Posti’s parcel lockers have become smarter than ever. The artificial intelligence (AI) system behind the scenes has significantly evolved, now offering comprehensive forecasting across several critical stages of the logistics chain. The updated system no longer focuses solely on parcel collection forecasts but evaluates the entire delivery process more holistically.
With these new AI models, Posti can predict:
When and to which sorting centre a parcel will arrive
How quickly the parcel will be shelved after sorting
When a customer is likely to pick up their parcel from the locker
This enables more precise logistical planning even before the parcel physically enters Posti’s system.
Predictive data optimises locker utilisation and sorting
The popularity of parcel lockers continues to grow, yet their effective use demands increasingly precise management. Lockers have limited capacity, and the pace at which space becomes available varies greatly between locations.
In especially busy areas, such as shopping centres or smaller grocery stores, physically expanding locker size is often not an option due to space constraints.
“Popular lockers can't always be expanded because they're located in limited spaces, such as inside shops. This was a key reason why we began developing and implementing AI technology,” says Kasper Wigren, who is responsible for Posti’s parcel control system.
Previously, locker capacity management relied on average utilisation rates, which overlooked specific locker usage patterns. This occasionally resulted in lockers remaining full despite incoming parcels, causing delays and unnecessary redirection.
Now, AI predicts locker-specific availability, considering the day of the week, time of day, and regional demand patterns. As a result, parcels can be routed directly to locations where space is reliably available, speeding up delivery and bringing parcels closer to customers.
AI models bring new precision to the delivery chain
Posti’s latest AI models provide greater visibility and accuracy at various stages of the delivery chain—even before parcels physically reach Posti.
AImodel forecasts when and at which sorting centre a parcel is expected to arrive. This prediction utilises advance shipment data from online retailers (EDI messages), allowing routes to be planned well in advance.
Another AImodel estimates how quickly a parcel arriving at a sorting centre will be shelved. The goal is to achieve 95% accuracy, and if successful, these predictions may soon become available to customers through the OmaPosti application. Currently, accuracy already stands at 91%.
Parcel pickup forecasts complete the intelligent delivery chain
AI advancements also extend to accurately predicting customer pickup behaviour. By understanding how quickly and when parcels are typically collected from specific lockers, Posti can better schedule and allocate parcels to the right lockers at optimal times.
This intelligent approach maximises the limited capacity of parcel lockers. By aligning locker availability closely with pickup patterns, unnecessary transfers are reduced, ensuring lockers have space precisely when needed for new parcels.
From the customer's perspective, this translates into improved service quality—parcels are more likely to be available at the preferred locker without delays or unexpected redirects. As forecasting models continue to improve, Posti is increasingly able to meet growing expectations for speed and accuracy in parcel deliveries.