The documented process is not the real one
Every operation has two versions of itself: the one described in the process document, and the one the team actually performs. The gap between them is where the workarounds live: the second spreadsheet, the WhatsApp group, the step everyone skips because it was designed for a constraint that no longer exists.
Software built against the documented version fits nobody. It is technically correct and operationally wrong, and the team routes around it within a month.
What discovery should cover
- Store operations: how sales, billing and customer interactions happen day to day, and where staff effort goes into repetitive manual tasks.
- Warehouse: how stock physically moves, how it is stored and picked, and where physical reality drifts from the recorded position.
- Intake: how new product enters the system, how it is catalogued and valued, and how intake accuracy affects everything downstream.
- Channels: how online and marketplace selling is managed alongside physical stores, and where the absence of synchronisation creates risk.
Look for repeated data entry
The fastest way to find a broken seam between systems is to count how many times the same fact gets typed. A product that is entered once at intake, again in the storefront and a third time in a reporting sheet is telling you exactly where the integration is missing. Duplicate entry is not just wasted time; it is three opportunities for the three copies to disagree.
Diagnose before prescribing
There is a strong commercial incentive to arrive with a product and find problems that fit it. The discipline worth holding is the opposite: establish what the operation needs, then decide what to build or deploy. Sometimes the answer is a platform. Sometimes it is removing a step.
The output is a roadmap, not a report
A study that ends in observations has not finished. What makes discovery worth the weeks it takes is a sequenced plan, setting out which inefficiency gets addressed first, what has to be true before the next piece can start, and which parts of the operation stay manual for now because automating them would cost more than they leak.
Staged this way, spend follows proof. Each release is justified by something the previous one demonstrated, rather than by a specification written before anyone watched the work happen.