Every second forwarder I talk to wants to automate right away. Everything, preferably yesterday. In one of our first projects it was supposed to be order confirmation in a customer portal. Three steps, they said: open the portal, check the orders, confirm. Counted, it was 23. Three systems, two Excel lists, a sticky note on the monitor with the rule that customer C is never confirmed on Tuesdays. Had we automated at once, we would have built three of 23 steps.
Since then we follow four stages in a fixed order. Here they are, with the rule for which process comes first.
Strategy& and BVL surveyed 115 logistics providers in November 2025. 96 percent have started digitalization, 10 percent scale. Half name missing priorities as the obstacle. A quarter do not measure results at all.
That matches what we see when we count. The managing director estimates an hour and a half a day for order capture. The clerk says three. Timed: three and a half. Whoever does not count prioritizes by gut feeling, and the gut prefers the process that annoyed it last.
One week, one tally sheet per process. How often a day, how long per case, how many follow-up questions. Not a consulting project, a sheet next to the keyboard. At the portal customer the result was one person, every morning in the portal, plus questions to dispatch. Every day.
The number alone decides nothing. But it ends the discussion about which process is the most expensive.
Every step, every exception, every unofficial rule. For a process with 23 steps this takes two to three weeks, and it is the stage nobody shows on LinkedIn. It decides what can be automated and what stays with a person.
At the portal customer, rules surfaced that only one person knew, and some were out of date. The reason for the Tuesday ban on customer C had long since disappeared.
Only now the service. It operates the portal the way the dispatcher does, checks each order against the TMS and confirms. What is not unambiguous goes to a list with a reason. Through the interface where there is one. Through the screen where there is not.
One process, not five. Only when the first runs in production does the next begin. (Yes, that takes longer than starting five at once. It just ends more often with five running processes.)
Every run writes a protocol: checked, confirmed, held for review. The alarm hangs on the result. A service that runs and confirms nothing is an outage. At the portal customer the protocol has shown between 900 and 5,300 checked orders per run for eight weeks. The variation does not come from the service. It comes from the customer: Wednesday is correction day there.
The protocol is also the return to stage 1. It keeps counting, every day, without a tally sheet.
Four questions per process, each scored 1 to 3:
Ten points or more: now. Seven to nine: second. Below: leave it, or sort out the data first. The rule is rough, and it is meant to be. It only prevents the most recently annoying process from winning.
Three cases where we advise against it:
A caveat: the scoring rule comes from about 80 processes at forwarders and warehouses with 20 to 500 employees. Whether it holds in a group with its own IT department, I cannot say from our own data.
Which of your processes scores ten? Send me your four numbers. I will tell you whether we can have it in production in four weeks, and show you the protocol of a comparable service. Ask for a process check: 30 minutes, no sales pitch.
