The question before the conclusion.
- Lens
- industry
- Journey
- 5 decision points
- Reading
- 7 min
The quotation is on the table. Conveyors, a sorter, an assisted picking station. The chief executive looks at the amount, then at us: ‘what do you think?’
It is a trick question for a software vendor. Answering ‘yes, and you will need a system to control all of that’ is commercially comfortable. More often, we answer: first show me your last six months of volumes, day by day.
Because a machine does not debate. It performs what it was asked to do, at the rate for which it was sized, indefinitely. That is its strength and precisely the problem: automation locks an organisation in place. If the organisation being fixed is not the right one, you have just paid dearly for the loss of your ability to change your mind.
The real criterion is not volume
Automation is generally assumed to become profitable above a certain volume. It is an approximation that has led to many poor investments.
The decisive criterion is stability—of flows, products, packaging and ordering patterns.
A warehouse handling high volumes but whose range changes every season, whose customers change their packaging requirements, and whose orders are gradually shifting from pallets to parcels should not automate. It must first understand where its activity is stable and where it is changing.
Conversely, a lower-volume warehouse where the same products have left in the same packaging for the same destinations for three years can automate with limited risk.
Ask yourself this question: if I freeze my current organisation for eight years, what will I lack? If you do not have a clear answer, you do not yet understand your variability well enough.
The five signals that say ‘not yet’
Your inventory discrepancies are substantial. A machine acts on the data it receives. A sorter given an order for inventory that does not exist creates a mechanical error, faster and more expensive than a human error—because an operator would have seen that the bin was empty.
Your procedures are not written down. Automation requires rules to be fixed at the level of detail of an industrial controller. If your rules live in team leaders’ heads, they will be invented during installation by an integrator who does not know your business.
Your seasonality is strong. An installation is sized either for the peak—and lies idle for the rest of the year—or for the average—and reaches capacity just when you need it most. Neither is satisfactory. There are answers, such as sizing for the high average and designing a manual overflow mode, but they must be designed from the start, not discovered in December.
Your product range changes a great deal. Every new item must find a place in a physical system designed for given dimensions and weights. An out-of-gauge product in an automated installation is not ‘an exception’; it is an entire parallel flow to organise.
You have never measured your times. This is the most revealing signal. An automation decision is justified by time saved against a known baseline. If you do not know how long picking takes today, you can neither justify the investment nor demonstrate its result.
What to do first, at one hundredth of the cost
Review the layout. Put the fastest-moving items closest to the operator and at easy picking height. It is free, reversible, and immediately reduces travel distance.
Batch picking. Use one route to serve several orders rather than one.
Make replenishment of picking locations reliable. A shortage in picking costs more than planned replenishment.
Remove inherited duplicate checks. Many warehouses check twice out of historical caution without ever establishing whether the second check catches anything.
Equip teams with mobile tools. Moving from paper to a device removes rekeying and produces usable time data for the first time.
That last point is worth emphasising: mobility produces the data needed to make an automation decision. That is the logical order. Measure first, then invest.
What we learned the hard way
We supported a go-live where the mechanised installation and the system were commissioned in the same week. The intention was sound: one operational stoppage, one disruption.
It was a mistake. When the first problems appeared—and they always do—no one knew whether they came from the machine, the configuration or the migrated data. Every incident required a three-way investigation. The teams spent six weeks in a fog that could have been avoided by separating the two go-lives by two or three months.
How to frame the decision honestly
Three figures are enough, and you can obtain them without a consultant:
Your average picking time today, measured rather than estimated. Observe it over two weeks, with a stopwatch if necessary.
The distribution of your workload—the ratio between your busiest day and your median day over twelve months. This ratio will show whether an installation sized for the average can cope.
The share of your activity that is genuinely stable—the items and packaging that have been unchanged for two years.
If the third figure is low, no financial projection will make the project sensible: you will automate an organisation that will no longer exist in eighteen months.
And when is the right time?
There is a right time, and that must also be said. A warehouse with stable flows, reliable data and written procedures, where picking capacity has become the limiting factor in growth—that warehouse is right to automate and will derive lasting benefit from it.
The clearest sign is that there are no organisational gains left to find. You have already reviewed the layout, batched picks and equipped the teams, and you are hitting a physical ceiling. That is when mechanisation takes over.
