What a two-to-four-week hall pilot actually costs to learn
The steering committee wants a plant-wide twin return from a slide that borrowed someone else's downtime percentage. Nobody in the room can price one starved hour, or one empty forklift day, on this hall.
Borrowed percentages fail before the floor does
Finance is right to reject that slide. Public economics work on digital twins, including the manufacturing investment framing from NIST, treats the buy as a decision under uncertainty. Costs and benefits have to be characterized for the system you actually run: your downtime hours, your maintenance spend, your margin on a lost hour, and a scope that maps to a real invoice. An outside reduction percentage is a hypothesis until those local numbers exist. It is not a forecast.
Credible return practice says the same thing in plainer language. Start from a downtime log you already trust and from the contribution of an hour on the constraint. Then ask what a specific intervention on a specific asset could change. A plant-wide ROI that cannot name this site's cost of one starved hour has not started the analysis. It has started a catalog. The business question has to be narrow enough to measure, which is the gate under clarify the business question first.
What a thin pilot actually spends
A useful learn window on an operational digital twin is typically two to four weeks, one zone, a defined set of assets, and success metrics that map back to one question. That is a different object from a one-to-three-month go-live on a modest whole warehouse. The short window is there to prove the picture is true and that someone will act. It is not a promise to instrument every aisle before anyone has agreed what 'pass' means.
The money in that window is small on purpose, and it is not only a license line. twinzo prices by data points rather than named-user seats, so a narrow zone stays a narrow commercial footprint. Current commercial detail sits on pricing. Beside the license, budget the hours of the people who must judge the picture: the shift lead who owns the waste, an industrial engineer or logistics planner for the baseline, and whoever connects the feed you already have. If those hours are not named, the pilot looks cheap and then dies in someone's spare time.
A documented 15,000 m² (about 161,000 ft²) logistics center is the entry-scale anchor for a small hall, not a template for a campus. That case invested €3,500 (about $3,850) in logistics visibility with a real-time location system (RTLS) and recorded €11,000 (about $12,100) of year-one operating cost saved, from live position and more even shift balance. The full story, including what it did not claim, sits under what the smallest documented RTLS case actually looked like. Do not paste a larger plant's return onto a two-week zone.
What those weeks must prove
Pass is one waste, one owner, and one Check that was blind before the pilot. Examples that survive a finance review are concrete. Can the lead see which trucks were busy at the same time before asking for another lease. Can the same person see a recoverable material wait before Saturday is penciled in. Can a KPI the plant already manages move in a way supervisors trust, on that zone only. A prettier map with no named owner is a failed learn, even if the positions are accurate.
Digital lean fails the same way when the tech catalog arrives before the waste has a name. Ownership is the prior decision under digital lean needs a waste owner before the tech catalog. The pilot should also show that the hall people recognize matches the shift they just worked: historical replay, a spaghetti diagram for the zone, and dwell where the argument actually is. If the picture and the floor disagree, fix that before any plant-wide conversation. That disagreement is a cheap finding. Discovering it after a campus purchase is not.
What a wrong plant-wide buy still costs
The license is the smallest bill when the scope was wrong. Integration work, extra feeds nobody uses, and a year of meetings to explain a model the shift does not open are the carrying cost. The harder cost is cynicism. The next honest proposal, even a thin one, inherits the failed slide. Sponsors remember the borrowed percentage, not the zone that might have taught them their own hour rate.
A plant-wide operational twin can be the right later buy when the zone already changed a decision. It is a bad first buy when the baseline is still a vendor chart. NIST-style investment analysis is sequential for that reason: characterize the local loss, test a bounded intervention, then scale what held. Rushing the scale relocates the uncertainty into a larger invoice. Capability once the question is clear sits on the features overview. Twin-type economics for static, simulation, or multi-hall programs are a different ledger and should stay on those pages.
How to write pass and fail before anything ships
1. Name the loss in hours or money you already have - One starved hour, one empty day, one Saturday crew. If you cannot price it roughly, you are not ready to buy a percentage.
2. Name one owner and one zone - The person who will change a roster, a route, or a call rule if the picture is true. A steering committee is not an owner.
3. Write the pass line - What Check becomes possible in two to four weeks that is impossible on today's screens. Tie it to one KPI the shift already believes.
4. Write the kill line - If the zone picture does not match the hall, or nobody acts on a confirmed loss, stop. Do not widen the buy to hide a failed learn.
5. Keep the commercial footprint as narrow as the question - Data points for that zone, not a plant-wide seat count. How a rollout is priced stays on pricing.
That sheet is the learn cost. It is also the only ROI story this plant can defend until the zone has numbers of its own. Get in touch if you want to size a two-to-four-week zone against one loss you can already name, before a plant-wide number hits the committee.