We know exactly what we're betting on, and how we de-risk each one.
Our entire sizing hangs on €100/asset/year, an assumption, not a market-validated fact. We don't yet know if a client pays €100 for a motor they ignore today, the real one-off/recurring split, or our true cost per asset.|| A paid pilot that validates price and real cost per asset before scaling.
Our biggest promise is also the hardest to prove. We don't yet know our false-positive/false-negative rate in real production, or how many weeks ahead we reliably predict. A false positive erodes trust; a false negative kills the contract.|| Validation against real failures in the pilot, measuring the lift from plant recalibration.
"15 minutes, no wiring, €100" is clean on one machine. We don't know how the Zigbee mesh behaves with hundreds of nodes in a real industrial environment, interference, range, metal, latency.|| Measure a full plant deployment, not a single machine.
We assume land-and-expand where the user pilots 10 motors and the operator signs the expansion. We don't know the true cycle length in heavy industry, who holds the budget, or if "don't touch the working DCS" slows adoption.|| Turn the Guerin validation into a letter of intent or first pilot with a real operator.
The motor TAM (380M) is anchored in data. But the SAM and the pump/compressor/transformer populations of Circle 2 are order-of-magnitude estimates, with no clean public census.|| Refine bottom-up per vertical as we close the first real clients.
We assume the moat is data + traceability + cheap coverage. We don't know how long an Augury or Siemens would take to launch a low-cost sensor and drop into our niche if we validate it first.|| Accumulate failure-fingerprint volume fast first and furthest wins.
The assumption we bet the most on: that the client pays €100 to see an asset they ignore today, and that our prediction of the never-seen is reliable in the field. If those two are true, the rest is execution.
