When people ask your assistant what kind of mood you are in before they come to see you, you have a problem that is already costing more than you think.
The leader in question is usually not difficult in any way they would recognise. Not abusive, not unfair, frequently well liked. Just variable. The same proposal gets an enthusiastic response in March and a sceptical one in May, and nobody can identify what changed.
What variability costs
An organization spends a certain amount of its energy forecasting the person in charge. Where the forecast is easy, that expenditure is close to zero. Where it is hard, it becomes a significant standing overhead that appears in no budget.
The specific behaviours are recognisable once named. People time their requests rather than making them when the matter arises. They test small versions first to gauge the response. They ask a colleague who saw you this morning. They hold a piece of bad news for two days waiting for a better moment, and the two days matter.
All of this is intelligent behaviour under the circumstances, and none of it is work. It is the tax an organization pays for a leader who cannot be predicted, and the leader is the only person who cannot see it being paid.
Consistency beats the content of the rule
This is the claim worth arguing with, because it runs against the intuition that the right standard matters most.
A reliably demanding leader is workable. People calibrate, they bring their best material, they stop proposing things that will not survive, and the standard becomes part of how the place operates. A reliably relaxed leader is also workable, with different consequences, but people know what they have and organise around it.
A leader who alternates between the two without a discernible rule is worse than either, and worse by a considerable margin. There is no stable standard to meet, so people cannot prepare. What they optimise against instead is your state, which means the quality of a proposal matters less than the hour it is presented. That is a corruption of the whole enterprise, and it arrives without anyone deciding anything.
The three that get modelled
People do not track everything about you. They track three responses, closely, and build their entire operating model from them.
How you receive bad news. Not what you say about wanting to hear it. What your face does in the first two seconds, and whether the person who told you had a worse day afterward.
What happens when someone disagrees with you in front of others. A single instance of this going badly is worth more than a year of stated openness, and it is observed by everyone present and relayed to everyone absent.
What you do about a missed commitment. Whether it is addressed, whether it is addressed the same way for the productive person and the marginal one, and whether it is addressed at all when you are busy.
Inconsistency in these three does more damage than inconsistency in anything else, because they are the inputs to every decision anyone makes about what to tell you.
Where the variation actually comes from
Rarely temperament. Three causes account for most of it, and two are fixable.
Fatigue and sequence: the answer depends on what meeting you just left. Real, human, and reducible by not scheduling consequential conversations in the wrong slot.
Information asymmetry: you responded differently because you knew something. This is the one that produces the most unnecessary damage, because from your side the two responses are entirely consistent and from theirs they are arbitrary. The whole fix is one sentence explaining the constraint you are reasoning from. Leaders skip it constantly, on the grounds that the reason is confidential, when the reason could be characterised without disclosure in almost every case.
Genuine change of view, unannounced. You have moved on something and never said so, and people are still working from the old model.
Predictable is not rigid
The distinction that saves this from becoming an argument for inflexibility: what should be stable is the basis on which you decide, not the conclusions you reach.
A leader who consistently weighs safety above schedule will produce different answers in different situations, and every one of them will be predictable to anyone who knows the criterion. A leader who reaches identical conclusions regardless of circumstance is not being consistent; they have stopped evaluating.
Which makes the practical instrument straightforward. Say the criterion out loud, and keep saying it, so people can predict you by reasoning rather than by reading you. That is the difference between an organization that has to watch your face and one that can work out your answer before they ask.
The second condition is where capacity actually comes from. People who can simulate your reasoning will make the decision themselves, correctly, without the meeting. Nobody can simulate a mood.
Edited by Patrick J. Wolf, PhD