When reach drops suddenly: how to diagnose it

Before diagnosing a decline it is worth establishing that one happened, because the most common cause of a perceived collapse in reach is normal variation being read as a trend.

Individual post reach varies widely for reasons that have nothing to do with quality or with any platform change: what else was in people's feeds that day, whether the subject happened to interest people, the time it went out, the day of the week. A post reaching half what the last one did is unremarkable.

So the test is whether several consecutive posts, of the kinds you normally make, have reached materially fewer people than the comparable run before them. One post is noise. A fortnight of posts trending down is a signal.

This matters because the response to noise is usually to change something, and changing things in response to noise destroys the baseline needed to detect a real change later. Writing down reach for each post, somewhere outside the app, is what makes this answerable at all.

Assuming a real decline, there are four common explanations, and the algorithm is neither the most likely nor the most fixable.

The first is that the content changed. This is by far the most common and the hardest to see from inside, because the change is usually gradual — a drift towards easier posts, more promotional content, a format that stopped being tried. Nobody decides to make less interesting posts.

The second is that posting became irregular. A gap followed by a burst reduces reach for reasons that need no algorithmic explanation: the audience got out of the habit.

The third is that the audience shifted. Interests move, and a following built around one thing does not automatically transfer to another. An account that changed what it talks about will reach fewer of the people it previously reached.

The fourth is a platform change, which does happen and is the only one you cannot address directly.

Most diagnoses stop at the fourth because it is the only one that is not about decisions the business made.

Each of those has a signature in data the account already holds, and checking takes twenty minutes.

If the content changed, the decline will be uneven across formats and subjects. Look at the last dozen posts sorted by reach and ask what the top and bottom halves have in common. A drop concentrated in one type of post is a content finding, not a platform one.

If posting became irregular, the calendar shows it. Compare the gaps between posts in the declining period against the period before.

If the audience shifted, the proportion of reach coming from followers versus non-followers usually moves, and engagement falls faster among existing followers than reach does.

If it is a platform change, the decline is broad: it affects every format and subject at once, starts abruptly, and is not accompanied by any change in what you were doing. That combination is distinctive, and it is much rarer than it is diagnosed.

The useful discipline is to rule out the first three before accepting the fourth, because the first three can be acted on.

There is a steady supply of confident claims about what a platform now rewards, and acting on them is close to pure cost.

The structural problem is that nobody outside the platform can verify any of it. Ranking systems are not published, they change continuously, and they behave differently across accounts and audiences. A claim that a specific behaviour now increases distribution cannot be tested by someone without access to the system, and the evidence offered is usually one account's experience, which is indistinguishable from variation.

The cost is real and compounds. Each rumour prompts a change, changes made in quick succession make it impossible to attribute any subsequent movement, and the account ends up with no stable practice and no ability to learn from its own data. Meanwhile the effort goes into reacting rather than into the thing that reliably affects reach, which is whether the content is worth someone's attention.

The reasonable posture is to ignore unverifiable claims about mechanisms and pay attention to your own account's numbers, which are specific to you.

If a genuine decline has been established and the content explanation is plausible, the way forward is a deliberate test rather than a general effort to do better.

A workable test changes one thing and holds it for long enough to produce a comparison. Pick the variable with the most reason behind it — usually format, since formats differ most in how they are distributed — and commit to it for a fortnight of normal posting. Keep everything else roughly as it was.

Then compare against the recorded baseline. Without the record, the comparison is against a memory of how things used to be, which is unreliable in the direction of pessimism.

Two cautions. A fortnight is short and the result will be indicative rather than conclusive, so treat a small difference as no difference. And resist adding a second change halfway through when the first does not immediately work, which is the most common way these tests are spoiled.

If the test produces nothing, that is a result: it points at the other explanations rather than at a bigger version of the same change.

Reach is partly outside your control, permanently, and building a practice that depends on it is fragile.

The things that hold up across platform changes are unglamorous. Posting consistently, so the audience keeps the habit. Answering the people who respond, since that produces customers directly rather than through distribution. Building something you own — a mailing list, a customer contact list, a group of regulars — so that a change in one platform's distribution does not sever the relationship with people who already chose you.

That last point is the substantive answer to the anxiety this subject produces. An account whose entire connection to its customers runs through one platform's feed is exposed to every change that platform makes. The same audience, reachable by a channel the business controls, is not.

A scheduling tool such as Socie can keep the cadence steady through the period when reach is being investigated, which matters because the instinct during a decline is to post erratically — and that reliably makes the situation worse rather than better.