# Checking in *Make silence mean something* --- Most of the advice on managing people is about keeping in touch. Hold a weekly one-to-one with each of your reports. Get the leadership team together every week. Review each function properly once a month, and hold some skip-level meetings with the people below your reports, so you hear what doesn't travel up. Every one of those is sensible, and for someone running a team of five they're close to the whole job. Add them up for a managing director with nine direct reports. Nine one-to-ones at forty-five minutes is nearly seven hours. A weekly leadership meeting is another two hours. Nine monthly reviews of an hour each come to about two hours a week, and skip-levels another hour a week. That's around twelve hours, more than a quarter of a working week, before the preparation each one needs, any customer, the board, or a single decision of your own. Most senior diaries look something like this, and most of the people living in them feel busy and somehow behind. Nearly all of those meetings are the same kind of thing, though: a check you've scheduled in advance to find out whether anything has changed. Most weeks, for most of your people, nothing has, or something has and they'd have told you anyway. ## Polling and interrupts Computers have the same problem with the devices attached to them: how to find out when one needs attention. Engineers have names for the two answers. Polling means going round each device on a schedule and asking whether anything has changed, paying for every check whether or not anything has. Interrupts mean getting on with the work and letting a device speak up when it needs attention. Polling is simple and predictable, and wastes most of its effort when little is happening. Interrupts cost almost nothing in a quiet week, as long as the device can be trusted to raise an interrupt when it matters. Most of the common advice is polling, and polling costs the same whether anything is happening or not, so the bill grows with every report you add. The twelve hours above is what going round all nine costs, every week. <iframe src="https://anishshailpatel.github.io/instruments/checking-in.html" title="What scheduled checking costs a week, against how many of your reports you still check on" width="100%" height="1060" style="width:100%;border:0;display:block;margin:1.2em 0;" loading="lazy" sandbox="allow-scripts allow-same-origin"></iframe> Which of the two you can use with someone depends on what their silence tells you. With some people, if they're stuck they'll say so, and until they do you can assume the work is moving. With others, silence could mean anything, so you check. So the number that sets your real span of control is how many of them you have to check on. [[The missing middle]] is about getting spans wrong in the structure; polling is how a span that looks fine on the org chart can still be one you can't carry. ## When someone does speak up You have more say over what someone's silence means than it feels like you do. Engineers would say you write the interrupt handler, the part that decides what happens when someone does speak up. If someone's first piece of bad news got a bad reaction from you, or you stepped in at the first wobble and took the work back, they'll have learned that raising things is expensive. They'll go quiet, their silence will stop meaning anything, and you'll be back to checking on them. ## Telling which silence to trust The quickest test is where your last few surprises came from. If bad news about someone's area reached you first through the monthly management pack, a customer or a peer, their silence isn't trustworthy yet, however good the work usually is. The pack is also what makes fewer meetings safe: [[Reading the numbers|a monthly pack read properly]] is a cheap check that covers everyone at once, and the conversations can concentrate on what the pack can't show you. Relying on people to speak up has its own failure. Someone who brings you every wobble is nearly as expensive as someone who brings you nothing, so relying on them only works once you've agreed what's worth raising. When something does get raised, the useful version comes with their view of what's causing it and the part they need your help with, which keeps the problem with them. Their view doesn't need to be right - you can correct a wrong one in a sentence, and without one you're starting from nothing. ## How often to meet None of this means cancelling the one-to-ones. They catch the things nobody knew to raise, and they're where most of the coaching happens. The change is to decide how often to meet per person and per piece of work, rather than run everyone at the same cadence because that's what the advice says. A trusted direct report on familiar work might need a fortnightly conversation; someone new, on something nobody has done before, might need two a week for a while. When you start seeing someone less often, say why, and agree out loud what they'd bring you in between - "tell me if the launch slips more than two weeks" - or they'll take it as you losing interest. --- So it's worth looking at your own diary and counting who you still check on, and why. For some the answer is that they're new, or the work is. For the rest, think back to the last time one of them brought you something early, and what you did with it.