Nobody wants a range

I started quoting my own work in 2014, on the side, for small clients, and I have been quoting or scoping something ever since. Different vocabulary in different decades. Same failure, and I did not see the mechanism for a very long time.

The failure is this. You give a range, because a range is the truth. Some number of days, or some number of dollars, between here and there. The person you gave it to writes down the first number. Not out of bad faith. Because a plan cannot hold two numbers, and something downstream of them needs one.

Ranges do not survive being repeated

Watch what happens to the sentence after it leaves.

You say between four and seven. The person you told says roughly four to seven when they repeat it internally. The person who hears that puts four in the plan, because a plan wants a figure and the low one is the figure that fits. Two steps out, the range is gone and the number is four, and it now belongs to someone who never heard you say seven and would be surprised to learn you had.

Then you deliver at six, which was inside the range you gave, which was honest, and you are two over. The overrun is real to everybody who is looking at it and it does not exist in the conversation you actually had.

There is a second version that is worse. The high end is not merely dropped, it is remembered as padding. Once somebody believes the top of a range is what you say to protect yourself, every future range is read as a low number with an excuse attached, and the honest estimator has been converted into a person who is thought to inflate. That reputation is very difficult to shed, because disproving it requires being right at the high end, and being right at the high end looks exactly like the thing you are accused of.

So the person who quotes a range and the person who quotes a single optimistic number get the same treatment. The range is not doing what it was supposed to do. It is transmitting its low end and its author’s caution and nothing else.

What I tried, both of which were worse

For a few years I quoted the top. It is defensible, you say prudence out loud, and it means the number you gave is the number you can hit. It also loses work, and more importantly it is a quiet dishonesty: I was telling people something I did not believe was the likely outcome, and dressing that up as conservatism. The clients who accepted it paid for my discomfort with uncertainty.

Then I moved to the single number with a buffer folded in, which is what most people actually do. That is worse in a specific way. It removes the uncertainty from the only person who could act on it. If a piece of work might take four days or seven, that difference is often the whole decision about whether to do it at all, and burying it inside one confident figure means the decision gets made without the information. The estimate looks better and the choice gets worse.

The only thing that has helped

Not a narrower range. A range attached to what would move it.

Between four and seven, and here is what determines which: whether the source data is what we were told it is, whether one particular approval takes a day or a fortnight, whether the two systems agree in the first pass. Now the second number has a reason to exist. It is not caution, it is a named condition, and the person receiving it can go and resolve the condition, which is a thing they can actually do.

I want to be careful about how much I claim for this. The conditions get stripped in transmission too. They are longer than the numbers and they travel worse. What they give you is not protection, it is a sentence you both said out loud in the same room, which turns the later conversation from a dispute about someone’s competence into a check of whether the condition held.

That is a smaller benefit than I would like. It is the only one I have found that does not require somebody to be dishonest.