Steganox Engineering · 18 June 2026
The estimate you get for free is the one you cannot trust
Every agency has been asked the same question on a first call: roughly what would this cost? It is a fair question, and the honest answer is uncomfortable. At that point nobody knows. The requirements have been described in a paragraph, the existing systems have not been opened, and the edge cases that will consume forty percent of the effort have not surfaced yet.
What usually happens next is that a number gets given anyway, because the vendor who refuses to answer loses the deal to the one who does. That number then becomes an anchor. Six months later everyone is arguing about scope creep, when what actually happened is that a guess was treated as a commitment.
Discovery is the estimate
We separate the two questions. The first is: what would it take to understand this properly? That is a bounded piece of work, usually one to two weeks, and it produces an architecture direction, a scoped backlog and a costed plan. We do not bill for it.
The second question, what will the build cost, gets answered at the end of that. By then it is not a guess. We have opened the legacy database, seen how many integrations there really are, and found the compliance requirement that nobody mentioned because it felt obvious to them.
The order is the whole point. An estimate produced before anyone has looked at the system is a sales artefact wearing a number. An estimate produced after is a position you can hold someone to. If a vendor will quote your project on a paragraph and a phone call, what they have told you is how they will handle every other unknown on it.
What makes an estimate hold
Three things, in our experience. First, the estimate has to come from the people who will do the work, not from a salesperson applying a multiplier. Second, it has to be decomposed. A single number for a six-month project tells you nothing, whereas thirty numbers show where the risk is concentrated. Third, it has to carry explicit assumptions, written down, so that when one turns out to be false everyone can see immediately which part of the number moved and why.
The last point matters more than the accuracy of the original figure. Projects rarely fail because an estimate was twenty percent light. They fail because the estimate was wrong in a way nobody could see until the deadline arrived.