← All articles

Mikhail Polyakov

A warehouse where the system thinks one thing and the floor shows another: what the "reality gap" is and how to measure it before buying new automation

Level: BeginnerReading time: 8 min

Topics: Automation, Industry, Small business

A warehouse aisle with shelves and boxes, with a shifted glowing digital copy of the same shelves above them, showing boxes where the real shelves are empty

The short answer: Cecile Dick-Calmes, inbound marketing manager at Daifuku Wynright (The Robot Report presents her as an employee of Daifuku, a company that automates warehouses and production lines), wrote a column for The Robot Report on 12 October 2026. Her point: many warehouses invest in automation but do not get the expected gains, and most often the reason is not a lack of equipment but the fact that the warehouse software and what happens on the floor diverge. She calls this divergence the "reality gap." The remedy, in her words, is to coordinate the work of people, equipment and software based on current conditions rather than outdated data. The column contains not a single figure and not a single verifiable example, and the author works for a company that sells such solutions. But the problem itself can be measured in your own operation, and below are three measurements that do not require buying anything.

Who is writing and what exactly is claimed

The author of the column, according to her byline in The Robot Report, is responsible for inbound marketing at Daifuku Wynright; previously she was a community manager at Boeing and led digital marketing at Hainan Airlines. The column is written as an opinion, without data. It is worth reading as a description of a problem as a vendor sees it, not as proof.

The main claims of the column:

  • Warehouses collect huge volumes of data: inventory movements, equipment status, people's work, transportation, orders. But the data, according to the author, arrives too late, sits in separate systems or lacks the necessary context. So managers react to what has already happened.
  • What this looks like, the author lists as follows: inventory records do not match what is on the shelves; congestion is noticed once it has already developed; a bottleneck in one process spreads to neighbouring ones; people and equipment are deployed based on an outdated picture; delays build up across the whole chain.
  • The problem, in her words, is less about technology than about execution. A warehouse may have good automation and good software, but if they are not aligned with what is happening now, the losses remain.
  • Hence her main thesis: automation provides movement, and coordinated work of all the parts provides results. Success, in her words, is increasingly measured not by the amount of automation or equipment utilisation, but by how quickly a warehouse notices changes, makes decisions and carries them out.

What the author calls real-time execution

By this the column means the ability to continuously see what is happening in the warehouse, assess changes and coordinate the response of people, processes and equipment. The author lists five foundations:

  1. constantly knowing the current state of operations;
  2. immediately noticing disruptions and deviations;
  3. reallocating work and resources on the fly;
  4. quickly adapting to changes in demand and workload;
  5. coordinating decisions across connected systems.

According to the author, the goal is not to collect data faster, but to make the right decisions faster.

What is useful here and what is worth checking

The useful part: the column frames well a question worth asking before buying another robot or conveyor. Is the warehouse held back by a lack of equipment, or by the fact that the records and reality diverge? If the latter, new equipment working from the same wrong data may not deliver the expected gain. This is my reasoning based on the logic of the column, not a result shown in it.

What to check: the column has not a single figure, not a single warehouse where the gap was measured, reduced and the effect calculated. The benefits the author lists (resilience to disruptions, consistent quality, better use of resources, fewer inventory discrepancies) are stated without evidence. If a supplier promises them to you, ask for the same figures as for any automation: what it was before, what it became after, at which warehouse and over what period.

How to measure the gap yourself: three measurements

These are my suggestions, not a method from the column. All three can be done by your own warehouse staff, without a new system.

  1. Records versus shelves. Take a sample of storage locations, for example 100 bins from different zones, and recount the goods by hand. Match rate = the number of bins where the quantity in the system equals the actual quantity, divided by the number of bins checked.
    • Who does it: the shift supervisor together with a storekeeper who did not work in those zones that day.
    • How to check: a recount sheet with bin numbers, quantity in the system and actual quantity, signed by the checker; the match rate is calculated from this sheet.
  2. Recording delay. For 30-50 random operations per shift (receiving, moving, shipping), write down when the operation was physically completed and compare it with the time it was recorded in the system.
    • Who does it: the warehouse manager or an employee they assign, with a watch and a list of operations.
    • How to check: a table "time on the floor / time in the system / difference"; look not only at the average but also at the largest delays.
  3. Manual corrections. Count over a week how many tasks the dispatcher or team leader reassigned by hand because the system issued them based on a wrong picture (the goods are not in place, the aisle is blocked, the equipment is down).
    • Who does it: the dispatcher, noting every such correction in a log.
    • How to check: a log with the date, the task and the reason for the correction.

When it makes sense to buy something

If the measurements showed that records and shelves almost match, entries appear in the system quickly and there are few manual corrections, then by these three indicators you do not have a noticeable gap. This does not mean there are no other manifestations the column writes about, such as congestion or bottlenecks between areas: the three measurements do not cover them. But buying a solution for a gap you do not see in the figures is premature. If the discrepancies are noticeable, it is worth first understanding the cause. One possible cause, my assumption and not a conclusion of the column: the work procedure, for example operations are recorded in a batch at the end of the shift rather than as they happen. If that is the case, it can be fixed with work rules without a purchase. But the measurements will only show where and by how much the records diverge from the shelves, not why. The cause is worth checking separately: for example, go through several discrepancies from the recount sheet with the storekeepers to see when and how the entry was made, and only then change the work procedure.

Buying software or equipment to coordinate the work makes sense when it is clear how much money is lost because of the gap. An important caveat: the time people spend searching for goods is money only when reducing it actually lowers costs, for example paid overtime goes away, you no longer need temporary workers or do not have to hire another person. If wages stay the same, the freed-up hours will not by themselves pay for the purchase. A simple calculation:

  • Monetary losses from the gap per month = pay for overtime and temporary workers needed because of searching for goods and rework + the cost of incorrect shipments and correcting them (repeat delivery, returns, customer penalties) + losses from equipment downtime, if because of it orders are not fulfilled on time.
  • Payback period in months = (price of the solution + implementation) / (the part of the monetary losses that the solution will actually remove, per month − its support costs per month). If the denominator in brackets is zero or less, the solution does not pay off.

An example with hypothetical figures, not from the column: because of searching for goods that are not where the system shows them, the warehouse pays for 160 hours of overtime every month at 750 roubles, which is 120,000 roubles; incorrect shipments and their correction cost another 30,000 roubles. Monetary losses are 150,000 roubles per month. If the solution costs 2 million roubles including implementation, removes half of the losses, that is 75,000 roubles, and support costs 15,000 roubles per month, payback is 2,000,000 / 60,000 ≈ 33 months. If it removes only a quarter of the losses, 37,500 roubles, the saving after support is 22,500 roubles, and payback grows to almost 89 months. What share of the losses the solution will remove should be shown by a trial period at your warehouse, not by the vendor's promise.

I wrote about the same approach, the task and the figures first, then the equipment, in the article on why a robot is chosen based on the task.

Summary

The column by a Daifuku employee in The Robot Report suggests looking at a warehouse not through the number of robots, but through the gap between what the software sees and what happens on the floor. The idea is sound, but the column has not a single figure, and the author represents a vendor of such solutions. Before buying anything, the gap can be measured with your own resources: records versus shelves, recording delay and the number of manual corrections. If the gap is noticeable, first find the cause, and evaluate a purchase by the losses it will actually remove.

I work on AI agents and automation. If you would like to see my projects or discuss your own task, take a look at my portfolio.

Sources