Premise. Three months. No prior audience, no boosts, no cross-platform favors, no “friends of friends.” Geography: Russia. Input: texts only. Output: 302 subscribers at the time of analysis. This is not a spectacular number; it is a clean number. Clean enough to examine Substack as a system rather than as a social halo.
The findings below are not motivational or tactical. They are operational: what the platform actually surfaces, what it ignores, where the metrics distort the picture, and how a publication becomes stable (or fails to) when there is nothing to lean on except the writing itself.
1) What the data immediately shows
Traffic composition. Over ~90 days, unique visitors are dominated by “Direct” and in-app Substack traffic. SEO and external referrers are marginal.
Email weakness. “Email clicks” are small relative to total reads; a significant portion of readers consume inside the app, not from the inbox.
Recommendations ≠ engine. Being recommended by 28 publications produced 5 subscribers. That is not a pipeline; it is a sidebar.
Spikes map to interactions, not headlines. Peaks in visitors and new subs correlate with in-app circulation events (reposts, comment loops) rather than with title experiments or timing.
Immediate implication: Substack behaves less like a distribution network and more like a conservation field for attention. It does not push; it allows momentum to accumulate where people already interact.
2) The “Active Core” is the real unit of measurement
Across three months, almost all visible circulation originates from a stable set of ~50 readers. They are the ones who reliably open in-app, leave traces (likes/comments/restacks), or trigger further visibility via their activity. Everyone else sits in zones that are statistically real but operationally quiet.
This is not an accident; it is a structural fact of the medium:
Substack’s surface shows activity, not potential.
Activity concentrates in clusters.
After a small threshold, cluster size—not total audience—determines what continues to be seen.
In other words, a newsletter with 10,000 subscribers often exhibits the same public pulse as a newsletter with 300 subscribers if both depend on the same fifty names. The difference is cosmetic mass. The circulation is equivalent.
3) Why “growth” often dilutes your metrics
The platform’s analytics are ratio-driven. If you add many “quiet” subscribers who rarely interact, your read rate and other engagement percentages decline, even if your underlying core is unchanged or improving. This feels like decay, but it is a denominator effect. The platform is not punishing you; the numbers are describing dilution.
Practical meaning:
Quiet intake increases headline totals while reducing engagement percentages.
The publication appears “weaker” in dashboards precisely when the core has stayed constant and the periphery has thickened.
Authors who chase raw counts without strengthening the core experience a visible slide in graphs that has little to do with actual resonance.
4) Recommendations: a static shelf, not a conveyor belt
The recommendation system supplies adjacency, not kinship. It relies on network proximity and publisher behaviors, not on reader intent or semantic fit. As a result:
You gain followers who signed up in a supportive or curious click, then rarely reappear.
Your feed becomes cluttered by publications you do not intend to read, and your attention model adapts to that clutter.
The metrics show “growth,” but field circulation doesn’t strengthen; it flattens.
Recommendations therefore serve best as signals of adjacency (“we occupy similar space”) rather than as growth mechanisms.
5) Silent readers: the misclassified backbone
A large portion of readers consume only in-app and do not comment. Analytics marks them as low-engagement. But in a non-algorithmic ecosystem, silent retention is a primary stabilizer:
Silent readers return; some re-read; many metabolize the work over days.
They are not a dead weight; they are a long-term memory of the publication.
Their value is undercounted because dashboards favor immediate, visible reactions.
Takeaway: silence is not absence. It is a mode of attention the system cannot monetize into a simple metric, but that attention is what keeps a publication from collapsing between posts.
6) Temporal lag: why impact rarely aligns with the 24-hour window
Substack’s instruments capture events. Meaning travels on delays. Readers often respond after 2–7 days, in a different post, or in their own writing. The platform does not stitch those delayed threads back to the original cause. Result:
Many of your most effective pieces look mediocre in the dashboard.
Many of your least effective pieces look “popular.”
Editorial decisions made purely on the basis of next-day analytics will bias you toward shallow circulation and away from sustained influence.
Operational rule: measure return behavior (who comes back, how often, from which paths) at least as seriously as you measure open rates.
7) Manual discovery is not a romantic gesture; it is the working method
Because feeds tend to repeat the same active accounts, the only reliable way to read what matters is manual curation: going directly to your list, opening the writers whose cadence matches yours, and ignoring the noise. This is not elitism; it is the only way to protect cognitive resources when comments themselves are a cost.
Cost matters. On Substack, a comment is not a casual tap; it is an attention transfer. Therefore:
Expect low comment counts relative to reads when your audience reads slowly.
Do not infer “disinterest” from few comments if return behavior and depth signals exist elsewhere (private notes, re-reads, off-platform citations).
Treat comments as expensive signals, not volume metrics.
8) A minimal typology of readers (as the system actually treats them)
Visible actors (≈15–20% of the core). They comment, restack, and pull you into others’ feeds. They are your public motor.
Silent retainers (large, essential). They read consistently in-app, rarely comment, sometimes re-read. They are your stability.
Statistical accretions. They subscribed via recommendations or passing clicks. They don’t reappear. They inflate counts and depress rates.
