The quarterly Retail Tech Bulletin, published by the Retail Analytics Council, Northwestern University, includes articles on current retail trends, case studies, artificial intelligence, and related retail technologies. The Bulletin also includes updates on Retail Analytics Council activities, the Retail Robotics Initiative, and the Retail AI Lab at Northwestern University.
If you would like to be a part of research conducted at Retail Analytics Council, Northwestern University, please contact us.
|
|
|
|
|
A Consumer View of the Grocery Market Landscape and the Potential Impact of Change on Retailer Market Strategy and Mergers |
|
|
|
|
|
|
|
|
By Dr. Martin Block, Professor Emeritus, Northwestern University, Retail Analytics Council, and Ronald Lunde, former senior executive in the grocery industry
In this article, six characteristics of the retail grocery market from the perspective of consumer self-reported shopping behavior are examined. Included is an analysis of the potential Kroger and Albertsons merger using the Herfindahl-Hirschman Index (HHI). The Department of Justice and the Federal Reserve commonly use the HHI as an anti-trust screening device in their analysis of the competitive effects of possible acquisitions or mergers.
Retail leadership and management teams must view the consumer economy not as a system that is constantly in equilibrium, but rather one where consumers, retailers, providers of goods and services, and the entire global supply chain constantly change their actions and strategies in response to the micro and meso-outcomes they mutually create. Therefore, change is largely an endogenous phenomenon, not necessarily a reaction to unexplained “Black Swan” shocks from outside the system.
The mental models of “old box” or “outside of the box” thinking is no longer adequate in our current era of rapid change. Leaders and mangers must embrace a new form of strategic creativity with the ability to continuously develop and test hypotheses, including new ways to embrace complexity, navigate uncertainty, and prepare for the disruptions that will inevitably materialize.
New tools are requisite to sense and respond to change. The objective of this article is to provide leaders and managers with a new toolbox. A “new box thinking” data analytics capability that utilizes both consumer data insights and market situations to create models which can either use or fuse first party, anonymous first party, corporate data, or industry data resources. Near-real-time and consumer privacy compliant enhanced perspectives on the potential impact of significant changes in consumer shopping patterns from both historical and or predictive analytics viewpoints can be explained. These superordinate insights will potentially provide organizations with the ability to sense and respond to change and create market strategy models that deliver not only market effectiveness, but also market efficiency.
Six areas of grocery market retail analysis using self-reported shopping behaviors are illustrated:
- Where Consumers Shop
- Average Number of Stores Where Consumers Shop
- A Linear Regression Model to Predict Monthly Consumer Grocery Spend
- A Geographical Representation of the U.S. Grocery Market
- Consumer Shopping Visits by Census Regions
- The Potential Impact of the Kroger | Albertsons’s Merger Using the Herfindahl- Hirschman Index (HHI)
For the purposes of this article, grocery retailers are put into six categories. To keep it relatively simple, some retail categories that sell grocery products such as convenience stores and drug stores, are omitted. Club stores such as Costco, Sam’s Club, and BJ’s, typically have a large footprint and a wide range of items beyond groceries.
The data used in this article is taken from the August 2022 monthly Consumer Intentions and Action Survey (n=7,552) conducted by Prosper Technologies of Worthington, Ohio. The data has been granted to Northwestern University for academic purposes. The data is collected online, weighted, and balanced according to age and gender, and limited to adults over the age of 18. Respondents are presented with a checklist of 56 retail grocery outlets that they have shopped in the last 90 days. Only 3.9% checked none, grocery shopping is then reported by 96.1% of the respondents.
1. Where Consumers Shop
Table 1 shows that 18.8% of respondents in August 2022 reported shopping at Costco in the last 90 days. The aggregate Club stores is the unduplicated sum of the three. Any duplication between the three has not been removed. It is analogous to the concept of gross audience in media, not net audience or reach. The Club category shows the largest decline from 2021. The Grocery category shows a 2.5% decline, although there are exceptions such as Food Lion and Winn Dixie. Overwhelmingly the increase in shopping is shown by the Dollar category at 19.4%. This is followed by Hard Discounters at 8.4%, led by Aldi. The Internet, consisting entirely of Amazon properties is at 2.3%. Mass Merchants show a modest increase, although Walmart by itself is at 1.1%.
