Four items, but one is the wrong model
Hypothetical example: an order contains four spare parts. There are four in the box, but one is the wrong size. The count matches and the shipment is still incorrect. Checking quantities and checking product identity are different tasks for a small distributor.
Before buying cameras, identify the error you want to prevent: a missing item, a substituted product or a box linked to the wrong order. Each needs a different check. This guide proposes a test; it does not describe a customer installation or guarantee that a camera can recognise your catalogue.
1. Separate product identity, quantity and order
The product reference identifies the item. Quantity tells you how many units should ship. The box or order identifier connects them to a shipment. Counting objects in an image does not answer the other two questions by itself.
Define the expected contents before packing starts. If someone changes the order during preparation, decide how that change reaches the workstation and which version is being checked.
- Example order: two units of A and one of B.
- Observation: three visible units, but one product identity cannot be confirmed.
- Result: review required; three objects are not enough to approve the order.
2. Check whether a barcode already provides the identity
GS1 defines the GTIN as an identifier for trade items, while barcodes allow identifiers to be captured by scanning. If your products already have readable codes, consider that data before trying to distinguish nearly identical packaging by appearance.
Scanning also needs process rules: what happens if a unit is scanned twice, removed afterwards or has an unreadable label? A scanned code and an item ultimately packed are not necessarily the same event. Software must connect the reading to the actual workflow.
3. Test the camera on the products most easily confused
Vision can be evaluated for presence, quantity or visible differences. The question is whether the image contains enough detail under workstation conditions: position, lighting, movement and obstruction. A hidden difference cannot be resolved by promising higher accuracy.
Include similar references, rotated labels, several items together and typical lighting changes. Keep preparation samples separate from evaluation samples. Compare results with manual review and record situations where there is insufficient information.
4. Decide what happens when something does not match
An alert should support action: identify the box, explain the discrepancy and request a review. Agree who can correct a result and when a shipment may be closed. Evidence must be linked to the right order, with agreed access and retention.
Measure missed substitutions, alerts on correct orders, duplicate units and wrong box associations separately. Also measure how long staff take to resolve an issue. The aim is to help the workstation rather than add another screen nobody can attend to.
5. Start with one workstation and a feasibility decision
An initial test can cover one workstation and an agreed set of product references. Deliverables would include the tested conditions, results, observed failures and a recommendation to proceed, adjust or stop. Sample size, acceptance criteria and price are agreed before contracting.
Adaris Vision Studio supports preparing and evaluating detection models, while Player runs the selected configuration. Comparing expected order contents, managing closure and integrating warehouse software require a specific scope. These are not presented as an already deployed order-packing system. Cameras, installation and production integration are assessed separately.
Related services
Sources
Which error do you want to catch before closing a box?
Tell us what you pack, what gets confused and how you check orders today. The initial enquiry does not need images or customer data. Adaris works from Elche with businesses in Alicante and across Spain; commercial discussions are currently in Spanish.