It seems retail today is about as dynamic as it has ever been. Competition from online retail powerhouses and non-traditional channels is intense. This is especially true in grocery, which was once the exclusive purview of ‘pure-play’ grocers,1 where now almost every retail format is selling one form of grocery or another (increasingly fresh foods) to draw people into their stores and venues. Home delivery has really taken hold. Amazon Go is expanding, providing checkoutless convenience.
Taken altogether, these industry dynamics are forcing grocers and other retailers to reinvent themselves, including providing excellent (not mediocre) omnichannel services (fully integrated channels, ultra-fast free or low-cost shipping, returns, etc.) and reimagining their most expensive physical asset, their stores (see Stores’ Identity Crisis: The Reimagining of Physical Stores). Hyperlocal assortments, locally sourced produce, clustering by buying behavior rather than demographics, and other changes are increasingly necessary but very difficult to achieve without more advanced systems and processes in place. It is against this backdrop that I attended Xcelerate, Symphony RetailAI’s annual user conference in Dallas. Symphony RetailAI is helping their retail and CPG customers meet many of these challenges, using Artificial Intelligence (AI) as the centerpiece of their strategy.
Figure 1 – Symphony RetailAI’s Vision for the Store of the Future
A Comprehensive Suite of Solutions for Retailers and CPG Brands
In early 2018, Symphony RetailAI was created from the merger of Symphony Gold and Symphony EYC, each of which had their own long and storied history.2 The combined entity offers a comprehensive suite of capabilities for retailers and CPG manufacturers. Symphony’schairman and CEO, Dr. Pallab Chatterjee, was the first keynote speaker at their user conference, Xcelerate. After being introduced by CMO Kevin Sterneckert, Pallab laid out Symphony RetailAI’s promise to their customers as follows:
- We will deliver revenue and margin growth to you in the first year.
- We will be the easiest company for you to do business with.
- We will provide the best quality and service.
- We will deliver breakthrough AI-enabled innovations.
Kevin said that Pallab was really holding the company accountable to these expectations throughout the entire organization. Pallab went on to describe the three layers comprising Symphony RetailAI’s solution offerings:
- CINDEEUREKA — This is Symphony’s data managementand AI/ML platform, which ingests data from many sources, including internal data (e.g. retailers own POS data, inventory, etc.) and external data such as item master data from suppliers, syndicated data or retailers’ POS (sent to CPG firms), and data from non-Symphony systems. It also includes master data management tools, which are important in an environment with such diverse data sources.This platform includes enterprise AI/ML that is leveragedand embeddedthroughout the entire suite.
- CINDEAPPS — The applications for Retail and CPGs are categorized by role/persona, within Retailer and CPG manufacturer organizations, as shown below. For retailers, this includes solutions for category managers, marketing, store managers, and supply chain. For CPG manufacturers this includes sales, category managers, and brand managers. Within these applications, CINDE AI logic is running to help answer specific questions and provide proactive recommendations relevant to each targeted role.
- CINDEUX — This provides a common UI, including a common way to view and interact with CINDE AI’s insights and recommendations. CINDE will show the predicted consequences and outcomes of differentscenarios to help the decision-making process.
Figure 2 – Symphony RetailAI Solution Stack
Role-based Rollout
Symphony RetailAI is positioning themselves as the ‘insightand recommendation system’ for retailers and CPG manufacturers. They are embedding AI into functionality and modules across the suite starting first with merchandising/category managementand demand forecasting and replenishment. Their next focus is the store, putting power onto the mobile devices of store managers and associates. After that supply chain. However, firstSymphony is focusing on getting merchandising and category management deployed and working well.
A Day in the Life of a Category Manager
In order to more deeply understand what AI functionality would be most valuable and useful for category managers, Symphony RetailAI’s product managers spent time not just interviewing category managers, but also conducting a ‘day-in-the-life’ ethnography — literally followingcategory managers around and observing how they do their work, where they spend their time, what kinds and sources of data they use in their job, what systems they use, and how they do analysis and discovery. They observed common weekly patterns and four areas where category managers spend the majority of their time:
- WeeklySales Analysis — the category manager typically starts Monday morning running a bunch of reports to see how they did in sales the previous week. They then spend four to six hours trying to figure out what happened and what to do about it. So, one of the goals for SymphonyRetailAI’s system is to eliminate all that running around and bring the answers to the category manager before they even ask — tell them “here’s what happened” and “why” and provide actionable recommendations.
- Vendor Meetings — These are dialogs with vendors, negotiating shelf space and volume commitments, and working through issues. Here CINDE provides detailed data on how the vendor is performing down to store/shelf/product level, as well as providing recommendations for improvement.