|
|
|
|
2. Average Number of Stores Where Consumers Shop
Overall, as shown in Table 2, the average number of grocery stores shopped in the last 90 days among the 56 listed in the last 90 days is 3.7. As shown in Figure 1, the distribution is skewed, with the average number visiting four times or less is about 60% of adults, or the bottom three quintiles. The upper 40%, especially the top quintile, as shown in the next table, averages 8.1 stores shopped. The average monthly spend consistently rises with the store shopped quintiles, with the lowest at $268 and the top quintiles at $330. It is interesting that the lowest shopping quintile reports the highest percentage of online grocery, with it dropping in the low middle quintiles. Online grocery shopping rises again in the top quintile. This might be interpreted as an indication of how well the one store fulfills the needs of the respondent.
|
|
|
|
3. A Linear Regression Model to Predict Monthly Consumer Grocery Spend
Figure 1 shows the average number of stores visited during the last 90 days by store reported. Overall, the average is 7.07 stores. This average is biased in the direction of larger stores and excludes all those that did not select any store. Clubs, Grocery, Dollar and Hard Discounters all show a lower number of stores visited as expressed by the index. This lower number of stores visited might be interpreted as an indication of how well the one store fulfills the needs of the respondent. Mass Merchants show the lowest other store shopping rate. The Internet and all other stores, which tend to be smaller, shows higher average stores shopped.
|
|
|
|
Using a simple linear regression to predict monthly grocery spend is summarized as a chart in Figure 2. With an overall multiple correlation of .334, annual income is the strongest predictor indicated by a standardized regression coefficient, followed by household size. Being married and having children in the household follows, with being younger and shopping at more stores at the bottom of the list of statistically significant predictors.
|
|
|
|
4. A Geographical Representation of the U.S. Grocery Market
The grocery market varies considerably by geography. Without knowing each chains marketing areas, a reasonable approximation is by U.S. Census Divisions. Table 4 shows a Census Division map and table. The table shows the size of the region, as well as monthly grocery spend, percent purchased online, average household income, age, percent with children and the percent female.
|
|
|
|
The following Table 5, Marketing Characteristics by Census Division, shows multiple key marketing characteristics. It is interesting that buying online and pick-up in store (BOPIS) is highest in the west south central.
|
|
|
|
5. Consumer Shopping Visits by Census Regions
As Table 6 shows the shopping visits vary considerably by census region. However, Walmart, based on Stores Shopped in the Last 90 Days, demonstrates the highest percentage of ‘Shopped’ in every Census Division. Strong regional chains begin to appear in the regions such as Stop & Shop, Shoprite, Kroger and Meijer, Hy-Vee, Public, H-E-B, Costco and Safeway, ALDI is very strong where it has stores.
|
|
|
|
6. The Potential Impact of the Kroger | Albertsons’s Merger Using the Herfindahl- Hirschman Index (HHI)
Using the proposed Kroger and Albertsons as an example merger, the chains different banners need to be aggregated. Kroger includes Food4less, Fred Meyer, Harris Teeter, King Sooper, Kroger, Pick ‘n Save, Ralphs, Smith Foods, and Fry’s. Albertsons includes Albertsons, Safeway, Jewel-Osco, Vons, Acme, and Shaw’s. The next table shows the percentage of shoppers for both Kroger and Albertsons, as well as their other brands or banners. Also shown in Table 7 the percentage of each banner that is shared with the parent Kroger and Albertsons. The average banner parent shopping overlap is just under 15%.
|
|
|
|
A respondent is given credit for selecting one or more in each credit. In August 2022, 20.1% shopped at any Kroger banner, 16.2% shopped at an Albertsons flag, and an additional 8.4% shopped at both for an overall 44.7% for the combined total. This assumes that there would be no store closures and all stores would be identified as the combined brand. Note that this places the proposed store ahead of Walmart in only two census divisions.
The key statistic used to assess competition is the HHI. To calculate market shares the shopping proportions need to be normalized. In this case divided by the unduplicated total for the 56 stores. The percent shares are then squared and summed. An HHI below 1,500 indicates a competitive industry, an HHI between 1,500 to 2,500 indicates moderate concentration. An HHI above 2,500 indicates high concentration. As Table 8 shows, currently the total US and each of the nine census divisions clearly demonstrate competitive marketplaces. If the proposed Kroger Albertsons merger occurs, and the stores all branded the same, the worst-case scenario, then the HHI still remains in the competitive marketplace range although the HHI always increases. Overall, the HHI increases just under 37%. The biggest impact is the Mountain with a 216% increase and the West South Central with a 151% increase. As an aside, non-metropolitan areas (D counties) are already slightly less competitive with an HHI of 682 but increase only 24.3% with the proposed worst-case merger.
|
|
|
|
Using stores shopped as a criterion, the grocery market is very competitive and remains so even if two major competitors, such as Kroger and Albertsons were to merge, and rebrand all their stores with one brand. The merger creates a supersized grocery company with about 5,000 stores that currently operate under approximately 50 banners. However, there are considerable differences by geographic regions which might be subject to further market-by-market government review resulting in market-based adjustments to the proposed merger.
