Expertise · Communication
Clear is not the same as understood
Two capable people leave a meeting convinced they agreed. A week later each is angry that the other did not deliver. The problem may not be how the message was phrased. Perhaps the two people heard different things, did not check their understanding or have different interests.
Engineering and operations each think the other is holding something back
At a fictional manufacturer the engineering manager and the operations lead leave a planning meeting agreeing that “the new line will be ready by the end of the month”. Operations understood “ready to run”. Engineering meant “installed and tested”. Ten days later each has privately concluded that the other is unreliable, and the emails have turned formal.
Their director is asked to step in. She can sit them down and help them settle it themselves, she can decide for them, or she can tell them to sort it out. Which of the three is right depends on what kind of problem this is, and that is not yet clear.
Illustration · where meaning gets lost
From a message to a shared agreement
Why engineering and operations talk past each other
Better wording fixes only the first. The others call for something else, and applying the wrong remedy makes the next conversation harder.
1 · The message was ambiguous
Mechanism: a word like “ready” carries a different meaning in each role, and nobody noticed.Cue: ask each to write down separately what was agreed. The two versions differ on a specific word.Move: end every agreement with a read-back: who does what, by when, and what “done” looks like.2 · An interaction loop
Mechanism: each person’s behaviour invites the other’s. Pressing tends to draw withdrawal, and hostility tends to draw hostility. Each then describes the other in terms of character.Cue: escalation by one follows retreat by the other, in a repeated sequence.Move: change your own move first, for example a question in place of a demand, and watch whether the other shifts.3 · Listening that closes down
Mechanism: solving too early, advising unasked, or “helping” in a way that signals the other is not capable.Cue: replies start with “yes, but” or a solution, the other says “that is not what I said”, and open questions are rare.Move: say back what you heard before replying, and ask open questions such as “what would make this workable for you?”.4 · A real conflict of interest
Mechanism: the targets conflict (a quality gate on one side, output on the other), and better wording cannot resolve that.Cue: after a clear read-back they still disagree, and the disagreement is about priorities or resources.Move: stop trying to communicate better. This is a decision about priorities, and someone with the authority has to take it.It is worth separating conflict about the work from conflict about the people. The distinction is useful, but the popular claim that conflict about the work is generally productive is contested: meta-analyses found both kinds, on average, go with worse team results, with benefits only under specific conditions.
When the director steps in: mediator or judge?
A leader can help two parties find their own agreement, or listen and decide. Both are legitimate. What damages trust is switching without saying so: a leader who mediates and then decides teaches the parties to hold back whatever might be used against them.
Mediate when
- The parties have comparable power and the decision is genuinely theirs.
- The problem is interpretation or a pattern between them.
- Both are willing to take part, and the leader has no stake in the outcome.
Decide when
- The question is a priority that only the leader can set.
- Power is unequal, or the rules, safety or someone’s dignity are at stake.
- The leader has a stake, or one party will not take part in good faith.
Say which role you are taking, and why, before anyone speaks. If conduct such as harassment or abuse is involved, neither role fits. It goes through the formal route.
The work
Practising a difficult conversation
- Pick the recurring exchangesTwo or three situations where things regularly go wrong: a weekly handover, a review meeting, an email chain. You get: a short list of exchanges to work on, so that the work does not drift into generalities.
- Compare what each side thinks was agreedWritten separately, then side by side. You get: a list of divergences, and often the first surprise.
- Observe the exchange in its channelWith permission, we sit in on a meeting or read a thread: who asks questions, who interrupts, who summarises. You get: observed patterns, described as patterns of behaviour and not as personality.
- Practise two or three specific movesA read-back, an open question, a clear request (who, what, by when, how will we know), the opening of a difficult conversation. Practised on a real upcoming exchange. You get: moves you have already used once.
- Choose a role for any live conflictMediator or decider, announced in advance, with the reason given. You get: a conflict handled in a way that leaves trust intact, or a decision with its reasons recorded.
How we check that it helped
Counted by the people themselves
- BaselineBefore the work, over a few weeks: how often did two versions of “what we agreed” surface, how often was something redone because of a misunderstanding?
- AfterThe same count over the same length of time, with the same people counting.
- TargetSet together. Not zero: some misunderstanding is the normal price of speed.
- If divergences persist after clear read-backsThe problem is not communication. Switch to a decision about priorities.
How certain are we?
What research says about communication
- EstablishedSenders overestimate how well their message, especially its tone, comes across, particularly in writing.
A robust finding from experiments. How large it is in a particular working relationship is an open question.
- EstablishedInterpersonal behaviour can be described on two axes, roughly dominance against submission and warmth against hostility, and behaviour tends to draw a patterned reply.
This is a descriptive tradition from the study of interpersonal behaviour. It describes patterns; it does not classify people into types.
- Plausible applicationStructured read-backs and open questions reduce misunderstandings at work.
Sensible, widely recommended, with modest direct evidence in organisations. We count rather than assume.
- Our hypothesisWe want to test whether briefly confirming an agreement helps prevent misunderstandings.
We would drop it if our own counts did not show the difference.
What communication cannot resolve
- Personality or style typing. We do not hand out questionnaires or labels. We describe what people do in a given exchange.
- A promise that better listening settles a conflict of interests. It can make the conflict clearer, which is often the useful result.
- Formal mediation. We are not a certified mediation service, and where a formal route is needed we say so and refer on.
- A way of fixing abuse or harassment by talking. Those go through the formal route.