Parcel intelligence: automate the inspection of inbound parcels

Conventional methods for inspecting and categorising items come with the risks of human error and slowing sortation speed – even when parcels are nicely labelled with adequate information. With the aid of parcel intelligence and machine learning technology, categorisation becomes easier and more effective.

By Jan Schroeder


Registering a parcel’s profile – its type, size, weight, sticker, danger, odd shape and so on – is essential to a smooth sortation process.

It enables differentiation, for example, between polybags and flat items, parcels and postal bags, those items with ‘dangerous goods’ stickers on them or limited quantity dangerous goods or odd objects such as tires that cannot be handled in some types of sortation systems.

The importance of registering parcel data

Given the rules that apply when handling dangerous goods, being able to categorise inbound parcels based on non-barcode information is important for distribution centres.

The question is how do you sort these ‘registration’ profiles? If a hub has information in a database, then this is fairly simple; bar codes on parcels can be correlated with entries in databases and those entries will inform the hub about not only the receiver but these registered values.

How to register the unmeasurable

A very simple solution is to separate inbound parcel flows according to lines – one line for bags and one line for parcels, for example. But while this works well enough, it’s far from an ideal solution from an operator’s perspective. It lacks scalability because it requires an additional line for every different type of parcel. It’s also space-consuming and requires more personnel.

Another straightforward solution is a centralised system using just one line for a mix of parcels that are registered within the line. While this solution also works, it too has its limitations in that the operator can only register so many parcels at a time and will, after a period, lose concentration.

Learn more in our introduction to digitalisation of distribution centres.

So what if we could centralise and automate the system even more? This is where parcel intelligence and machine learning processes can be a smart way to overcome the shortcomings of these more conventional ways of registering parcel profiles.

What is parcel intelligence?

Parcel intelligence is letting the system decide the category of a parcel, instead of a human. It uses software to categorise a picture of an item into its type, through image analytics software that employs machine learning technology.

In a distribution centre, it looks like this:

  • A photo is taken of the inbound parcel.
  • The photo is sent to a computer for analysis by its software.
  • The software’s algorithm will determine the nature of the parcel, categorise and register it.

How parcel intelligence works

The types of items that algorithms can categorise are limitless. But first, algorithms must be built to identify specific types of items. How is this done?

With very specific details and explanations of what the item consists of, an algorithm can be built and trained to identify the item. Through the process of machine learning – of training, running, checking and improving – the algorithm is constantly improved and strengthened. Ultimately, it’s able to identify the parcel in all its variations and possibilities so that sortation of items using this technology becomes very, very accurate.

Applying this type of intelligence is in fact straight forward and can do away with some of the more expensive scanning solutions. At this point, expensive equipment is not needed to know every single profile detail of the parcel. It just needs to know what kind of parcel it is so the system can determine what to do with it – and that’s what machine learning can do very quickly.

How to implement parcel intelligence

Parcel intelligence is easily implemented in both new and existing systems. Taking images, training the system and conducting test runs can all be carried out during normal operations without interfering with the distribution centre’s production. Once the algorithm is put through test runs, analysed and improved upon, it can then be implemented on actual flows and sortations. It doesn’t require high quality images in order to work, so will perform just as well using the types of cameras found in existing sortation systems.

Read “The parcel distributor’s guide to data analytics and digitalisation”.

Benefits of parcel intelligence

The great value of adopting parcel intelligence in distribution centres is that the system can be used for any item the human eye can see – important for hubs that are increasingly receiving different types of items driven by the boom in e-retail.

The system is constantly improving, so if inbound flows change and new types of parcels are arriving, the system can adjust to those changes through the image analytics and computerised categorisation.

As mentioned, parcel intelligence will work with any camera technology – from high quality line scanners to simpler cameras used only for this purpose of taking pictures of individual items.

The system can be installed anywhere and on any type of equipment, whether it’s at the inbound receiving point, the booms, the point at which the items are placed on the sorting system, over the conveyors, over the chutes or on the gates in and out of the facility.


Parcel intelligence technology and machine learning is an easy and effective way in which distribution centres can register immeasurable parcels in their sortation processes, without slowing down or reducing inbound capacities. It’s a very agile technology that can easily be tested prior to its installation to prove its potential benefits. And it just gets better and better over time as the algorithm improves!

Do you want to know more about the potential of digitalisation for your distribution centre? Download our e-book about digitalisation for parcel distributors.

Subscribe to our newsletter