Outforz team has presented PICSELL

The trend of developing and integrating IT solutions in the retail sector opens up new opportunities for developers to create IT products. In the field of trade, the main emphasis have always been on CRM system for customer interaction. The foundation of successful business in production and distribution is the rightly selected and integrated ERP system. The number of offers of finished products CRM + ERP and companies involved in the integration of such solutions, is large enough as well as the number of customers who are actively using such solutions to improve the efficiency of production processes and build a personal relationship with the consumer.

How to improve the efficiency of a well established business that already uses CRM + ERP? The answer was searched by the Outforz team before creating the future product. In fact, it lies in the plane of the capabilities of AI and neural networks, because in the work of field trading teams, merchandising structures and retail auditing, there is still a share of the so-called human factor, due to which data are collected and analyzed that supplement CRM/ERP systems. Merchandising resource is an important aspect of controlling brand representation in retail. This is an expensive service based on the work of the field team and usually performs service support for several clients at the same time. In some cases, manufacturers/distributors are forced to maintain their own field resource, which significantly increases the cost of executing/supporting merchandising KPIs in proper form. A slow and poorly motivated merchandiser or a team of merchandisers directly affect the correctness of data generation and business efficiency in general. After all, if the goods are not on the shelf the goods are not being sold in fact.

PICSELL is a software product presented by Outforz team, the main purpose of which is to optimize the time spent by merchandisers in retail chains through automation and the use of new technologies. This is a kind of AI injection into an existing client business process that does not require complex integration or improvement of the ERP system.

The main goal of the PICSELL project is to optimize the costs of maintaining the field structure for clients by several times, to increase the accuracy and speed of the information provided by the field structure, to automate manual labor and to eliminate human factor errors.

PICSELL helps to solve a number of the most common tasks and KPIs of merchandising:

  • automatic detection and verification of the correct location of the SKU on the shelf;
  • detection of out of stock on the shelf;
  • comparison of the location of the SKU with the planogram;
  • estimate the number of facings on the shelf;
  • estimate the length of the shelf;
  • monitoring the prices of competitors' products;
  • monitoring the performance of merchandisers, including task management system and motivational program.

How does it work?

The basis of the product is a self-learning neural network, the result of which is the automation of verification processes for SKUs with an accuracy of 95-98%. Authorization of a new client in this software takes no more than 2 weeks, and the product itself is provided on a freemium subscription.

In fact, the use of a neural network is only the tip of the iceberg. The main difficulty faced by all IT developers in the field of neural network training is the process of dataset formation (a set of data, photographs, structured in a specific format, which allow the neuronet to “learn”). The neural network supervisor loads this data through the admin panel and while being processed it receives new pieces of information, thereby increasing the accuracy of verification.

In order to form datasets, you need to use a human resource. There is no other way to train the neural network to verify SKUs in retail; accordingly, these sets of photos are formed manually. It can be thousands, millions of photos, structured according to certain criteria. As the large photos are marked in manual mode, they are turned into a set of clear data, which is necessary for the neural network to improve its accuracy indicators. For different product categories, the training period of the neural network can be different.

What are the results for clients who use PICSELL?

Let's imagine how much time it takes to manually enter prices of about 50 SKUs (approximately 10 seconds each price tag). In the case of PICSELL, a merchandiser only needs to take 1 photo. All other man-hours are replaced by this system. Eventually client receives a report on everything that has happened during the whole day: what products are presented, whether prices are correctly indicated in dynamic analysis by networks, by cities, by addresses, by individual merchandisers, up to a specific SKU, exactly how it is presented in networks in different cities. This is important information, because retailers do not always have time to correctly make changes in prices, correctly place promotional POS-materials, place products on shelves. This is very important because PICSELL allows you to make the right management decisions for a more efficient display of goods, it also helps to make an analysis of the effectiveness of the entire chain until the goods appear on the shelf.

