How many direct reports is too many?

This question rarely gets asked from a comfortable spot. It gets typed the week the org chart hands you reports seven, eight and nine, and what you want to know is whether the drowning feeling is normal. The stock answer is a number: five to eight, or seven, plus or minus two. The number arrives with more confidence than its sources can back, and the sources are worth twenty seconds, because the honest version changes what to do about the roster you already have.

Where the folklore numbers come from

The seven is borrowed. Miller's 1956 paper measured how many items — digits, tones, words — people can hold in immediate memory. It says nothing about org design, and Miller himself closed it wondering whether the recurring sevens were anything more than "a pernicious, Pythagorean coincidence." The number migrated into management writing because it had authority and a memorable title.

The five-to-eight has an older source. Graicunas, a management consultant writing in 1933, counted the relationships a manager is party to: not just you and each person, but the pairs and groups among them, and your relationship to each of those. By his arithmetic, four reports means 44 relationships, five means 100, and eight means over a thousand. It's a stylized model rather than a measurement (nobody experiences "a thousand relationships"), but the direction survives scrutiny: the load grows combinatorially while the roster grows by one.

And in practice, spans aren't set by any of this. Neilson and Wulf tracked CEO spans doubling from about five in the mid-1980s to nearly ten two decades later, moved by delayering and budgets, not by any new finding about attention. Your span was probably set the same way.

What actually breaks

So there's no blessed number, and asking for one slightly misses what degrades. Meetings scale badly but visibly; you can see a full calendar. What breaks quietly is the read: the current, evidenced picture of each person you manage. The signs are consistent:

1:1s drift into status meetings, because status is the only thing you can still hold for everyone. Project state has tools behind it; a person's state has only you.
Reviews start reading generic. If two paragraphs could swap between two people without anyone noticing, the reviews are describing your memory, not the people.
Attention starts following noise. The person with the loud problem gets your thinking; the quietly competent get a nod a month. The resignations "nobody saw coming" tend to come from the second group.
You could pass the ten-question test for two people on the team, and you know which two.

The fixes, in order of honesty

Shrink the number, if you can. A lead who takes three people, a team split, a real conversation with your own manager about the org design. When this option exists it beats every note system, and a page like this one has no business pretending otherwise. The rest is for when it doesn't exist, or not this quarter.

Move the memory out of your head. The method and the template are the same at any span; what changes past six or seven people is that they stop being an upgrade and start being the difference between having a read and having impressions. Four adaptations for scale:

Sweep on a rotation, not on demand. With three reports, re-reading before reviews is enough. With nine, put it on a schedule — two pages a week touches everyone in a month — because "when it comes to mind" is exactly the trigger that stops firing.

Triage by staleness, not by noise. Sort the pages by last-touched. The one that's gone longest unwritten is your next 1:1 prep, and it will keep being the quiet person the noisy week crowded out. This one habit redirects attention from where it's pulled to where it's thin.

Keep the why next to the read. At this span you act on reads you formed months ago, and the basis fades before the conclusion does. "Ready for the next level" is worth little in a calibration room if you can't reconstruct what convinced you; a read that kept its dates and evidence still has the reasoning attached when the calibration meeting or a skip-level asks for it.

Lower the bar per entry. A line or two per person per week beats the thorough version you'll abandon by Thursday.

Stop budgeting on effort. The part that fails at nine reports is storage, not diligence; the forgetting curve is indifferent to how much you care. If keeping notes feels like an admission that you can't hold nine people in your head, it is one, and it's accurate. That's the one thing the borrowed research got right: heads are small.

There's no number at which this page starts applying; there's a symptom. When the reviews go generic, or the test fails for half the team, the span has outgrown memory, whatever the org chart considers normal. A record has a ceiling of its own: fifteen shallow reads aren't better than eight real ones, and no notes system fixes a span that's genuinely too wide. Shrink it if you can; write it down either way.

Full disclosure: I make Continuum, a Mac app whose opening screen is the triage above, running continuously: who's gone longest unwritten about, which beliefs have gone stale, where something looks like it's shifting. Honesty about pricing, since this is the page where it bites: the free tier covers three people, the team sizes this page describes are the paid case, and Pro is a one-time purchase if you'd rather not subscribe. Everything above works with a folder of docs sorted by date modified; the app is for when the rotation has to survive quarters too busy to run it by hand.