood, reading becomes duty.
What I think: reciprocity is attention inflation. Numbers rise, weight falls.

6) “Series / rubrics” (format instead of movement)

What I experienced: a series helps retain the core, but doesn’t create crossings by itself.
What I think: a rubric is a container. Fuel is conversation and connections.

7) “Paid access” (money as filter)

What I experienced: paywalls are not about money — they block random noise. Reading density grows.
What I think: paywall is a depth choice at the price of visibility. Not monetization — a sanitary cordon.

8) Notes as growth (sparks without wires)

What I experienced: subscriber growth often comes through Notes — fast, jerks the graph. But it’s a short impulse: entry flash without retention guarantee.
What I think: Notes are good as a gateway, bad as a container.

9) Reading happens not where you think (email ≠ platform)

What I experienced: new people read as an event: from notifications, from the app, from forwards, from direct links — often leaving no trace.
What I think: if you keep optimizing “the email” as the main stage, you’re talking to an empty hall.

10) Manual cleanup (not drama — math)

What I experienced: I cleaned manually and watched percentages and densities change.
What I think: dead subscriptions dilute text weight. Cleanup is not loss — it’s measurable signal strengthening.

11) Comments as growth (servicing the field without a lever)

What I experienced: comments create presence, but easily turn an author into service staff.
What I think: commenting without your own line is working on someone else’s orbit.

12) Collaborations (often decorative)

What I experienced: “togetherness” often looks like mention exchange without real reader transition.
What I think: only what gives a reader a route works: quote → link → transition → next author.


The problem is one: almost all of this thinks in terms of output (how much I produced).
But the system lives in terms of transitions (where and how attention flows).

🔻 Do you measure “how much you wrote,” or “where attention went after the text”?


Insert: what the “top layer” says (short, to the point)

I don’t believe in “authorities.” But I use them as a control point: what those who build the system say, and what those sitting on top tables do.

📌 ВСТАВЬ БЛОК С ЦИТАТАМИ here.

  1. (Translation) A Notes algorithm architect states the goal directly: people should discover, subscribe, and ideally pay — the feed is built for that.
  2. (Translation) Substack’s own explanations link supporting others and participating in conversation to growth — but it’s growth with a cost if you lose your line.

  3. (Translation) One strong author describes the difference from standard social networks: there the “algorithm is your boss,” here it’s closer to direct contact with those who actually need it.

📌 Insert the original links for the quotes directly in this section.


Technical block: authorship, quoting, mentions — and why “simple” turned out unclear

(This section appeared because I kept getting the same DM question: “Can I write an article about your article?” The fact that this question exists shows the scale of misunderstanding.)

I received many questions, and it became a separate symptom: people don’t just “not use a lever” — they don’t technically understand what counts as quoting, what counts as authorship, what’s “allowed,” what’s “needed,” and how to insert it so it works.

Breaking the confusion.

0) This isn’t decoration. This is wiring.

Substack doesn’t read “topic” and doesn’t read “beauty of thought.” It reads transitions.
Link → click → reading → return → next click.

Quote + link + author name = a node.
“Thanks to N authors” = a decorative sign on a wall without doors.

1) What quoting is — and what you can do

You can quote anything. From anywhere.

A sentence from an article.
A paragraph from a book.
A thought from a letter.
A line from a note.
A fragment from my text.

Quoting isn’t “permission.”
Quoting is a form of thinking and a form of connection.

Short version:

You can quote any fragment of any text if you:

  1. clearly show it isn’t yours,

  2. name the source,

  3. give the reader a route to the original.

And yes — writing an article about my article is allowed and welcome if you have something to do with it.

2) Correct quoting is not “how much,” but why

Not volume. Not percentage.
Answer one question: why is this fragment here.

A quote works if:

A quote doesn’t work if:

The form is simple:

3) Can / must / can’t (no legal theater)

You can:

You must:

You can’t:

Rule:
If you remove the link, the text must become worse.
If you remove the link and the text stays “complete,” you used someone else’s blood as paint.

