A demonstration on our own sample data establishes very little, because we selected it. An assessment is worth running on your history, over a period in which the operations team already knows which machines caused trouble, so that the comparison has a control.
Two things are being measured. Whether the findings correspond to events already known to the team, which establishes that the models are reading the fleet correctly. And what the findings nobody knew about are worth, which is the product itself.
Nothing is installed to do this. The assessment runs on exported SCADA history, so it can be completed before any decision about deployment is taken.