AI Roadmap for REWE Digital

Introduction

The current roadmap for REWE Digital was developed during the MIT executive program. It is a short version of the roadmap. A full version is available upon request.

Executive Summary

The current AI roadmap will enable REWE to transform its existing retail business to an AI-first business model to strengthen its differentiation strategy and develop entry barriers against new competitors. The roadmap provides a high-level 10 year plan on the deployment of AI technologies with the focus on REWE’s Digital online food shopping and delivery service.

There are three main areas of AI on which REWE will be focusing:

The roadmap is divided into two parts by time factors. The first part deals with the implementation of machine learning and natural language processing technologies. The second part deals with the implementation of robotics. The division by time was made because the technologies mentioned in the robotics phase need further technological advances. Also, the roadmap includes success metrics. The progress of the project will be measured with the following metrics:

Since REWE is not an AI-first company, to achieve the goals REWE needs to undergo transformational processes. There are three main steps to transform REWE’s operations:

  1. Building an in-house AI team
  2. Providing broad AI training and upskill program to employees
  3. Developing external and internal communications

Current State

REWE Digital is a German online fresh food retailer with 200 million euros in sales in 2018 [1]. The company was founded in 2013 and is a subsidiary of the REWE Group. According to Statista, REWE’s online delivery service takes the first place in terms of market share [2]. Technology plays an important role in all steps of the value chain process. Starting from forecasting, ordering, stuffing, and delivering food. The company relies mainly on the differentiation strategy. The company differentiates in three areas: the quality of the fresh food, a large selection of food online available and educational content related to nutrition. REWE Digital operates in a high price segment and has a well-recognized brand in Germany.

Generally, AI technologies in retail can be used in areas such as demand planning, sales, and operations planning, warehousing, logistics, supplier management, pricing, and promotions. According to the latest job offerings on the REWE website [3], the company is investing in agile processes, big data processing, and intelligent supply chain management systems. Also, REWE Group is hiring software engineers with a focus on machine learning to improve human resource planning.

New AI-enabled systems fit into the current differentiation strategy and can be used in other areas to introduce new technological advances so REWE stands out from the competition, such as the creation of a unique shopping experience. The current focus of using AI technologies lies in automation and optimization of internal processes. The implementation of AI with the focus on the customer shopping experience, such as autopilot food delivery, driverless last-mile delivery and meal kit service can strengthen the differentiation strategy of REWE and grow the revenue for the company. REWE Digital operates in an ecommerce market, and the food segment shows the fastest growing rates in Germany [4].

Another main area to drive business impact is through the reduction of food waste. According to the European Commission, 88 million tonnes of food waste are generated annually. The associated costs are 143 billion euros per year [5]. Through the implementation of AI-enabled systems that allow intelligent on-demand forecasting, REWE Digital will be able to move towards a cost leadership position in specific segments of the product range. In the case of food waste, REWE Digital, as well as society, can benefit from the effects of AI.

Proposed Initiative

The following proposal to achieve competitive advantage includes three main technologies: machine learning, natural language processing, and robotics.

Machine Learning

Machine learning can be used to find patterns in data to make predictions about customer orders and inventory levels.

Food delivery on autopilot

The current fresh food delivery service of REWE operates like an ecommerce store. Customers need to visit the website, add products to a basket and pick the date for the delivery. The suggestion is to remove the manual shopping process. Fresh food will be delivered on autopilot based on a customer’s historical purchases. Customers don’t need to shop anymore, the food will be delivered automatically. The supervised learning algorithms can help with time-series forecasting and unsupervised learning algorithms can be used to build clusters of customers based on the goods they regularly order. While the whole process is via an online ordering system, there is enough relevantly structured data for machine learning algorithms. As customers tend to order the same type of food with little variance at regular intervals, there is a high chance to identify patterns in the data.

On-demand ordering

The industry suffers from food waste. With autopilot shopping, there will be a possibility to optimize the supply chain, especially the production, processing, and distribution parts [6]. Based on the data from the autopilot shopping model, a supervised learning model can be developed to make time-series forecasting for up to six weeks upfront to meet customer demand. The structured data will come from the autopilot shopping model. To make the model work more precisely, data augmentation techniques such as the inclusion of data from public holidays and seasonal food calendars can be added. The system will enable REWE Digital to introduce on-demand ordering, similar to Amazon’s on-demand printing, and by doing so, food waste will be minimized and higher profits will be achieved.

Natural Language Processing

Natural language processing technologies can analyze and extract data from natural language and speech.

The new market of meal kits

REWE has already developed a virtual voice assistant [7]. The current version of the virtual assistant can read out recipes or add some products to the basket, but it is not able to convert dish names into a set of products. The proposal is to develop a service that can process customer requests such as a dish name and construct an order with all the needed ingredients, a so-called dish name converter service. Once the customer request has been deconstructed into an order, the ingredients and the matching recipe gets automatically shipped to the customer’s home address. To further improve the dish name converter service approach, REWE can introduce a weekly meal suggestion service. The virtual assistant can suggest a set of meals for a week based on customer preferences. The service can help REWE to enter a completely new meal kits market. According to statista.com, in 2017 the meal kits industry was valued at 4.65 billion US dollars.