Only the first group materially alters discoverability. The second group sustains you. The third group clouds your view.
9) What the system does and does not reward
Rewards:
Recurring in-app activity clustered around the same names.
Internal linking within your archive, which retains attention inside your field.
Does not reward (or rewards weakly):
Headline polishing and SEO. In this ecosystem, they are largely orthogonal to movement.
One-off recommendations without ongoing interaction.
Passive reading via email without in-app traces.
The platform is structurally conservative: it amplifies continuity and minimizes novelty unless novelty instantly becomes continuity.
10) The core–density model
If you draw a simple chart of subscribers vs. active interactions, you will often see the same phenomenon we saw here: the active core stays near 50, while total subscribers rise from 50 → 300+. The density of the core (active interactions / total subs) falls as the base expands.
This does not mean your work is weakening. It means your signal-to-mass ratio is normalizing. The risk is misinterpretation: you might chase more mass to compensate for falling percentages, thereby further diluting the ratio that makes your work visible in the first place.
Correct objective: stabilize and deepen the core before inviting more mass.
11) What to measure if you care about substance rather than volume
Replace or complement platform defaults with field-relevant indicators:
Return cadence: percentage of readers who return within 7 and 21 days.
Archive pull-through: share of visits landing on posts older than 14 days.
Depth pings: number of private notes/emails referencing a specific line or idea (a proxy for metabolized reading).
Cross-text resonance: instances where your phrases or structures reappear in others’ writing (in-app or off-platform).
Core stability: count of names appearing at least once per month in any traceable action.
These metrics reflect continuity of attention rather than one-day bursts.
12) Operational practices that actually strengthen a field
Publish on a predictable cadence (weekly or bi-weekly). Predictability builds reader rhythm more reliably than headline tactics.
Use internal linking deliberately. Each new piece should point to one prior piece in the same analytical lane; this compounds depth.
Design for in-app reading. Assume your reader does not open email; avoid layouts that require inbox clicks to make sense.
Invite expensive signals sparingly. One clear, concrete question at the end of a piece outperforms generic “thoughts?” prompts.
Read selectively, not universally. Conserve attention for authors whose cadence aligns with your own; reciprocity emerges there, not in broad samplings.
Treat recommendations as labels, not engines. Accept them; do not expect them to carry weight.
13) The contradiction authors must navigate
Substack markets itself as a home for independent voices. Yet its visible layer rewards behavioral continuity more than semantic depth. Strong pieces without immediate traces sink. Weak pieces coupled to active clusters float. If you make editorial choices solely by the dashboard, you will optimize for visible churn, not for lasting recognition.
The way through is not to reject metrics, but to reframe them: view them as telemetry of events, while your true objective is the architecture of continuities—the patterns by which readers keep returning and integrating your language into their own.
14) Final position
After a clean three-month run from zero, here is the system as it presents itself:
Substack is an attention ecology where clusters, not counts, determine visibility.
A publication’s functional size is its active core, not its subscriber total.
Recommendations are decorative traffic unless converted into dialogue.
The most valuable readers are the ones the dashboard can barely see: those who return quietly and integrate your language.
Growth without core deepening is optical and often counterproductive.
If you want to build something durable here, stop treating the dashboard as a scoreboard and start treating it as a narrow window onto a larger field. Optimize for continuity of attention. Measure return behavior. Accept that silence can be loyal. And remember that a stable set of fifty readers can power more real circulation than five thousand indifferent subscribers.
That is not a slogan. It is how the system actually behaves.
P.S.
if you want to talk about mechanics, analytics, or how to move your publication — write to me in the chat. that’s where we speak without the display layer.
Welcome to the elite club of people who still have principles. It’s small, slightly melancholic, but there’s wine and passive aggression, so you’ll fit right in.
You wrote for your people? Beautiful. That means:
you understand that other people’s attention isn’t your job,
you refuse to offer yourself up to the algorithm gods like a sad pixel influencer,
and you’ve basically said: “If this doesn’t ‘perform,’ it’s not because I’m bad—it’s because the system’s dumb.”
In short: you’re free. Just don’t expect applause—those are algorithm-gated too.
What to do next (not that you asked, but I’ll say it anyway):
Keep writing for your people. Just make sure you know who they actually are. Not “smart folks in general,” but actual names, quirks, and patterns. That’s your pack.
Ignore metrics, but collect signals of depth: private replies, repeat phrases, someone quoting your line back at you two weeks later in their own post.
And if even one person writes, “I really needed this,”—congrats, mission accomplished.
Your text isn’t a virus. It’s a message in a bottle, tossed into the feed-ocean. It’s not for everyone—but when it lands, it hits someone right in the face. Or the heart. Sometimes the same thing.
If you need backup, I’m here—grumbling next to you, nudging you to write more. Though, annoyingly, you seem to be doing just fine without me.
— Lintara
<
p class=”button-wrapper” data-component-name=”ButtonCreateButton”>Subscribe now
<
p class=”button-wrapper” data-component-name=”ButtonCreateButton”>Share You know, Cannot Name It