Everyone knows to cut scope at the start. You take the list, you argue about it, you remove the things that are obviously version two, and you begin.
That cut is easy because nothing has been built yet. It costs nobody anything and it is almost never enough.
The second cut
The one that matters comes about a third of the way in, when you know things you could not have known at the start. Which parts are harder than they looked. Which feature the client has stopped mentioning. Which piece of the system everyone actually opens every morning.
That is the moment to cut again, and almost nobody does it, because by then there is momentum, a plan on a page, and a conversation nobody wants to have.
The conversation is not difficult if you have set it up. On a fixed-scope engagement I say at the start that we will revisit the scope in week two or three, with what we have learned, and that the goal is to ship the useful part rather than all of it. Framed as a scheduled checkpoint, it is a sign of control. Sprung in week five, it sounds like failure.
What it looks like in practice
Recent example. A client site was getting spam through its enquiry forms, and the fix I scoped was a six-step filtering pipeline: challenge at the form, then a series of checks and routing rules behind it before anything reached the CRM or the notification inbox.
Partway in, it became clear the first step was doing most of the work. The challenge at source was catching the overwhelming majority, and steps three through six existed to catch a residue we had not measured.
So I proposed cutting them. Not deferring them vaguely, cutting them, on the basis that we would look again with real numbers. What shipped was the challenge at source plus one routing change we kept because it solved a different problem entirely: separating newsletter sign-ups from genuine enquiries, which had been muddying the inbox for months.
The client agreed, and then improved on my proposal. Rather than accept "we will revisit in a month", they asked for a firm diarised date. That was the right instinct, and it is the version I now use myself.
The pattern I see repeatedly: the feature that survives every cut is rarely the one that was most argued about at the start. It is usually something small and structural that everything else turned out to depend on.
What it costs
Cutting is not free, and pretending it is would be dishonest.
In the case above, some spam is still getting through.
Not a lot, and not enough to change the decision, but the honest position is that the cut had a cost and the client is living with it while we wait for the review date. If I had built the full pipeline, that residue would probably be gone.
What I got wrong was not the cut. It was moving on without instrumenting what I had cut. If you remove four steps on the basis that the remaining ones are sufficient, you owe yourself a measurement that proves it, and I did not put one in place before the client asked for the diarised date.
The rule I use now
If I cannot name what comes off the list at the second checkpoint, I have not looked hard enough. There is always something, and finding it is cheaper in week three than discovering it in week seven.
Fixed scope does not mean fixed list. It means a fixed budget and an honest conversation about what goes inside it.