- Ad Planning and Preparation — Category managers typically look at what they did a year ago for promotions and advertising. They examine how that past promotion performed and often keep it mostly as-is withminor tweaks. CINDE takes into consideration other business conditions and trends and recommends the optimal promotion (media, pricing, placement), based on existing conditions (e.g. weather, competitive actions, etc.) and past performance. Now, as various category managers sit in a room together, they can have insight- and recommendations-driven conversations when negotiating who gets the ads on the front page and other prime spots in print and digital campaigns.
- Store Visits — If the category manager manages to get through everything else, they will try to do some store visits. When the other activities consume too much time, store visits may be neglected. CINDE frees up time formerly consumed in gathering, organizing, and digesting data from the other activities, thereby giving the category manager more time to visit stores. CINDE also helps them decide which stores they should visit and where their attention will deliver the most value.
A Walkthrough ofCINDE’s Category Manager Suite (CMS)
Figure 3 – Category Manager Suite Home Screen with ‘CINDE FEEDS’
At Xcelerate, I got a demo of CINDE CMS to see how the pieces work together. The data in the demo was real data but anonymized, from a combination of several retailers. As shown above, the home page of CMS (Category Manager Suite) displays high level KPIs, key pieces of information, insights, and areas needing attention or action — all brought to the category manager in one place, rather than them having to login to various systems to export and massage the data, create reports, and dig through the reports.
The CINDE feeds displayed on the home page span the main activities of a category manager. Clicking on any one of them will drill them down into that item. Across the top of the page are five top-level menu items, corresponding to the category manager’s main activities: 1) Weekly Sales, 2) Daily Sales, 3) Vendor Analysis, 4) Store Visits, 5) Ad Planning. The series of screenshots below shows these activities from the perspective of a category manager that is responsible for Fresh and Frozen Pizza and provides a feel for how these tools might be used by them.
Figure 4 – CINDE Weekly Sales View
The weekly sales screen provides a summary of last week’s performance for the category (in this case Pizza). Across the top are KPIs, such as sales, volume, margin, and coupon spend vs. last year. Below those is a breakdown of subcategories performance; in this case it is showing that Frozen Pizza sales are down 2.1% compared with last year, and that caused the overall decline in sales. Under each subcategory are key insights and recommended actions. For example:
- “A decline in penetration with price sensitive customers,”
- “70% of the overall decline is due to promotions not performing as well as last year,”
- “Lower discounts on promotions led to a decline in traffic to the category, although with better margin (3.4%),”
- And “I have identified the promotions that are ineffective and have recommendations for better tactics.”
It likely would have taken the category manager hours — possibly days — of gathering and analyzing data to come to those conclusions on their own, plus more time figuring out which promotions are ineffective and what to do about it. The system is saving time and recommending actions that potentially outperform the actions the category manager would have take otherwise. CINDE’s analysis of promotion performance is sophisticated and takes into account things like cannibalization and the halo effect from other promotions and/or competitors’ actions, as well as weather and events.
Figure 5 – CINDE’s Daily Sales View
Similarly, CINDE can update results each day, allowing the category manager to keep up during the week, tackling issues as they arise, and monitoring the results of their ongoing actions.
Figure 6 – Vendor Analysis, Meeting PreparationView
Above we see the Vendor Analysis screen helping the category manager prepare for a meeting with one of their vendors, in this caseThe Schwan Food Company about their Freschettabrand. It shows vendor/brand performance metrics along the top, with more detailed information about long-term and recent performance below. There are sections recommending immediate action, as well as joint opportunity areas. For example, it notes, “10% of regular Freschetta customers are starting to shop the category less often” and then recommends “Target these customers with reward offers to maintain their engagement.” Clicking on the latter brings up a specific recommended plan. The retail category manager’s counterpart at the brand can be provided access to this same data and tools to facilitate the discussion and a joint action plan.
Figure 7 – Store Visit PreparationView
The store visit screen similarly shows detailed information and suggestions for actions in preparation for a store visit. It provides insights such as “This store has 20% more Quality Driven customers, relative to the rest of the division” and areas needing attention such as “5 lines have regular out of stocks. I have generated a product list.”
Figure 8 – Ad PlanningView
The Ad Planning Meetings screen provides tools to view the past performance and forecast the expected performance of planned ads.
Store View – Analyzing and Visualizing Planograms and ‘Realograms’
Symphony RetailAI’s CINDE also has Store View, that provides a visual view of the shelves for a category and enables analysis of performance to a granular level.
Figure 9 – Store View — Planogram Visualization, KPIs, Intelligence
Figure 10 – Store View — Performance by Brand
Figure 11 – Store View — Drill Down to Single Brand, Analyzing the Types of Buyers
CINDE Category Management Rolled Out First in Europe, Then the U.S.
