Discovery Phase.

Discovery is about understanding, not about describing features. Typical components are conversations with the people who perform the workflow every day, a recording of the current process including its detours and paper workarounds, a review of existing data and systems, and the collection of constraints: interfaces, data protection under the revised Swiss FADP, retention obligations, availability of internal contacts. The important distinction is between what people say and what people do. Observation at the workplace regularly yields different findings from the meeting held to discuss it.

What comes out at the end is not slides but working material: a sketch of the target workflow, a list of use cases prioritised by value and risk, a list of open technical questions with a proposal for how each will be tested, and a rough cost range with the assumptions behind it stated openly. A preliminary make-or-buy check belongs there too. If a product on the market already covers eighty per cent of the need, the remaining twenty per cent has to justify building anything at all. That check belongs at the start, because later on nobody asks it with an open mind.

An example: a trading company wants a customer portal because the telephone is eating too much time. Discovery shows that a large share of the calls concerns a single question — delivery status — and that the information already sits in the ERP system. Instead of a portal, an automatic status email is built first, followed by a lean status page. The portal remains an option for later. Discovery did not shrink the project here; it exposed the actual problem underneath the requested solution.

Honesty also means keeping the proportions right. A discovery that runs for months and ends in a specification nobody reads has become the problem itself. For an SME project, one to three weeks of concentrated work is usually enough. It also does not replace testing: discovery reduces uncertainty about the problem, not about the solution. That is why it ideally ends with one clearly cut next step — a prototype, a proof of concept, or the reasoned recommendation to buy a standard product instead.