Estimating a project properly takes real work — understanding the current systems, the constraints, the integrations, the people. Firms routinely do this unpaid, in the hope of winning the build. The cost is absorbed, the estimates are rushed because the work is unfunded, and the resulting number is the one everyone is held to.
A paid discovery phase reverses this. The client buys a short, defined engagement that produces a scope, an architecture, a plan and an estimate. It is priced modestly relative to the build, has its own deliverables, and ends with a decision the client is free to make either way.
The objection is that clients will not pay for it. Some will not, and that is useful information. A client unwilling to fund a small piece of work to define a large one is telling you something about how the larger engagement will go. The clients who do agree are self-selecting for seriousness.
Make the deliverable genuinely valuable independent of the build. A scope document, integration inventory, risk register and estimate should be useful to the client even if they take it to another supplier. That has to be true and it has to be said, because it is what makes the phase feel like a purchase rather than a toll.
Keep it short. Two to four weeks, with a fixed price and a defined end. Discovery that drifts becomes a project without a build, and clients notice.
The commercial effect is that estimates improve. An estimate produced with access to the real systems, by people who spent time with the client's team, is a different quality of number from one produced from a requirements email. Fewer overruns follow, and overruns are where services firms lose the margin they thought they had.
