Where the thirty-five weeks actually go

The industry average for deploying an industrial robot cell is roughly thirty-five weeks. Best-in-class operators do it in two. That gap is the most interesting number in industrial automation, because almost none of it is explained by hardware.

Arms don't take thirty-five weeks to build. Vacuum tooling doesn't take thirty-five weeks to specify. What takes thirty-five weeks is four organizations (the plant, the integrator, the component manufacturer, and whoever is funding the asset) exchanging incomplete information over email, in series, with no shared record of what was agreed.

Watch a single deal closely and the time breaks down into four buckets, none of which appear on a project plan.

Serial question rounds

A specification is assembled by asking the customer things. In practice this happens in rounds: the integrator asks, the customer answers partially, the answer raises a new question, and the cycle repeats. Five rounds across ten days is entirely normal. Each round is a day of work and several days of waiting, and the waiting dominates.

Dead air

Nobody owns the follow-up. A fifteen-day silence between parties is not a rare failure; it is the default behaviour of a process whose system of record is four separate inboxes. The customer assumes the integrator is working. The integrator assumes the customer is deciding. Nothing is happening.

Sequential tracks

The capital package is typically assembled only after the quote lands, on the reasonable-sounding logic that you cannot finance a number you don't have. But financing partners will state plainly that assumptions are acceptable when end-customer figures aren't yet available. Waiting for the quote before starting the capital track costs about ten days for no informational gain.

Custom engineering

Historically around half the cost of a deployment was design and programming work performed fresh for each project. On one recently benchmarked cell, engineering was $36,925 of a $140,000 quote (twenty-six percent) against assembly labour of $7,500. Design, programming, and test dwarfed the physical build.

None of those four buckets is manufacturing. All four are coordination latency, and coordination latency is a software problem.

The fixes are unglamorous. Run the supply track and the capital track in parallel from the moment the specification is approved, with assumed values clearly labelled as assumed. Detect dead air per counterparty and chase it automatically. Batch questions into one round instead of five, and drop the questions that don't apply to this kind of customer at all. Build cell templates so the second deployment of a pattern carries a fraction of the engineering of the first.

Do that consistently and thirty-five weeks becomes four. Not because anything was built faster, but because the waiting stopped.