The automation business case: what has to be true before the numbers mean anything

Stabilise the process, then mechanise it

Automation freezes a method. If the method is inconsistent, the equipment reproduces the inconsistency faster and with fewer people around to notice. Teams reach for a robot because a station is troublesome, when the trouble is variable incoming material, an unclear standard, or a fixture that lets the part sit two ways. Fix that first and two things happen: sometimes the problem disappears and the capital is not needed, and where it does not, the specification becomes writable because somebody finally knows what good looks like. Projects that skip this step spend their commissioning period discovering process variation the business had been absorbing manually for years without ever documenting it.

The shape of the demand matters more than its size

A dedicated cell wants long runs of similar work. What undermines it is not low volume but choppiness: many variants, short batches, and a schedule that changes weekly. Look at run length and changeover frequency rather than the annual total, because a large yearly figure delivered in dozens of small mixed batches gives a machine very little productive time. Also test the downside: if the largest customer for this part left, would the equipment still be loaded, or would it become an expensive fixed cost on a line running below capacity? Automation converts variable labour cost into fixed asset cost, and that trade is only attractive when the load is dependable.

Incoming part consistency is where budgets actually go

Quotations price the robot and the controller. Overruns come from part presentation. Castings with variable flash, mouldings that shrink differently between cavities, film reels that vary in tension, components delivered loose in a bin rather than in trays — each forces bowl feeders, vision, compliant grippers or an operator loading the machine by hand, which quietly removes the saving. Before signing, take a genuinely representative sample, including material from every supplier and cavity, and prove the handling method on the worst of it. A cell proven on hand-picked golden samples will fail its first real production week and everyone will blame the integrator.

Payback horizon against product life and contract length

Purpose-built automation is only as valuable as the product it was built for. Set the payback window against the shorter of remaining product life, customer contract term and the interval before a design change makes the tooling obsolete. Automotive and appliance programmes have defined end dates; consumer products can be redesigned with no warning at all. Ask the design team what is in the pipeline before committing, because engineering changes that seem minor to them — a rib added, a fastener moved, a material switched — can require the fixtures to be remade. Where product life is genuinely uncertain, favour reconfigurable equipment and accept the lower peak rate.

The costs the spreadsheet leaves out

Beyond the quoted price sit installation, services, guarding, safety assessment, validation, training, spare parts holding, software support agreements, and the production lost during ramp-up while the cell is learning to run. There is also a skills cost: the plant now needs somebody who can fault-find control systems, and in many regions those people are hard to hire and easy to lose. On the other side of the ledger, count what is genuinely gained beyond labour — consistency, traceable records, the ability to staff a shift you currently cannot fill, and reduced exposure to repetitive-strain injury on a station that has been generating claims.

Frequently asked questions

Can automation be justified on labour saving alone?

Sometimes, but a case resting only on removed headcount is fragile, because the saving rarely materialises in full. Somebody still loads, clears faults, changes over and inspects. Cases that survive scrutiny usually combine several effects: steadier quality, output that no longer depends on filling a shift nobody wants to work, capacity released on a constrained operation, and reduced injury exposure. If labour is the only benefit you can name, examine whether the station is genuinely the constraint or merely the most visible cost.

How do we tell whether our process is stable enough to automate?

Run the manual station under observation and record every deviation: how often the part needs coaxing, how often the operator adjusts something, how often material is rejected before it reaches the operation, and how much the cycle varies between people and shifts. If the answer is that experienced staff constantly compensate for something, that compensation is the specification you have not written. A machine will not improvise it. Automate once the station runs the same way regardless of who is on it.

Should we automate one station or convert the whole line at once?

Starting with one station is usually the cheaper education, provided you pick a station that is genuinely constraining and not simply the easiest. A single cell teaches the plant what its parts and utilities are really like, builds the maintenance skill, and exposes the documentation gaps at a survivable scale. Whole-line conversions concentrate risk into one commissioning period and leave nowhere to fall back to. The exception is when the operations are so interdependent that automating one merely moves the queue somewhere else.