Shop floor feedback: quantities and times without paper

The usual route runs through the office: cards are collected, bundled and transferred into the merchandise management system once a day or once a week. This creates a delay that no amount of diligence can remove. Stock figures in the system are older than the stock on the shelf, the order status is older than reality in the hall, and every delivery statement to a customer rests on data that has already been overtaken. On top of that, retyping ties up office hours without producing a single new value.

The second effect weighs heavier. During transfer, information is lost that was never on the sheet in the first place. An operation was interrupted twice because material was missing, but the card shows only the total time. A batch was changed halfway through the lot, but the card shows the number used first. A part was reworked rather than scrapped, but it appears as good quantity. Such details are known at the workstation and can no longer be reconstructed in the office. What is not captured where it arises cannot be restored later, only estimated.

This leads to a simple design rule: the capture point belongs where the information arises, and the effort there must stay small enough for the message to fit into the working rhythm. Three or four entries at a terminal two steps from the machine do get made. A form with fourteen fields in an office at the other end of the hall gets filled in afterwards, and filling in afterwards means estimating. This trade-off influences data quality more than the choice of software.

The digital traveller does not abolish paper outright; it changes its role. Printing continues, but only as a carrier for an identifier: the job paperwork carries a barcode or a two-dimensional code for each operation. Whoever starts the operation scans the code at the terminal, and the system immediately knows the order, the operation, the product and the target quantity. Nobody has to type in what is already known, and that is precisely what determines whether the message actually happens in daily work.

The rest of the interaction consists of a few steps: report the start, enter good quantity and scrap quantity at the end, pick a downtime reason from a short list if needed, close the operation. Times follow from the timestamps and do not have to be entered. Where several people work on one operation, each signs on with their badge so that time is apportioned. Where one person operates several machines, several operations run in parallel, which also needs to be settled beforehand because it changes how time is allocated.

Two design decisions determine acceptance. First, the number of mandatory fields stays small; all further entries are optional or filled from master data. Second, the screen shows large buttons and little text, because it is operated with gloves on, in noise and under time pressure. A form designed for a desk does not work at a machine, and a message that takes too long is caught up from memory at the end of the shift.

Material issue is the point where stock management and production meet. As long as issues are posted retrospectively, the book stock differs from the shelf stock and purchasing works with figures that are wrong. Capture at the workstation resolves this, because the issue is reported in the same interaction as the operation: scan the container, confirm the quantity, carry on working. The additional effort is a matter of seconds; the benefit for purchasing and stocktaking is considerable.

For companies with traceability obligations, the batch or serial number is added. What counts is the link in both directions: which batch of raw material sits in which lot of the product, and which customer received which lot? This chain can only be closed if the batch number is captured at the moment of issue. Reconstructing it later from delivery notes leaves gaps precisely where a change occurred within a lot, and that is the case that matters when it counts.

Technically the effort is manageable, provided the material master data holds up. Every item needs a unique number and a readable code at the storage location or on the container. Where labels are missing, where several numbering schemes coexist or where the same item is held under two descriptions, the project starts with cleaning up master data. This step is regularly underestimated and often shapes the schedule more than the implementation itself.

The actual return lies not in capturing but in what becomes possible afterwards. As long as quantities and times land in the system bundled and delayed, evaluations at operation level do not hold up; you can calculate them but you cannot decide on them. With feedback per operation, figures emerge that previously circulated as assumptions, and assumptions differ noticeably between production, sales and management.

The first and most important evaluation is the comparison of planned and actual effort per order and per operation. It answers whether a calculation holds, not as an annual average but for individual products and lot sizes. In almost every company this surfaces items that have been calculated below value for years, and others where there is room. This evaluation changes prices and lot sizes more often than any measure taken in the hall.

Further evaluations follow from the same data: scrap rate per operation, machine and shift; setup share of total time and with it the question of economic lot size; distribution of downtime reasons by frequency and time lost; real lead time from order release to completion; delivery reliability per product group. Capacity utilisation in manufacturing is surveyed regularly at the level of the economy as a whole (ifo Institut); inside a single company a comparable statement is only possible once occupancy times per machine exist. How this becomes concise, understandable reporting is covered by metrics and reporting, and the decisive point is to start with few figures and keep their definition stable.

Feedback data describes operations, but technically it can also be attributed to people. That touches codetermination and data protection. Systems capable of recording the behaviour or performance of employees are subject to codetermination in companies with a works council; independently of that, purpose limitation, access rights and retention periods have to be defined. The specific arrangement, for instance a works agreement ruling out evaluation at person level, should be checked professionally in each case and cannot be derived from an article.

What works in practice is a clear commitment communicated before the start and then kept: evaluations happen at the level of operation, machine and cost centre, not at the level of individuals. Where signing on with a badge is necessary to allocate time correctly, evaluation can still be aggregated. If that commitment is broken, data quality deteriorates within a few weeks because messages are delayed, rounded or omitted, and the damage then hits the whole project.

Two further limits deserve mention. First, capture does not replace planning: without maintained routings and standard times you get actual values but no target-versus-actual comparison. Second, not every production process can be broken down equally well. With very short operations, the reporting effort can end up in an unfavourable ratio to processing time. In that case, combining several steps into one capture point is the better solution, even though detail is lost. This trade-off should be made deliberately and documented, rather than settling in quietly on the shop floor.