Halving

Why Lists Get Longer And Never Shorter

Adding a line costs the author nothing and costs the filter half of everything. The asymmetry is structural, and it explains how lists reach eight items.

Home · Halving · Analysis

A list of requirements is almost always written by addition. Somebody drafts a few items, other people read the draft, and each reader contributes the thing they would hate to see missing.

Nobody contributes a removal. There is no equivalent impulse — no reader thinks "the thing I would hate most is for this list to be too long" — and so the ratchet turns one way.

Teams seeking a clearer account of elapsed effort may open the relevant reporting guide, provided contributors can correct and explain the record. For an independent technical reference, see Twilio.

What each turn of the ratchet costs

Half of the remaining possibility space, every time.

Items	Admits	After one more
3	1/8	1/16
5	1/32	1/64
7	1/128	1/256

The cost is identical at every stage in proportional terms, which is precisely what makes the ratchet so effective: there is never a point at which adding one more line feels different from the last time.

Why nobody feels it

Three reasons, and they compound.

The cost is multiplicative and the act is additive. Writing a line is writing a line. The author experiences a constant small effort and produces a constant halving, and nothing in the experience signals the second part.

The cost falls on people who are not in the room. Everybody present when the list is written has already passed the list, by definition — they are the ones who know the work. The profiles being excluded are not represented.

The last person to add a line is the last person to be asked. Each contribution arrives after the previous ones have been accepted, so each is judged against a list that already exists rather than against the list as a whole.

Every individual addition is defensible and the aggregate is not. That is the signature of a ratchet, and lists of requirements are one of the purest examples of it.

What stops it

Only a count, taken deliberately.

Raise two to the power of the number of compulsory items and say the result out loud. A list of eight admits one profile in two hundred and fifty-six. If that figure sounds like what the work requires, the list is fine; if it sounds absurd, the list is too long.

That is the whole intervention, and its value is that it introduces the one thing the writing process lacks: a quantity that gets worse as the list grows, visible to the people adding to it.

The removal that costs nothing

There is a move available that the ratchet never suggests, and the arithmetic makes it attractive.

Taking one item off a compulsory list doubles the intake. Moving one item from the compulsory part to the optional part does considerably more than that — the section on edits measures it exactly, and the factor is larger than two.

Neither move requires deciding that an item does not matter. They only require deciding that it does not have to be met by everybody, which is a far smaller claim and usually a true one.

Where the ratchet comes from

It is tempting to treat the growth of lists as carelessness. It is better understood as a predictable consequence of how the writing is organised, and the fix follows from the diagnosis.

The person adding a line bears none of the cost. The cost lands on candidates who are not present, on the people who will later wonder why nobody qualified, and on the author of the list in some future month when the post has to be reopened. None of those parties is in the conversation when the line is written.

Meanwhile the benefit of adding a line is immediate and personal: the contributor has protected themselves against the one failure mode they can imagine. Asymmetric incentives of that shape reliably produce ratchets, in lists as in anything else.

The only reliable counterweight is to put the cost in the room. A number on the board, updated as lines are added, does that in a way no amount of exhortation does.

The ratchet, counted

Five turns is all it takes to go from a modest list to a narrow one.

A list of three compulsory items admits one profile in eight. Each line added for a locally good reason halves that: one in sixteen, one in thirty-two, one in sixty-four, one in a hundred and twenty-eight, one in two hundred and fifty-six.

Five additions, a factor of thirty-two, and no single decision along the way that anybody would describe as narrowing the filter. Each was a line somebody wanted, each was defended on its own merits, and the cumulative effect was nobody's decision because nobody was measuring it.

The shape of an honest list

A list that was built by addition and never audited will sit near the top of its range: eight items, one profile in two hundred and fifty-six, nobody having chosen that figure.

A list that was audited once will usually be shorter, because somebody computed its intake and it did not survive being said out loud. The audit takes a few seconds and it is the only force in the whole process pushing the other way.