In conclusion, we are beginning to understand that markets are not necessarily static or linear but are essentially always changing—a recursive, reflexive loop. We need to have ‘new tools’ and ‘new box’ thinking, the data insights, and mental models to sense and respond to the evolving patterns that shape individual decisions which ultimately create the patterns for the systems of tomorrow. With enhanced consumer data insights and metal models, leaders and managers will have a better understanding on how change might impact both retailer market strategies and mergers.
|
|
|
|
|
Supply Chain and Autonomy |
|
|
|
|
|
|
|
|
By Timothy Wei, former Dean of Engineering from the University of Nebraska
The correlation between labor force and supply chain was no more evident than in the meat processing industry in the early days of the coronavirus pandemic. Within weeks of the pandemic spreading across the US, the meat processing industry was in crisis. By May 2020, the rate of infection in US counties with 20% or more employment in meatpacking was as much as ten times higher than other non-metro counties (Figure 1). At the same time, between workers falling ill, plant shutdowns, and workforce reductions, there was a precipitous drop in meat production (Figure 2). A year that had started with such promise for the meat industry devolved into chaos.
|
|
|
|
The pandemic, of course, unleashed a perfect storm. International trade routes were paralyzed by port closures. Panic buying exacerbated shortages. Factories and retail outlets alike were locked down. Employees in retail (and across society) either chose or had to find alternative sources of income. In short order, the entire retail supply chain fell into disarray.
As the US emerges from the pandemic, there have been dramatic shifts in the distribution of the nation’s workforce. While statistics are returning to pre-pandemic levels, many have chosen not to return to traditional jobs; this raises questions about the meaning of percentages across a disruptive event. The trend toward alternative sources of income, e.g., the gig economy, was emerging even before the pandemic. Whatever the cause, retail and service industries have had to adjust to doing business with fewer workers (Figure 3). This poses tremendous challenges to industries where the in-person customer experience is of utmost importance.
|
|
|
|
Labor challenges, however, are not unique or limited to retail. Indeed, manufacturers are having similar difficulties in attracting and retaining workers. Shortages in labor, materials, and parts are therefore combining to reduce production of goods and impede recovery of inventory-to-sales ratios (Figure 4). The challenges faced by the retail industry today are therefore ultimately a culmination of challenges across the supply chain. And their resolution will require holistic solutions across that supply chain. This is not something that any one industry or company can solve on its own.
|
|
|
|
One increasingly viable technology opportunity is autonomy. As distinguished from automation, which is machines doing repetitive tasks, autonomy uses data and analytics to enable devices and systems to intelligently carry out complex functions in dynamically changing environments. Automation requires clear separation between machines and humans to ensure worker safety and operational efficiency. Autonomy allows safe collaboration between humans and machines, i.e. co-robotics, to leverage relative strengths and mitigate weaknesses.
Autonomous mobile robots (AMRs), for example, are mobile platforms that have been deployed in factories and warehouses. Equipped with an array of sensors and signals, AMRs are being used to transport material throughout facilities with full situational awareness of objects that are both stationary as well as moving, e.g., people. They are programmed with onboard intelligence to complete their missions without endangering people, themselves, or their cargo. Further, instead of carrying cargo, they can transport machinery, such as robotic arms, which allow them to make decisions and carry out nuanced tasks much in the way humans do. In the context of retail, the technology exists to deploy an autonomous system to manage a stockroom with a capital expense on the order of $10,000.
In larger, more complex environments, fleets of autonomous devices can be deployed to accomplish an array of activities. Most recently this is being and has been done in a number of food and beverage processing facilities both in the US and abroad. In these applications, a central command center provides assignments to individual AMRs for efficient operation across the facility. Each AMR, however, has its own onboard intelligence to optimize its own path to minimize distance, time, and energy usage while avoiding collisions and areas of congestion. At the completion of each assignment, the AMR receives a new and potentially completely different task.
