Skip to content
All field notes
From practice24 August 20266 min read

Packing lists for quality control: the real lesson is in the source

Automating packing lists for quality control is about more than reading a document: it is about what is in that document, how it looks and how reliable the source is. At HDG, active in food and quality inspection, dottle now processes 200,000 packing lists per year with 90 percent less manual work. The biggest lesson was not in the technology, but in the variation of the source data itself.

By Yeslin Beljaars

Why packing lists for quality control are so difficult

A packing list sounds straightforward: product name, quantity, weight, maybe a lot number. But in practice, that same information arrives as a clean PDF from one supplier, as a screenshot taken on a mobile phone from the next and as an Excel file with its own formatting from a third. For an inspection system like QC4U, that makes no difference: it expects structured input. The employee in between types everything over. Every time. That is not only time-consuming, it is also a source of errors in a place where errors have direct consequences: quality control.

What goes wrong when you try to automate this directly

The first instinct is: we put OCR on it and the system will read it out. That works as long as the format is fixed. The moment a supplier adjusts their layout, moves a field or reformats a table, the template reads nothing. And in food and fresh produce trade, suppliers change formats regularly without notifying anyone. You are essentially building a maintenance burden, not a solution. What you need is a system that understands document context: what is the article number here, what is the weight, what is the batch number? Regardless of where on the page that data appears and how it is formatted.

What you learn from variation in source data

The real lesson when automating packing lists for quality control is this: the quality of your automation equals the quality of your weakest input source. A screenshot taken in poor lighting, a PDF generated from an old system without a text layer, an Excel file with merged cells and hidden rows: all of these variants need to be recognised and correctly converted. That does not happen in one go. You start with the suppliers who deliver most frequently and most consistently. Those are the first to tackle. Then you build volume. The harder cases, suppliers with poor PDF quality or highly variable formatting, are handled separately. That way the automation rate rises step by step, rather than trying to solve everything at once and getting nothing right.

When does automating packing lists for quality control make sense?

It pays off when you have sufficient volume and when the input into the inspection system follows a recognisable pattern: the same fields, recurring suppliers, a consistent process. At HDG, that means 200,000 packing lists per year. At that volume, automation recovers nearly a full FTE, now available again for the actual inspection work. If your volume is lower or your process is highly situational, the more honest approach is to start with your busiest suppliers and measure the time savings there first. Automation is not an all-or-nothing decision.

What human-in-the-loop means in quality control

In quality control, you do not want a system that makes its own decisions and conceals errors. dottle flags deviations: if a field cannot be read with certainty, or if the extracted value does not match an expected reference, a staff member receives it for approval. The staff member approves, not the system. That is intentional. In an environment where an error in batch registration directly affects inspection outcomes, predictability matters more than speed. 85 percent flows through automatically; the rest requires a check. That is a better division of labour than entering everything by hand.

Seeing this in your own document flow?

Book a demo on your own documents

Frequently asked questions

Can packing lists submitted as screenshots be processed automatically?

Yes, but recognition quality depends heavily on image quality. Sharp screenshots with sufficient contrast work well; blurry or poorly lit images more often require a manual check. Start with the most readable sources and work from there.

How long does it take before packing lists are processed automatically?

dottle is live within two weeks, without a months-long implementation project. The first suppliers are onboarded quickly; more complex formats are added step by step as volume grows.

What if a supplier changes their packing list format?

Unlike OCR templates, dottle does not require you to build a new template. New or changed formats are added in a controlled way, not learned automatically. This keeps the system predictable and auditable.

Does packing list automation work if I use QC4U or similar inspection software?

dottle delivers structured data into the system where the work happens. Whether that is QC4U, a WMS or another system: the connection is part of the setup. Ask about the specific integration via the integrations page.