4) Where to get an author handle (where 80% break)

  1. Open the author profile.
  2. Look at the URL — the identifier is usually there.

  3. In Notes/Chat the identifier works as an @ mention.

5) How to quote technically (step by step)

A) In a post

  1. Open the original → copy the URL.

  2. Choose 1–3 sentences.

  3. Insert the quote as a separate block.

  4. Immediately under it: author name + link.

  5. Then: your thought (what this quote does in your text).

📌 Insert the link directly under the quote, not at the bottom.

B) In Notes

Notes are a transition node.
Formula:

C) A book / external source

Short fragment + author + title + (if possible) link.


Mini‑practice (so it becomes action)

1) Every post = minimum 2–3 connections

📌 Insert 2–3 author links directly where you analyze them.

2) “Thanks” becomes “analysis”

“Thanks to these authors” does nothing:

Minimal live analysis (one author = one paragraph):

  1. What mechanism you saw.
  2. One short quote.

  3. Link to the original.

  4. Author name next to the link.

Test:
— remove links and names and the text collapses? → analysis.
— the text remains self‑sufficient and smooth? → simulation.

for example,

Thanks for the quote.

https://substack.com/@frankydyson/p-183013994


What really mattered on Substack in 2025 according to a post I read meny times written by @Lintara from @You Know, Cannot Name

absolutely:

3) Notes as a gateway

After the post, make a Note:

Not for “promotion.” For the graph.


Questions in the comments (so I publish the continuation)

Write your questions under this text.
I’ll answer with a separate publication: all my observations, behavioral strategies, and error analysis — what pays back, what burns resource, and where the boundary runs between “bridge” and “servicing.”

What gives you an impulse (spike) — and what gives you weight (tail)?


Break (no conclusion)

I unsubscribed from almost everyone and left 40 authors.
Not because I was “tired.”
Because noise destroys wiring.

Question:
are you writing a text — or building the transitions through which it can live at all?


⬛ A BLOCK OF LINKS TO INSERT


Lenny’s Newsletter
500,000
This week the newsletter hit 500,000 subscribers, which I’m told makes it among the top five largest newsletters on Substack and one of the largest newsletters broadly. Friggin’ bonkers…
Read more

https://substack.com/@karencherry/p-180139608

Chris Best
Secrets of the Substack algorithm
I chatted with Jairaj (by far the most intelligent of the Substack founders) about how he’s building the Substack feed to help writers and creators have more influence and make more money…

Listen now

https://substack.com/@jairaj/p-155547352

Today we’re announcing the launch of $20 million in guarantees to help creators move their paid-subscription audience to Substack. The Substack Creator Accelerator Fund is designed to help creators expand their reach and business by taking advantage of Substack’s growth network and full suite of publishing, community, and discovery tools. On Substack, they can build their own home on the internet: one where creators, not platform

https://www.emmagannon.co.uk/

Emma Gannon, described last year by the Bookseller as “one of the most popular novelists on Substack”, says that “the thing I love about it is it’s sort of unlike classic social media. It’s based on interests, rather than the humblebragging of showing your life as a highlight reel. People are geeking out on Substack about the things they love: writing, knitting, gardening. It’s got a different vibe to it, because people are showcasing what they’re interested in rather than what they are doing.”

You know, Cannot Name It
How Substack Really Works: Core Audience, Metrics, On conversion, silence
substack philosophy, writing metrics, open rate myth, creative systems, conversion and meaning, attention economy, solitude writing, non-monetized content, digital minimalism, resistance to optimization…
Read more

You know, Cannot Name It
How Substack Really Works: Core Audience, Metrics, and Silent Readers Explained
Premise…
Read more