A key point of using food manufacturing as the exemplar here is that the technology now exists to safely deploy autonomous systems and robotics to handle highly variable objects such as food, as opposed to highly standardized objects such as ball bearings. But more importantly, autonomous systems represent a solution to the labor vacuum that has been emerging (and exacerbated by the pandemic) since the beginning of the twenty-first century. This is no longer a question of replacing human labor with something ostensibly more effective and efficient. Rather it is a question of filling the void needed to accomplish manufacturing and retail tasks necessary to provide essential goods and services to society. Finally, while autonomy in manufacturing will be essential for sustainably meeting the needs of the retail industry, it represents a powerful tool to fill labor needs directly in the retail space.
|
|
|
|
|
Retail Robotics & AI Conference Recap |
|
|
|
|
|
|
|
|
By Zain Rahman, Graduate Student Northwestern School of Communications
The Retail Analytic Council's held its annual Retail Robotics and AI conference on November 2. The conference focused on topics that provide opportunities for growth and efficiency for retailers and brands. The program began with Spiegel Research Center’s Executive Director Larry DeGaris announced the RAC is officially merging with The Spiegel Center. The resources brought by Spiegel and the RAC team will create an environment of technological innovation and applied research. Larry addressed how the collaboration between the two organizations will create a powerful resource for partners, providing technologies, automation, and analytics to engage shoppers and drive performance.
The conference continued with an overview of past research projects by Martin Block. Professor Block demonstrated how customer transaction data transactional data can be utilized for deep insights and strategic alignment. The integration and understanding will play a key role in online and in-store sales. Martin also discussed future technological developments making use of quantum computing that will dramatically change research practices. Emerging technologies and key aspects such as quantum computing and the web 3.0 can give RAC and its retail partners an advantage in understanding and serving shoppers.
The next speakers were Professor Edward Malthouse, Research Director at the Spiegel Center, and Sinan Seyman, a doctoral student in Engineering. Their presentation's key focus was recommender systems and their use for serving advertising and purchase suggestions. Professor Malthouse researched the "rating matrix" used in recommendation systems, and how Amazon ads filter based on clicks and interests. Popularity bias, the tendency for models to recommend popular items, plays a crucial role in product suggestions and advertisements presented. Sinan spoke in-depth about the coding element of creating software that can identify search and purchase trends.
|
|
|
|
Ronald Lunde, former senior executive in the grocery industry, delivered a presentation about how the digital consumer is disrupting the traditional retailing and branding models in place. Ronald was adamant about analytics-oriented retailers and brands have a competitive advantage through the use of data. He referenced that the shopping environment is both physical and digital at the same time. He advised that understanding what the customer truly desires is crucial.
In keeping with the conference theme of technological innovation, Professor Mario Savvides of Carnegie Mellon University addressed the coming role of artificial intelligence in the consumer-packaged goods industry. AI recognition systems can help retailers in two critical ways. The first is reducing shrinkage cost by better tracking inventory. The second is creating an efficient checkout process that is quick, simple, and easier current scanning technologies. AI-based optical scanning tracks updated product lists and observes each item without scanning the barcode. The process requires large data dictionaries to identify each item. The technology can reduce shrinkage because the item's weight is evaluated, so two safety methods are embedded into the software (vision and weight). The current constraints are storage capacity and updating of product lists.
Professor Frank Mulhern, Director of the RAC, spoke about the importance of smartphones in the shopper experience. Utilizing smartphone functions for shopping-related benefits can enhance the shopping experiences and better connect shoppers to a store’s digital ecosystem. He presented a segmentation analysis that categorized smartphone users into quintiles based on number of functions used. Results showed that people who use smartphones for the most functions are more sophisticated shoppers and heavier users of most forms of media.
Professor Timothy Wei, former Dean of Engineering from the University of Nebraska, discussed the use of robotics in warehouses for improving efficiency. He described how a robotics sensor can be used to navigate products through a warehouse in driverless vehicles. He also described innovations in food packaging whereby clean and recyclable food packaging material are utilized for improved safety and package integrity, including the detection of spoiled foods to replace date stamping.
|
|
|
|
The program concluded with Martin Block discussing the changes in the fashion marketing that are driven by shifts in shopper preferences, and implications optimally managing apparel inventory. The RAC thanks the speakers for showcasing their work and contributing to a very informative conference.
|
|
|
|
Retail Analytics Council (RAC) is the leading organization focused on the study of consumer shopping behavior across retail platforms and the impact of technology. Established in August 2014, RAC is an initiative between the Medill School of Journalism, Media, Integrated Marketing Communications and the McCormick School of Engineering, Computer Science Department, Northwestern University.
This document is not to be reproduced or published, in any form or by any means without the express written permission of Northwestern University. This material is protected by copyright pursuant to Title 17 of the U.S. Code.
|
|
|
|
|
|
|