Estimates, budgets and vendor selection
Money conversations in software break at the same point every time: the client wants a number, the vendor says «it depends», and both sides get frustrated. Both are right. The number does depend, not on abstract complexity but on specific things that can be listed and checked.
Collected here is what we use ourselves: three point estimates instead of a single number, the blocks of work that go missing from cheap quotes, planning from real developer output rather than the 160 hours a calendar month suggests, and the spike as a way to test the main technical risk before the full budget is discussed.
All of it has been tested on our own projects and on quotes clients brought to us for review. In one of those the scope was understated roughly threefold, and even the vendor's pessimistic scenario sat below our trimmed optimistic one. The review took an hour and saved the client months.
FAQ
Where do I start if I only need a ballpark
With the format and the list of external systems. The format sets the order of magnitude, the integrations decide where inside the range you land. Those two inputs are enough for a range without a full discovery.
Why do vendors differ by several times on the same brief
Usually because they are pricing different scopes. One quote includes discovery, infrastructure and stabilisation, another covers screens only. Compare the list of work rather than the totals.
What if a vendor refuses to break the number down
Treat that as a signal in itself. A quote that cannot be taken apart cannot be verified, and it also cannot be trimmed in a meaningful way when the budget is short.