We’ve all been there: walking through supermarket aisles searching for items that should be there, but aren’t. Sometimes we can’t find something because it’s in the wrong place or hidden behind a sign. Sometimes we can’t tell what the price is. Whatever the situation, customers don’t care. They have countless shopping options and are going to find what they want elsewhere. With CB4, retailers make sure every shopper leaves satisfied.
CB4 helps bricks-and-mortar retailers grow sales by making the in-store experience easier for store teams and their shoppers. “Running stores is quite complex and difficult,” explains Rafi Rainshtein, Vice President of R&D and General Manager Israel at CB4. “We use machine learning to help store teams make quick, easy fixes—like taking a product out of backstock or ordering more units—to gain new sales and satisfy shoppers .”
CB4’s solution sends stores a list of recommendations for SKUs that they can sell much more of now. Each store’s unique selling patterns and operation conditions determine which products are on the list — so no two stores get the same set of recommendations.
CB4 follows a technology approach Rafi calls “sophisticated, but simple,” combining state-of-the-art machine learning technologies with an intuitive user interface that anyone can manage. With the CB4 app, store managers need to be able to identify issues in seconds. For that to work, Rafi and his team have to ensure that every aspect of the technology stack operates in harmony.
“We can divide our solution into three parts,” explains Rafi. “The data pipeline, where we collect sales and product data from our customers. The machine learning operations, where we generate and prioritize recommendations. And the application, which our clients use daily.”
To help optimize all three aspects, CB4 works with technology partners who tailor their support to the size and unique demands of the business. “When we have issues or queries, we need someone to take that call immediately, in person, instead of just sending links.” says Rafi. “And that’s the kind of partnership we’ve developed with DoiT International.”
In 2019, when Rafi joined the company, CB4 did not have a robust central data pipeline to extract, process, and transfer customer data to the machine learning environment. In addition, the team was using different scripts and processes that were custom-built for each client, which got the job done, but held back growth during a time when CB4 was greatly expanding its customer base.
Rafi and his team decided to leverage Google Cloud tools to develop a more robust solution. They selected Apache Airflow with Google Cloud Composer to generate a scalable and agile data pipeline by orchestrating the flow of data reliably, before funneling that data into the machine learning models. Replacing legacy processes with an entirely new tool was a difficult task, and DoiT International helped ensure a smooth implementation.
“We have to move a lot of data, and move it quite fast, and our data pipelines need to be robust to scale with our growth,” explains Rafi. “Google Cloud Composer helps us achieve a fully automated data pipeline we can use for all of our customers, and DoiT International helped ease the difficult implementation of the tool.”
Besides flexibility, the new system also helps CB4 ensure that data is stored securely, while complying with international data protection regulations, such as GDPR. In the data repository on Google Cloud, Rafi and his team can apply parameters, and geo manage data to certify that EU data is stored in the EU only, as required. “We need to trust that data remains in the right place,” explains Rafi. “Our new data pipeline helps with that.”
To analyze all this data, CB4 has developed ML models in-house, using Google Compute Engine to host the analysis infrastructure. As heavy users of Google Compute Engine, spread across different regions, CB4 was faced with a growing cloud spend, and prompting Rafi and his team to search for ways to track and optimize that spend.
“If we’re going over budget, that’s usually because something is wrong in the stack, but without transparency, we can’t identify these budget issues, wasting money,” says Rafi.
Today, CB4 uses DoiT International’s Cloud Management Platform to analyze and predict cloud costs, balancing the company’s growth with cost-efficiency. “Our cloud spend is going up, because we’re growing,” says Rafi. “DoiT International helps us track more precisely where our money goes, and identify and resolve issues quickly, saving money.”
Most recently, DoiT International helped CB4 migrate from N1 to N2 machine types within Google Compute Engine, making its machine learning models more performant. “We’ve been able to measure a 30% performance increase of our machine learning,” says Rafi.
With its streamlined data pipeline, 30% more performant ML operations and improved cost visibility, CB4 is ready to bring its AI-powered retail revolution to many new clients around the world. The team can easily integrate new retailers into the data solution, and ensure high availability even during peak times and when scaling.
On their mission to make in-store operational issues a thing of the past, Rafi and his team already have their eyes set on new technologies. Like with Cloud Composer and Compute Engine before, the DoiT team will be always just a call away, ready to provide support on every step.
“To master technology, you really have to learn its basics first, and learning from experience is usually the best,” says Rafi. “As our go-to partner for all things Google Cloud, DoiT International provides that experience, helping us save time, while making it easier to explore and implement new technologies.”