Choose one process to demonstrate
Three minutes rarely allow you to explain an entire technical product. Choose a process that makes its value visible. For software, that might be turning an input into a usable report. For a component, it could be one clear difference in assembly. Let the most important question your visitors bring determine the choice.
Write one sentence explaining the starting situation. Then identify what changes during the demonstration and how the audience will recognise the result. Features that do not contribute to that sequence belong in a more detailed conversation afterwards, rather than being squeezed into the same short performance.
Rehearse the real sequence
Time a complete run, including the greeting, equipment startup and reset. The transitions between steps often take longer than the feature itself. If a process loads slowly, show a prepared intermediate stage and explain honestly where it came from. Do not make the demonstration appear faster through a misleading comparison.
Ask someone unfamiliar with the subject to interrupt with a question. Can you continue calmly afterwards? Decide which step may be omitted and which is essential. This turns a rigid script into a sequence that can survive a normal conversation, rather than requiring the audience to remain silent.
Connect the result to the next question
Leave the result visible at the end instead of immediately returning to the start screen. Ask whether this process occurs in the visitor's own work. That can lead to a useful extension: different input data, a larger scale or a technical question. Avoid promising performance that the demonstration cannot establish.
Prepare the following run as well. Give test files clear names, remove personal data and create a reliable starting state. A brief reset between conversations prevents the next visitor from seeing the previous person's results or arriving halfway through an unfamiliar workflow. Practise that reset with everyone who will operate the stand.