Robotics

Robotics and robot process automation is a branch of AI that deals with the automation of manual tasks through robots.

Warehouse automation

The suggestion is to buy Ocado’s bot technology [8] for warehouse management and fulfillment processes. To improve the operation by warehouse bots further, REWE will need to develop new food packaging standards for robotics. The current generation of robots has problems with different shapes of items. Companies are trying to build robots to mimic human picking behavior. The goal of the initiative is not to build robots that can handle different shapes but to provide a packaging standard made for robots, where everything is packaged in a way that a basic robot can handle. As mentioned by MIT Professor Daniela Rus, currently it’s easier to send a probe to Mars than to build a robot to clean the table.

Driverless last-mile delivery

Fully automated food delivery without human drivers, driverless delivery, will increase the margins, make the business more profitable, and introduce entry barriers for new competitors. The entry barrier plays an important role in securing future business position. There is a danger that other companies from other industries will enter the food delivery market. An example of the trend is Amazon Fresh, but also other companies from other verticals such as the meal kits industry are trying to enter the grocery food market too. An important feature of the driverless last-mile delivery service is the inversion of control. Instead of REWE having to send drivers to deliver food, a customer will be able to manage the driverless delivery bot. A customer will take control via a mobile app and will order a bot themselves. The inversion of control will solve one of the biggest issues of the current delivery model, where the customer has to adjust its schedule to receive the delivery.

Plan of action and criteria for success

The plan of action will be timely and divided into two parts. The first part describes the strategy to 2025, and the second part describes the strategy to the year 2030.

Strategy 2025

Strategy 2025 includes three main initiatives:

Strategy 2030

Strategy 2030 includes two main initiatives:

Criteria for success

The following metrics will be used to measure the progress of the project.

The separation of initiatives into two different parts is needed due to technological challenges. While initiatives from the strategy 2025 can already be solved with off-the-shelf machine learning and natural language processing technologies, the initiatives proposed in strategy 2030 require technological advances. Since REWE is not an AI-first company and does its business in an “old-fashioned way”, REWE will need to undergo transformational processes. An approach for transformation is described by Andrew Ng in an AI transformation playbook [9]. The proposal is to focus on three main areas described in the playbook:

  1. Building an AI inhouse-team
  2. Providing broad AI training and upskill program for employees
  3. Developing external and internal communications

Building an AI inhouse-team

According to the current state of the company, REWE is not widely investing in AI-technologies. To create proposed services from the strategy 2025 REWE will need to hire a new type of employee. People with two extremes of skill level will be needed. The highly skilled employees that codify the processes and the highly skilled employees who will develop and manage machine learning systems. Since machine learning systems are not deterministic and may deliver wrong predictions, another type of lower skilled employees is needed. The lower skilled employees will handle the approvals for orders in a micro-task format to control the predictions of the machine learning systems.

For the proposed strategy 2030 REWE will need to hire robotic experts who will be in charge of choosing, deploying and managing warehouse bot technology. The same advice applies to the driverless last-mile delivery service. It is an area that needs more technological advances. Since the proposed initiatives are complex, the suggestion is to build a technologically versatile management team that will be in charge of finding and acquiring the needed technologies.

Providing broad AI training

The focus of the initiative is to focus on providing AI training and upskill training for the whole organization. It is assumed that AI will change the current type of organization at REWE Digital. People will be at the beginning and the end of the value chain. The middle layer and everything in between will be replaced by machines. It’s a system where people provide inputs and receive the outputs. The new type of organization will enforce a combination of people and machines. The structural changes and elimination of the middle layer will force the company either to upskill the people or to make redundant people who are overqualified for low skilled jobs and underqualified for high skilled jobs. REWE will need to develop an AI transformation program to provide people with the chance to acquire new skills for the new type of jobs.

Developing external and internal communications

Today, AI technologies are often poorly understood by people. There are many ethical concerns about robots, self-driving cars, and machines that will replace people. To avoid external and internal speculations and to increase the adoption rate of AI, REWE will need to develop a comprehensive communication program. The communication program will need to explain short, middle and long-term implications for REWE’s employees, customers and investors.

Conclusion

The described roadmap is a tool to organize people around the vision. The roadmap clearly communicates the direction of the company and all necessary steps to kick-off the transformational processes. The roadmap shouldn’t be treated as a static document, and it should be updated as new information and new initiatives become available. The roadmap will help managers and leaders to move towards a common future to build an AI-first company.

Footnotes

  1. https://www.onlinehaendler-news.de/online-handel/haendler/130797-rewe-umsatzrekord
  2. https://www.welt.de/wirtschaft/article194485581/Lebensmittelhandel-Picnic-will-Deutschland-erobern.html
  3. https://karriere.rewe.de
  4. https://www.gruenderszene.de/food/onlinehandel-wachstum-lebensmittel
  5. https://ec.europa.eu/food/safety/food_waste_en
  6. https://www.researchgate.net/figure/The-Food-Production-and-Supply-Chain_fig1_304709744
  7. https://www.rewe.de/rezepte-ernaehrung/voice-assistant/
  8. https://www.youtube.com/watch?v=4DKrcpa8Z_E
  9. https://landing.ai/ai-transformation-playbook/