AmRest transforms its operations with
Azure-based Enterprise Data Warehouse

Whether you’re on your way to a long-awaited vacation, staying late at the office, or perhaps just hanging out with your friends, nothing makes a situation better like your favorite food.

You might drop by your chosen restaurant, or have your meal delivered to anywhere you want. It’s fast, easy, and you know it will always be just the way you like it.

As you’re waiting patiently for your order, there is much more happening in the background. Large amounts of data flow from one place to another, carrying information on which products sell most often, which supplies need to be ordered, or how many employees are currently on (or off) duty. This kind of information holds enormous potential.

Building up the power brands

AmRest has been steadily building its leader status as an independent chain restaurant operator since 1993. It brings people’s favorite brands, such as KFC, Pizza Hut, Burger King or Starbucks, to customers around the world, from Central and Eastern Europe, through Spain and France, to Russia and China.

The company started with one restaurant in Wroclaw, Poland. Now, AmRest manages 2,346 restaurants in 26 countries, employing over 50,000 people. With the growing complexity of the business, the company decided to invest in new technological solutions which would help it make the most of data coming from its ever-expanding network.

The growing potential

AmRest generates hundreds of GBs of structured data a day. As the company grows both organically and through acquisition, with its rapid growth, it has managed to create a wide portfolio of data processing platforms which required a lot of integration efforts. Information is gathered from multiple systems spread across different domains, making it very difficult to align.

The company comprehensively oversees each business domain, empowering employees through accountability for areas of its performance, leveraging its data assets in the process. Ensuring that the right information is delivered to the right people at the right time is crucial in making data-driven decisions.

With the successful growth of the company’s online channels, the need for gathering and effectively processing unstructured data became more relevant across virtually all departments, from marketing to operations. Understanding customers’ eating habits and general brand perception allows the company to take a more personalized and individual approach to each order and marketing campaign, providing a unique, customer-focused experience.

Understanding the potential of its data and limitations of its current systems, the company decided to strategically re-invent its reporting and create a unified, consistent data platform where everyone will have access to the required information.

Taking the first step towards a new data strategy

AmRest decided to modernize its data platform in 2018. The goal was to simplify its operations by bringing together data from all domains into a single platform, thus enabling easy access and analysis of data across the company.

In response to AmRest’s inquiry, Predica, in close cooperation with Microsoft, proposed a modern data warehouse architecture leveraging Azure Data Services. After a successful proof of concept, during which the company assessed solutions from top BI solution vendors, the company decided to establish a strategic cooperation with Microsoft and choose Predica as its implementation partner. The project began in January 2019.

The services and components that form a Modern Data Warehouse (adapted from the Microsoft model)

As with any new technology, the Azure cloud was a cause for some doubts. The issues of managing the cloud, its privacy or compliance often raise questions. However, we presented clear advantages of building an infrastructure using Azure Data Services, which addressed AmRest’s concerns. Those included:

• PaaS services requiring fewer administrative and management responsibilities from the infrastructure staff
• No maintenance needed (version upgrade, patching, etc.)
• Scalability, HA and DR functions available right out of the box
• Easy and quick delivery and configuration
• Low entry costs
• Pay-as-you-go pricing
• Some services may be suspended when not in use (Azure SQL DW, Azure Analysis Services).

Identifying the project goals

AmRest wanted an infrastructure that could adapt and scale to the rapid business growth, increase in data volume and growing user base. The company aimed to establish a single source of truth that could be accessed quickly and easily by literally every employee.

Flexibility

The first goal of the project was to provide sales, marketing, and operational departments with a highly performant data model that would enable users to slice and dice any sales-related data at a push of a button while maintaining clarity and convenience. All daily reporting, ranging from the analysis of key performance indicators such as sales revenue, number of transactions, average transaction value, to the rankings of best-selling products or meals, must be fast and easy.

Performance

Performance was of the utmost importance to AmRest. The solution had to handle rapid creation of rankings: employees, restaurants, regions, districts based on sales revenues, in several dimensions such as time, geography, sales and ordering channels, etc. As the data is related to different countries, the rankings also had to support daily fluctuations in exchange rates, which had to be brought to a common denominator.

Example of a new dashboard (with no real data)

Reporting

The company had a working BI solution prior to our engagement. Each user had access to an online portal to generate ad-hoc reports upon request. The core reports were generated by the system as static reports and sent to subscribers each morning. As each user received several reports, this quickly added up to several thousands of files. The generation phase was preceded by data acquisition, reconciliation, and conversion.