The CINDE category management functionality has been launched in Europe and is in the process of being rolled out to every one of Symphony’s Category Manager Suite (CMS) customers in the U.S. At first, they will all get upgraded to the next generation of reports and the CINDErecommendation interface. Separate from that, the next generation ‘AI-decision coach’ (as described above), with AI models, prescription, immersive visualization, and mobile alerting, has been rolled out with one major U.S. customer.
Agile Merchandising
Category managers are pressed for time, and rarely find the time to do ‘what if’ exercises. Category resets can take 30 weeks from start to finish. As a result, most categories are reset no more than twice a year. Retailers often push category management responsibilities off to a brand owner, the category captain. However, in spite of rules and guidelines, the brand owner is inclined to provide biased answers that favor their brand,which may or may not be in the best interests of the retailer.
Symphony has the vision to help retailers and CPG companies start doing ‘agile merchandising.’ They believe they can get the 30 weeks it takes to do a reset down to under 10 weeks. The idea is to run an analysis on every category once a month and see where the highest opportunity is for a category reset for that month. The reset does not have to bea slave to the calendar, mindlesslydone at a regular period, but rather can be done at the optimal time.
Virtual Reality Store Layout / Shopper Experience Insights
Using virtual reality and online testing, these cycle times can be compressed even further to bring truly agile merchandising. Next, Symphony wants to collapse the cycle time to two days. To speed things up, Symphony works with InContext to provide virtual reality views of shelfs. Since most of the time is in the physical work, this is like doing virtual testing on a digital design.Through the partnership with InContext, Symphony customers are using virtual reality to test different store layouts as well. These are quite realistic, as seen in this video. Symphony’s Shopper Experience Insights solution allows rapidly trying out many different variations for fixtures, layouts, lighting, and merchandise placement. In one case, a major retailer tried out more dramatic lighting and found it improved customer engagement by 33%.
Shopper Experience Quotient
The Shopper Experience Insights can create a virtual store from models, then run tests with a representative panel of shoppers in this virtual setting. These shoppers are given a task or ‘trip mission.’ In the case mentioned above (testing out different lighting), their trip mission was to buy a suggested set of items for a party of 15 people. Based on the shoppers’ actions and reactions, a Shopper Experience Quotient (SEQ) is generated.
Figure 12 – Elements (Drivers) of Shopper Experience Quotient
The Shopper Experience Quotient incorporates many elements, such as a survey asking the shopper specific questions. It looks comprehensively at the virtual shopping trip that each person took, examining things like dwell times, as well as the emotional response through both facial and voice sentiment analysis (using the camera and microphone on the virtual shopper’s computer). These technologies allow rapid testing of many variations of merchandising and store layouts at a speed and quantity that were not previously possible.
Delivering Fresh and Local
Symphony RetailAI’s largest customer sector has always been grocery, though they have other kinds of retailers and CPG customers as well. That is why managing freshness (fresh foods, produce, fresh prepared, etc.) and winning the ‘freshness wars’3 is key for many of Symphony’s customers. The challenges of managing fresh foods is multiplied when a grocer is trying to source locally. Dealing with very few large, reliable suppliers, who you know can provide the volumes required nationwide is much simpler than having to deal with hundreds of small local farms. This is where an AI solution can help tame the complexityof local fresh food sourcing,ensuring proper pricing and continuity of supply without requiring an army of people to manage it.
Farmstead is a good example of the trend. They provide fresh local food via home delivery with three key value propositions: 1) Same price delivered as at the store; 2) Fast, reliable delivery, with a promise of ‘free delivery, forever’ — order in the afternoon and get delivery by dinner time; and 3) Refill staples each week, never run out of milk, eggs, bread, and basic staples — you select the refill items and date. Farmstead has serviced over 100K customers.
CINDE Keeps Getting Smarter
There are many things that Symphony provides that we did not touch on here. They told us that their AI-driven forecasting is improving forecast accuracies by 10%-15% for their customers — that’s money in the bank. They have an excellent set of supply chain tools and other capabilities as well. This combination of a comprehensive suite of capabilities with advanced meaningful AI capabilities represents a unique and differentiated solution offering. And, it keeps getting better as CINDE is continuously learning new skills (e.g. out-of-stock detection/management, pricing optimization, etc.) and becoming smarter each month. This bodes well for Symphony RetailAI and its customers.
_________________________________________________________
1 Walmart took over the top spot as the nation’s largest grocer more than a decade ago. — Return to article text above
2 Symphony EYC was founded in 2001 to provide retail customer engagement analytics. Symphony Gold was formed by the 2012 acquisition of Aldata Solutions (founded in 1988) , providing assortment planning, pricing, inventory management and more to CPG companies and retailers. — Return to article text above
3 For more on the role of freshness in the grocery sector, see Winning the Freshness Wars: Creating Shopper Loyalty and Profitability in Retail Grocery — Return to article text above
To view other articles from this issue of the brief, click here.