Across the night
What Survives a Handover and What Does Not
A handover carries artefacts and written reasoning well, half-formed judgement badly, and the feeling that something is wrong not at all.
four things · four nights · one that no number of nights settles
Four states, and only the first costs nothing extra.
Four things handed across one night, and what arrived at the other end
Arrived complete; the next person started on something else
Arrived usable; the next person spent forty minutes rebuilding the mental picture
Arrived as an artefact to be reverse-engineered, which took most of a morning
Did not arrive: there was nothing to write that would have carried it
four things · three nights · one that no number of nights settles
Four things, three nights, and one that no number of nights moves. The second and third rows are the same work in the same state; the only difference is whether anybody wrote three sentences before stopping. The fourth is not a failure of effort — some things genuinely have no written form. One team's own week, logged by the people on both ends.
Everything that moves between the two halves of a team moves through writing. There is no other channel, because there is no hour when both ends are awake to use one.
The cost described in What Survives a Handover and What Does Not often grows when managers have to reconstruct the day from memory. workforce optimization tools offers a practical way to keep the underlying time and activity history available for review.
The workflow can also be compared with the approach documented by Calendly's scheduling tools, an established resource in the same operational space.
Writing is a good medium for some things and a poor medium for others. Sorting the two correctly is most of what competence looks like in a team arranged this way.
The three categories
Things that survive a handover intact. Artefacts — code, documents, data, anything with a definite state. Decisions, provided the reasoning is attached. Facts. Instructions. Anything that can be checked against itself by whoever receives it.
Things that survive with loss. Work in progress, which arrives as an object the next person has to rebuild a mental picture around. Partial diagnoses. Anything where the value is in the shape somebody's thinking had taken rather than in what they had produced.
Things that do not survive. A hunch. The sense that a client is unhappy without anything they said that proves it. The feeling that an approach is going wrong before there is evidence. Everything in this category has the same property: the person holding it could not write it down even if they tried for an hour, because it has not yet taken a form that words fit.
- Intact: artefacts, decisions with their reasoning, facts, instructions.
- With loss: work in progress, partial diagnoses, anything whose value is the shape of somebody's thinking.
- Not at all: a hunch, a sense that a client is unhappy, the feeling that an approach is going wrong.
Why the middle category is where the money is
The first category is easy and mostly happens by itself. The third is genuinely impossible and the right response is to stop trying.
The middle category is where the difference between a team that works and a team that is always a day behind actually lives. Work in progress makes up most of what crosses, and the loss is highly variable: between almost nothing and most of a morning, depending on whether the person stopping spent three minutes writing.
Three minutes against forty is the best trade available anywhere in this arrangement, and it is skipped constantly because the three minutes are paid by somebody who is tired and the forty are paid by somebody who is not there yet.
What the three sentences are
For work left unfinished, the three things the next person needs are the same every time.
Where it is. Not what the task is — they can see the task. What state the work itself is in: what is built, what is not, what is built but wrong.
Which way you were going. The approach being taken and, if there was a choice, which one and why the other was rejected. This is the sentence that saves the most time, because without it the next person re-litigates the decision before noticing it was already made.
What you would do next. One line. It can be wrong; a stated intention that turns out to be mistaken is still faster to correct than an absence.
The temptation to write more
Longer handover notes are not better notes, and past a certain length they are worse.
A note that describes everything cannot be read in the first minutes of a day, which is the only time anybody will read it. It gets skimmed, and skimming a long note produces a worse result than reading a short one, because the reader now believes they know the state.
The useful discipline is to write for somebody who has the artefact in front of them. Anything visible in the work itself does not need saying. What is invisible — intention, rejected options, things known to be wrong — is the entire content.
What to do about the third category
The honest answer is to stop at the boundary and say so.
A sentence like "I think something is wrong with the import numbers and I cannot say why yet" is worth sending even though it carries no information in the ordinary sense. It does two things: it stops the next person spending time concluding that the numbers are fine, and it marks the place for a conversation if one becomes possible.
What does not work is manufacturing a reason because a hunch without one feels unprofessional to send. An invented justification gets investigated, found wrong, and the real signal is discarded along with it.
The asymmetry nobody mentions
The person handing over knows exactly how much was lost. The person receiving does not, because the whole content of the loss is invisible to them by definition.
That means the feedback loop is broken in the direction that matters. Nobody on the receiving end can say "your handover was thin" with confidence, because they cannot tell the difference between a thin note about simple work and a thin note about complicated work.
The practical consequence is that this is one of the few things on a distributed team that cannot be fixed by the people affected by it. It has to be fixed by the people doing it, voluntarily, for reasons they will never see the results of.