Opportunities in Retail

By 2020, 85% of customer interactions will be managed by Artificial Intelligence. Learn more about the origins of AI and its applications to the retail industry in this report, independently prepared by EIQ.


Today’s retail industry is far more fragmented and competitive than ever. Multiple store formats and an arsenal of digital tools are making shoppers more educated about choices. Digital channels also continue growing. This is particularly true in grocery, where heavy hitters like Amazon and Walmart continue to eat into the market share of traditional chains.

The landscape has also become more diverse, with a variety of household types and lifestyles having very different needs than the mom-dad with-kids target that dominated generations past. This is compounded by a burgeoning ethnic population, with each group having a distinct profile in every area from language and food to shopping style and economic status. Add to this revitalized inner cities, which are attracting young Millennials in droves, and the result is a seismic melting pot that never stands still.

Retailers and their suppliers need real-time, in-depth knowledge to attract diverse shoppers. But the advent of distinct devices, sensors, and machine-to-machine communications has made data sets so large that timely manipulation, management and analysis present significant logistical challenges for companies using on hand data management tools or traditional data processing applications. Other entities have incorrect or outdated legacy data.

This is compounded by the difficulties many companies face in recruiting the talent needed to implement complex technology tools, analyze the data and make effective recommendations. Many successful high volume retailers and consumer packaged goods (CPG) organizations have turned to artificial intelligence (AI) to navigate the muddle.

At the simplest level, AI machines or systems imitate human behavior in intelligent ways that can augment productivity and optimize business performance. AI applications include machine learning, natural language processing (NLP) and robotics.

AI allows retailers and manufacturers to gather customer insights in an automated fashion and predict next actions based on previous patterns or images. AI uses predictive patterns to help understand desires, motivations and actions across both physical and digital channels. This lets retailers and suppliers enhance many functions, such as executing more targeted and personalized marketing campaigns and improve trade promotion efforts. AI can also automate forecasting of inventory needs, more accurately predict out-of-stock incidences and ultimately help optimize supply chains.

Types of Artificial Intelligence

With artificial intelligence (AI), machines mimic or replace intelligent human behavior, like problem solving or learning. They “sense,” “comprehend” and “act” in accordance with the real world. In essence, machines learn from experience and make recommendations, learning and improving over time.

AI applications fall under three key areas

Machine learning

Machines automatically analyze large amounts of data and “learn” using rule-based algorithms that identify patterns and trends. As an example, this could mean combining 100,000+ data points from 75 million customers regarding shopping patterns and other habits.

Natural language processing (NLP)

NLP is a machine’s ability to understand, analyze and generate human speech. A computer listens to a natural language spoken (or written) by a person, understands its meaning and responds by generating natural language to communicate back (as opposed to a computer language like Java or SQL).


Involves full-scale automation of tasks traditionally performed by humans. Warehouse picking and packing, for example, can be performed by robots.

AI’s Growth

Machine learning first became a scientific discipline in the late 1990s. But it did not seriously take off until the 2000s. Growth was fueled by access to huge amounts of real time Big Data and the emergence of algorithms that make sense of that data for productive output. AI is continuing to grow, touching more industries and functions every day.

To date, much AI retail activity has revolved around machine learning in e-commerce, particularly for search analysis, product recommendations, promotions and analyzing consumer sentiments. Amazon is regarded as a pioneer here, and it is widely estimated that 25% of its sales are generated through recommendation-based product views and previous purchases. Today, Amazon is even marketing its easy-to-use, highly scalable search and other machine learning technologies to outside parties.

Other e-commerce companies have used search and recommendation tools for some time. But in recent years, e-commerce has reached new heights by using machine learning to make functions more comprehensive and specific. User choices and information can be cross referenced in numerous ways. Customers can locate merchandise faster, more products are sold per transaction and there are fewer abandoned carts.

Retail and consumer goods are among the top five industries in which AI is being applied1.

In 2017, the global AI market was estimated at $2.4 billion. It is expected to grow at a CAGR of 50% to over $59 billion by 20252.

64% of CIOs plan to invest significantly in cognitive or AI technologies over the next two years3.

By 2030, AI will drive Global GDP gains of $15.7 trillion (14% higher) through productivity and personalization improvements4.

Now, retailers and suppliers are applying AI to areas outside e-commerce. Demand forecasting that incorporates machine learning, for example, allows online and offline retailers to generate more precise forecasts than traditional time series approaches. Machine learning also facilitates warehouse management by helping to alleviate the over- and understocking scenarios that can erode a retailer’s bottom line. When applied to trade promotions management, AI could help suppliers improve timing, tracking and other aspects of retail marketing investments.

NLP is also making inroads by providing conversational answers in areas including category management through deep analytics, data mining and visualization at department, planogram and product levels.

