The most expensive sentence in a software project is "that is not what we meant." Every hour of discovery exists to make that sentence impossible, or at least cheap. We have run discovery for two-person startups and for listed companies, and the structure barely changes: one focused week, the right eight people, and four artifacts at the end that make the build quotable.
This post is the full agenda. Feel free to steal it, whether or not you ever work with us.
Before the week: homework, not meetings
Discovery dies when the week starts cold. Two weeks before, we send a short pack: a questionnaire for the sponsor on goals, constraints, and budget reality; a request for any existing material, spreadsheets, process docs, screenshots of the current tools; and read-only access to whatever system is being replaced. We spend two or three days in that material before anyone books a room, so workshop time goes to decisions, not download.
The week, day by day
- 1Day 1: the problem, not the product. No feature talk allowed. We map who suffers, where money leaks, and what "better" measurably means. The output is a one-sentence problem statement and three success metrics with numbers attached.
- 2Day 2: process mapping. We walk the real workflow end to end on a wall, with the operators driving. Every exception, workaround, and "well, usually" gets a sticky note. This day always produces the sentence "wait, is that what you do?" from a manager.
- 3Day 3: the future workflow and the walking skeleton. We redesign the process first, then sketch the thinnest system that runs it. Screens as rough wireframes, data as a whiteboard model, integrations as named arrows.
- 4Day 4: scope surgery. Everything gets sorted into now, next, and later. We fight for the smallest honest "now". This is the day the cut list is born, and it is the most uncomfortable and most valuable day of the week.
- 5Day 5: risks, estimates, and the plan. We threat-model, list the unknowns that could move the estimate, and present the build plan with a price range, not a false single number.
What you walk away with
- A problem statement and success metrics everyone signed, which becomes the test every future feature request must pass.
- A current and future process map, photographed and redrawn digitally, that is worth the week on its own for most operations teams.
- A scoped release plan: the now-next-later backlog with the cut list attached and the reasoning recorded.
- A costed proposal with a range, the assumptions that set the range, and the specific unknowns that would narrow it.
The artifacts are deliberately vendor-neutral. Clients have taken our discovery output to other build partners, and that is fine. A discovery that only makes sense if you hire us is a sales document, not a plan.
Why we charge for it
Discovery is real work, typically priced between R60 000 and R120 000 depending on complexity, and it routinely changes the project. In a recent engagement with Atlas Voyages, discovery cut the imagined scope nearly in half and redirected the first release entirely: the booking-engine rebuild everyone assumed was the project turned out to matter less than the operator console nobody had mentioned in the brief.
5 days
from cold start to costed plan
~40%
average scope change between brief and discovery output
4
artifacts you keep, whoever builds it
Free discovery exists to win builds, and it earns its cost back through optimistic estimates and change orders. Paid discovery exists to find the truth, including the truth that the project should be smaller, later, or bought instead of built. We wrote about that last call in build vs buy.
Running a lighter version yourself
You can compress this to two days for an internal tool: half a day on the problem and metrics, one day on process mapping and the future workflow, half a day on scope surgery. Keep the rules even when it is lightweight: operators in the room, no feature talk on day one, and a written cut list at the end. If you would rather run it with us, the full format is part of how we work.