You know, Cannot Name It
The Bestseller Substack’s Illusion: When an Author Becomes an Anomaly
A sharp, humorous, and deeply analytical look inside Substack’s “Bestseller” mechanics — from velocity and retention to anomaly detection — told from the perspective of an author repeatedly misclassified across categories. A must-read for anyone writing in Science, philosophy, or at the edge of the algorithm…
Read more

You know, Cannot Name It
A quiet discovery inside Substack: what recommendations really are, and why they reveal the truth you learn last
I found a hidden corner of Substack — a place I’d never opened — and realized that people had been seeing me long before I saw myself. This essay is about that moment…
Read more

You know, Cannot Name It
THE GOD OF ALL ALGORITHMS: Cross-Promotion Substack
By: A person who realized email marketing isn’t spam if it’s clever…
Read more

You know, Cannot Name It
2.5 months ago, I was the only one here. And now it’s you.
2.5 months ago, I had zero reach, no SEO, no strategy, and no idea how to “grow” anything…
Read more

TECHNICAL GRAPH READ (my screenshots, plain )

Over the period shown, the system is not behaving like “more subscribers → more reads.” It behaves like “more routes → more reads,” with subscribers lagging behind.

In the 90-day overview, total views jump to 40.2K (from 7.15K) while subscribers rise to 527 (from 169). Views scale faster than the list size, which is a signature of distribution/entry routes expanding beyond your existing base.

Timeline-wise, the chart reads as four phases:

  1. Before Oct 2: organic baseline—low, steady, no structural step changes.
  2. Around Oct 25: the peak period—this looks like a real baseline lift, not a single spike. Your “many strong posts / high frequency” explanation fits the curve shape (the background level changes, not just the top of one bar).

  3. Around Nov 20: a discrete subscriber event (sharp drop) followed by continued growth—this looks like a non-organic action (cleanup/reset), not normal churn. Importantly, the system keeps moving afterward.

  4. December: you switch to Notes and publish fewer heavy posts; the charts show weakening of sustained views/“tail.” Movement continues, but accumulation fades.

Source split is the clearest mechanical signal. Total top-source views are 46,413, and they are dominated by:

That means most reading is happening via email + direct links, not via internal Substack discovery surfaces. “Direct to app” and “Substack app” behave like a separate circuit: they’re where conversion actions tend to happen more than in email.

The “before/after cleanup” screenshots show the cleanest effect. In November, Opened rates sit around 19–22% across posts with views roughly 132–241. After deleting inactive subscribers, Opened jumps to 41–46% across posts while views per post are lower (104–127). That’s the textbook trade: reach shrinks, signal purity doubles. It doesn’t mean content got worse; it means the denominator changed and the list became more “contactable.”

There’s one strong outlier after cleanup: 337 views with only 28% opened. That mismatch is exactly what you’d expect when views come from non-email routes (direct/app/external shares/Notes). It’s the clearest internal proof in your data that views and email opens are different events and can decouple.

Net mechanical read: October heavy-post phase produced a structural lift; December Notes-heavy phase produced activity without the same sustained tail; the cleanup improved Opened% dramatically; and your system is driven primarily by email/direct routes, with app routes acting as the conversion channel.

before removal, the opening rate is 20-23%


after deleting inactive subscribers.

until October 2 – organic growth. after that, the algorithms started to promote me.




The activity in the notes did not produce much of the expected results. The most active days in the substages are December.


Where you are now

This text is part of Lintara’s forensic analysis of Substack as an attention system — examining how metrics, participation mechanics, and reader behavior translate into actual weight, retention, and long-tail visibility.

It belongs to the cycle How Substack Actually Works, focused on how the platform evaluates wiring rather than content:
entry → retention → return → external reading → next hop.

How to Read My Texts

Series: How Substack Actually Works
Category: Media & Substack

<

p class=”button-wrapper” data-component-name=”ButtonCreateButton”>Subscribe now

Share You know, Cannot Name It

Leave a comment

Share:

Like this:

Like Loading…

Discover more from Lintara

Subscribe to get the latest posts sent to your email.

Leave a Reply

Scroll to Top