As a result, the system was running at capacity and in need of maintenance and support. Additionally, static reports provided the end users with limited business insights. Requests for specific information had to be handled by dedicated teams. The new system had to provide an easier and more efficient reporting functionality.

Adoption

The developed solution had to measure its adoption across the organization. The data was to be gathered from services such as Azure Log Analytics, and Power BI and input to a dedicated report showing how each service is being used, and the increase in the overall user base.

Getting usage statistics on the new solution is useful for determining the ROI, but also for identifying areas where more training or support may be needed.

Defining the capacity of data solution

The previous system supported millions of transactions per day. The new solution, according to predictions, must be able to handle at least several times more over the next three years. Over that time, it also has to function efficiently for thousands of users and scale up to several thousands more. AmRest expected the solution to support thousands of simultaneous user sessions at any time, with consistent performance.

Currently, the data warehouse holds several TBs of data and is expected to grow at a rate of about 40% per year. Some of the data comes from non-integrated PoS systems, which the solution also needs to support. The new infrastructure is evolving in line with new business requirements, but with the Agile approach, we can introduce changes on an ongoing basis.

It is expected that ultimately, there will be close to a hundred GBs of raw data added each day to the new data warehouse, made available to business users in 30-minute packages loaded from PoS systems.

Putting a dam on the data flow with Master Data Management

To maintain data quality, we have implemented Master Data Management. Permission management works dynamically based on the organizational structure assignments, such as employee, restaurant, region, country, or brand.

The system must meet the requirement of full functionality: 24/7 throughout the year with SLA at the level of 99.5% of operation/availability. All reports should be generated in no more than 60 seconds. One of the key reports is issued in less than 3 seconds based on a model with billions of records, which shows the efficiency of the new solution.

Multi-dimensional analysis in just one click with Power BI

Power BI reports allow business users to slice and dice data according to various dimensions, such as restaurants, regions, parts of the day. The dashboards enable a top-down approach, investigating details, and analyzing data in multiple ways. Intelligent visualizations help quickly understand even the most dynamic scenarios.

Business users will be able to manually enter values for future estimates of sales and transaction data. The Power BI dashboard with an integrated value delivery form allows them to make estimates and immediately display results in reports.

Example of a new dashboard (with no real data)

Summary of the Enterprise Data Warehouse project so far

In addition to the new EDW infrastructure, in partnership with Microsoft, we were able to deliver:

• Point of Sales analysis
• Key analytical reports
• Profit and Loss statement & reporting process
• Cost analysis divided by organizational hierarchy and product category
• A dedicated interface for managing master data
• Access to a single source of truth using Excel or Power BI

Technologies we used include:

• Azure PaaS Services (Azure Data Factory, Azure Databricks, Azure Data Lake Gen2, Azure SQL Database including Azure SQL Database Hyperscale, Azure Synapse Analytics (former Azure SQL Warehouse), Azure Analysis Services, Azure Functions, Azure Automations, Azure Log Analytics)
• Azure IaaS (VM for Master Data Services, VM for Data Factory Integration Runtime)
• Power BI Premium

Project delivery model

Setting sails for the future

So far, we have completed the first of three parts of this long-term engagement. To enable AmRest to make the best use of its data, we will support it in going through three phases:

• 1^st wave – become a data-driven company
• 2^nd wave – enable real digital transformation
• 3^rd wave – other data usages

Machine learning (ML) and artificial intelligence (AI) with video recognition algorithms for detecting fraud will be used in the future. Our team will prepare forecasting and predictive models to drive better planning and budgeting. We will also assist with enhancing e-commerce capacities and customer experience. AmRest will then be able to carry out ad-hoc marketing campaigns.

The goal is to enable it to “run in the morning, measure and receive data on customers’ moods for lunch, and read analysis reports by the end of the day”.

Our extensive experience with advanced data analytics, including acquisition, cleaning & integration, processing, visualizing & sharing vast amounts of data using a modern approach (Modern Data Warehouse) utilizing Azure Data Services, meant that we were able to successfully assist AmRest in its project.

Our know-how related to delivering complicated, long-lasting projects for international and multi-entity organizations meant we could design a solution that was relevant to its needs and answering its requirements. Understanding business perspective and needs, we designed a flexible solution, delivered with business users in mind.

As a long-term loyal Microsoft partner and holder of 12 gold competencies, we have years of experience in using Microsoft products and delivering reliable solutions. As a result, we became the partner of choice for AmRest in upgrading its EDW and taking its data analytics strategy to a new level.

If you want to brainstorm your ideas, do not hesitate to book a free scoping call with our experts. We’re ready to help!