With a product range of over 5,000 products, Niko is Europe’s leader in electrical components, from switches to sockets. Every week, over 200,000 pieces of their most popular product come rolling off the conveyor belt. As inspecting each finished piece manually would take way too much time, Niko had a traditional quality control system in place.
However, the system yielded too many false positives as well as false negatives. In other words: a lot of good pieces were being thrown out, while the bad ones were left to continue their journey.
AI proved much more accurate than their existing solution, reducing the number of misshapen pieces that got the green light by 33%. For the good pieces previously wrongly accused of defects, the number even shot up to 72%.
Added bonus: it now takes Niko just about 3 hours to train their AI to recognise faults in a new product, compared to weeks using a traditional vision system.