PepsiCo in Russia implemented technology of neural network for goods control on shelves

The senior Chief information officer of PepsiCo in Russia, Ukraine and the CIS countries Mikhail Platonov at the TAdviser IT Retail Day conference which took place on March 14, 2018 shared experience of the company with startups. He told about the organization and results of the acceleration program of PepsiCo LAB for the startups developing the projects in the field of the innovation IT solutions for business (a track of Tech LAB) and also food and drinks (a track of Food LAB).

In total in the technology direction of an accelerator of 10 pilot projects were considered successful. One of solutions of the pilot passed to wide use. It is about the platform for control of outlets based on a sensing technology of images using neural networks, Mikhail Platonov specified.

The product developed by the Russian startup of Inspector Cloud - the participant of the acceleration program of PepsiCo LAB is the cornerstone of the solution. The company is a resident Skolkovo.

Mikhail Platonov at the TAdviser IT Retail Day conference

In this project the functionality of recognition of SKU was implemented (Stock Keeping Unit - units of an inventory control), Mikhail Platonov explained TAdviser. The sales representative does a photo of shelves with goods from the tablet or the smartphone, sends them to a system where they will be recognized, and quickly receives from a neuronet the answer as far as the situation with goods corresponds to what it should be.

At the expense of it the employee can take at once on site actions for improvement of a situation if it is necessary: for example, to put additional goods quantity on a regiment or to order it, Platonov explained.

According to Platonov, the number of users of a system in the company makes about 1.5 thousand employees. Mainly, it is sales representatives of PepsiCo.

Advantages of this startup in PepsiCo call that the solution managed to be translated very quickly from a test stage on tens and several hundred users to much bigger number and also feedback speed from neural network.