Distribution mechanics: how feeds decide

The algorithm is not a mystery. It is a prediction machine.

Strip away the folklore and every feed on every platform does the same job: for each person, each time they open the app, it ranks the available posts by a prediction — how likely is this person to engage with this, dwell on it, and stay on the platform because of it? The signals differ by platform and change constantly. The model does not, because the business behind it does not: platforms sell attention to advertisers, so the feed's task is to hold attention. Everything else follows from that one sentence.

This is why "the algorithm changed" is both always true and rarely important. The weights move; the objective doesn't. Reason from the objective and you stop chasing tricks that die monthly — hashtag rituals, posting-time superstitions, engagement pods — and start doing the few things that have worked on every feed for a decade, because they genuinely improve the prediction.

The test audience

A new post is not shown to your whole following. It is shown to a small sample first — some followers, sometimes some non-followers — and the system watches what they do. Strong early response earns a wider circle; weak response ends distribution quietly. Every post is an audition.

Two practical consequences fall straight out of this mechanism:

Hooks matter mechanically, not just artistically. The test audience decides in the same second or two your reader does. Lesson 7.2's craft is also distribution engineering.

Timing matters mechanically, not superstitiously. If your test audience is asleep or at work when the post goes out, it cannot respond, and the audition fails regardless of quality. There is no magic hour — there is only when your audience is actually looking, which is an evidence question. Lena answers it below.

The engagement hierarchy

Not all responses predict equally. As a durable rule of thumb, sends and shares outrank saves, saves outrank comments, comments outrank likes — because each step up predicts deeper interest. A like costs nothing and means little; a send means one human staked a sliver of their reputation to put your content in front of a specific other human. Ranking systems weight accordingly, because that is exactly the signal a prediction machine wants.

The design consequence you already met in 7.2: build content someone would send to a colleague. It is simultaneously the best content advice and the best distribution advice, which is not a coincidence — the hierarchy exists because feeds are trying to predict genuine human interest, and genuine human interest is what you should be earning anyway.

The honest part: organic reach declines, and why

Now the uncomfortable section, stated plainly because you will meet it in your own numbers.

Organic reach — the share of your own followers who see a given post without payment — has fallen across every mature platform, and it falls for structural reasons, not because anyone is persecuting you:

• Supply outgrows attention. Every year more accounts post more content into feeds whose users' hours are fixed. Your post competes against more candidates for the same slots.
• Recommendation displaces following. Feeds increasingly fill slots with content predicted to interest the user rather than content from accounts they follow. Good for discovery when you win the prediction; bad for the reliable follower reach you thought you owned.
• Scarcity is the product. A platform that gave businesses free reach forever would have nothing to sell them. Organic scarcity and the "promote" button are two ends of the same business model.

For a small account this means: most posts reach a fraction of your followers, the fraction is not guaranteed, and a follower count is best understood as permission to audition, not an audience you hold. Ostara's canonical numbers say it quietly: 14,200 followers, net growth of about 90 in six months, on an account posting four to five times a week. The followers are real. The access to them is rented — which is why M5.L3 changed Instagram's job to feeding the email list, and why Module 8 exists.

Why posting more is not the answer

The intuitive response to falling reach is volume. It fails, for mechanical reasons. Each post is auditioned on its own early response; doubling output with the same material halves nothing and dilutes plenty. Your test audience — the same few hundred people — starts skimming you, early response weakens, and the system's estimate of your account drifts down with it. Flooding also spends the one budget that actually binds you: your hours.

What the mechanism rewards instead is consistency at a sustainable cadence in recognisable formats — the commitments you made in M5.L3 and 7.2. A steady account in a stable format gives the system clean data and gives the audience a habit. An erratic account never leaves the audition stage.

Organic, boosted, paid

Three different tools, often confused:

• Organic builds trust and assets and compounds slowly. It rarely scales on its own — that is not failure, it is the structure described above.
• Boosting — the button on a post — is paying to show that post to a vaguely defined audience the platform chooses. For a targeted business it mostly buys impressions from the wrong people. Treat it as what it is: the easiest possible way to spend money, not the best.
• Paid social proper — campaign objectives, real audience definition, measurable outcomes — is a different discipline with its own module. When a post proves itself organically, that is a creative test you got for free; Module 9 is where you decide whether to put money behind proven creative, properly.

The discipline this lesson asks of you: write down the condition under which you would consider paying — "when a post exceeds my rolling engagement average by X and points at a working landing page" — so that the decision, when it comes, is triggered by evidence rather than mood.

Lena runs the experiment

The canonical case. Lena's four-post LinkedIn series from M5.L5 — the onboarding-data findings. Posts one and two went out mid-morning on Tuesdays, into the working hours when practice staff are buried in client work: around 420 and 510 accounts reached, a handful of reactions, distribution over by lunch. Before post three she went back to her M2.L3 desk-research notes and looked at the timestamps on the forum threads she'd mined: the tool-recommendation conversations clustered between 20:00 and 22:30. Post three — same format, same series — went out at 21:15 on a Thursday. It reached about 2,600 accounts and was shared 14 times, mostly practice owners tagging colleagues: the "send to a colleague" design doing exactly what the engagement hierarchy predicts.

Same account, same quality, five times the reach — because the audition finally happened in front of an audience that was awake. Lena then hovered over the boost button and closed the tab: "people interested in accounting software" is not "partners at 3–25-person Dutch practices", and she wrote her paid rule instead: consider paid only behind organically proven posts, with real targeting, after Module 9.

---