Strip away the vendor decks and every automation business case is the same model: cost divided by annual savings equals payback in years. The entire audit skill is knowing which lines get optimistic thumbs pressed on them.

The six lines

  1. Loaded labor baseline. The honest denominator: wage + benefits + turnover cost + absenteeism + overtime premium, per hour, times the hours the operation actually runs. Vendors like to use your peak wage; you should use your blended reality.
  2. Labor displaced (not eliminated). A 40% productivity gain does not remove 40% of headcount; it removes the marginal hours at the constraint, and only if volume holds. The model should book avoided hiring and overtime first — those are real cash.
  3. Throughput uplift (vendor-stated). The picks/hr claim. Treat it as the ceiling achieved in the reference site, then haircut for your SKU profile, your congestion, and your first-year learning curve. 70-85% of spec in year one is a defensible planning figure.
  4. All-in capex. Equipment is the visible number. Integration, software, WMS work, electrical, floor repair, network, spares, and contingency routinely add 25-40%. A quote without integration is not a price.
  5. Opex reality. Maintenance contracts (8-12% of equipment per year), software subscriptions, and the headcount you ADD (a robot wrangler is a real role).
  6. Utilization. Payback at one shift is double payback at two shifts. The single biggest lever in the whole model is hours-per-week the system actually runs.

The thumb spots

Optimism hides in lines 2, 3, and 6: headcount treated as fully eliminated, spec throughput treated as day-one reality, and utilization modeled at the shift pattern you aspire to rather than the one you run. A model that survives a 20% haircut on all three and still pays back inside your window is a real project. One that doesn't was a brochure.

The operator's list

Demand the vendor's model in a spreadsheet, not a PDF. Re-run it with your blended labor rate, 75% of claimed throughput, and your current shift pattern. If the vendor resists the haircut, that is data.

Independent automation industry analysis — not investment or procurement advice. Capex decisions warrant site-specific engineering diligence.