Another problem is traditional retail. Despite the fact that in the classic network retail these systems work, the traditional retail is not subject to full control today. And PICSELL also helps with this, since it became possible to monitor the availability of products, based on the minimum assortment matrix, which should be in this outlet.

How the idea was born?

In the process of analytics, it was found that the full cycle of preparing the final report takes months. In addition, those people who are in the "fields" do not always correctly do their work. The paper reports were replaced by software solutions that helped to partially automate the work of merchandisers. Technologies that allow to recognize not only QR codes and barcodes, but also images appear. And the team wondered why not recognize the real objects and train the neural network.

Previously, Outforz had the task to build a staff motivation platform called Push&Pull. A simple and at the same time effective tool was needed that would, on the one hand, control the location and KPI performance of field personnel. But, on the other hand, it was necessary to give a simple and clear loyalty program. Accordingly, they developed a simple solution that made geolocation detection possible. Merchandiser could choose from the nearest outlets from the database and after that could start his work. As a confirmation, they used a phone camera and 2 photos were formed: at the start and at the end of the working day. For each photo and comment the merchandiser received points, and on a specially designed landing page could choose his rewards or bonuses. An indirect result of the staff motivation system was a huge set of photos that came daily from each employee. It was decided to monetize this data. Such information can serve as a starting point for the formation of datasets and the team tried to use simple examples to conduct experiments with neural network training and get the first results. These datasets are an invaluable resource in order to train this neural system and automate the merchandising process, as well as save staff time at the outlet. This is how the PICSELL project was born.

What collaboration options are available for the client?

After gaining experience with different directions, the team came to the conclusion that there is no single approach to integrating the product into the business of different clients. When PICSELL is being introduced into business processes, clients receive 4-week trial, because for both the company and the client it is an opportunity to make sure that the system is working, to correctly deliver KPI and prepare reports.

The PICSELL solution initially  had the possibility for clients to use it by subscription, this is the competitive advantage of the service. For SaaS solutions are the future, it is convenient and profitable for customers. In addition, using SaaS solutions, the client does not need to carry out an expensive ERP system upgrade, since PICSELL data can be formed in any convenient, customizable format.

The client is offered several packages at the beginning. The first and basic functionality is very simple: control the KPI of the merchandiser’s visit, the formation of the first photo database. Several key merchandising indicators are taken into account:

  • validation of the SKU presence;
  • out of stock in comparison with the assortment matrix;
  • detection of shelf length in the context of product categories, as well as selected SKUs;
  • price tag monitoring.

Further, additional options for the client are offered:

  • planogram verification;
  • detecting the presence of competitors;
  • shelf share in your brand’s category compared to a competitor’s brand;
  • detection and verification of competitor’s price lists;
  • optimization of transport movements of merchandisers.

In the process of integration and training, new needs appear, on the basis of which additional micro services for clients are born. For the Outforz team, as for developers, this is a unique expertise and experience, since on the basis of such insights, a further vector of product development is built up and it is adjusted depending on the client’s needs.

Clients may request such data from research companies, but they must understand that this data is a one-time, and esearch company needs a certain time period to process such request. In the case of the PICSELL product, the client receives data in real time using an already existing field team.

Prospects for the development of PICSELL

The future of SaaS solutions is seen in the development of options and microservices. The basic version of the product should be simple and understandable to clients. And then, as a constructor, each individual client connects microservices that can solve individual problems.

The Outforz team suggests that this system in the future will become some kind of basic platform. Any consumer making a purchase at the point of sale and having an application can receive various tasks: making photos of shelves, commenting on display of goods, prices, etc. When performing these tasks he will be credited with bonuses. Consequently a “secret customer” feature will be easy, cheap and fast to implement using the PICSELL product.

We are also talking about product development in other markets. Working with different product categories, the team realized that there are no territorial boundaries for the PICSELL. Automatic recognition of SKUs can be carried out both in Ukrainian and Russian as well as in English, Arabic, etc.