1 PWC 2017 data | 2 Statista | 3 Deloitte’s 2016 Global CIO survey | 4 PWC 2017 data

AI is continuing to grow, touching more industries and functions every day.

Top AI Applications in Retail

AI is gaining an important place in retail with growth expected to reach $40 billion by 2025, up from an estimated $6.46 billion today. It is being driven by an increase in customer-centric initiatives, more social media advertising and heightened demand for virtual assistants5.

Among retailers, 16% already use some form of AI, while 20% plan to add it over the next 12 months; another 18% hope to implement it more than a year from now6. Following are the leading AI, machine learning and NLP application areas in retail.

AI in Personalized Marketing

Retailers typically use marketing automation software or campaign management solutions to target customers. Applied to CRM data, tools divide customers into groups according to shopping behavior, demographics, preferences or other criteria.

Usage of AI in personalized marketing

Social sentiment analysis

Elastic personalized search

Currently use
Plan to incorporate in next 12 to 24 months

Source: 2018 EIQ Retail Innovation Survey

The problem is that rules are chosen based on the marketer’s human assumptions and leave significant room for error. The process also leaves out potentially useful criteria, and it can be hard to segment customers who do not correspond to pre-designated buckets. Since information is historic, shopping behavior, income and other factors are prone to change.

Machine learning examines a full set of data, identifies patterns and organizes it into “clusters” of similar data. Assumptions and stereotypes about what is important are bypassed. Rather, information is determined by the analysis. Trends and connections are established that might have been overlooked by analyzing individual pieces of data simultaneously, and information can be used to send highly personalized offers to customers.

But retailers believe personalization still has a way to go, even though 39% say it is extremely important7. Most retailers (54%) gave themselves a low rating for executing personalization strategies at an omnichannel level, with just 4% rating themselves as high overall. The biggest hurdle, said 69%, is lack of appropriate technologies, followed by managing across channels (47%).

Analyzing store point-of-sale and e-commerce transaction history became the standard for classifying and targeting consumer groups. Now, advances in Big Data and AI are giving rise to highly personalized campaigns and other initiatives without major human intervention. These engagement tools factor in customer purchase history, browsing behavior, social media activity and overall channel engagement. The biggest difference is that today’s initiatives target people on an individualized basis, and with AI, retailers can do this at scale.

Personalization can grow revenue 5% to 15% and increase efficiency of marketing spending by up to 30%8.

5 ReportsnReports | 6, October 9, 2017 | 7 RIS News, “Closing Big Gaps in Personalization,” October 2017 | 8 Harvard Business Review

AI in Trade Promotions Management

In promoting products, CPG companies have historically made substantial investments with retailers to boost revenue and/or increase market share. Today, trade spending represents more than 15% of CPG companies’ total revenue9 and continues to grow. Consequently, spending volume has increased dramatically, and trade promotions have become more complex and harder to manage.

What is more, it currently takes the average business user four weeks to understand if a trade promotion was effective10. Yet 72% of promotions fail to break even11, and many new products fail.

According to the Trade Promotions Management Association’s website, the industry has been reluctant to adopt new technologies. Roughly 60% of companies still use manual processes and spreadsheet applications or proprietary software. Thus they lack real-time, accurate and meaningful insights for planning, managing and optimizing trade promotions.

AI and analytics can provide promotion-related insights and guidance to channel managers, category/brand managers and financial teams to help allocate trade fund dollars more wisely and alleviate margin erosion.

Improve promotion precision with NLP

By layering on NLP, consumer goods companies can facilitate experimentation with trade promotions criteria. They can verbally ask, for example, how results could differ if a promotion is run in July versus August. The answer is immediate and does not require using a PC or running massive calculations. And they avoid expensive risks.

Given the time and money invested in Research & Development, this feedback can go a long way when it comes to trade funds marketing, determining actionable price points, tweaking items and recognizing market voids that could be filled by new products. And it can yield results faster and at a lower cost.

NLP also recognizes natural, written language. Using sentiment analysis, it can determine whether consumer reactions to products are positive, negative or neutral. Given the hordes of posts consumers make daily on social media, blogs, e-commerce sites and other platforms (including CPG companies’ own social media sites), manufacturers have a wealth of information to draw from. predicts the NLP market will roughly double, reaching $16 billion by 2021 at a CAGR of 16.1%12.

Usage of AI in trade promotions management

Pricing optimization

Promotion optimization

Currently use
Plan to incorporate in next 12 to 24 months

Source: 2018 EIQ Retail Innovation Survey

Trade Promotions Management Association, 2017 | 10 Consumer Goods Technology, “Tech Trends 2017: Redefining Trade Promotion” | 11 Nielsen price/ promotion survey, 2017 | 12 Natural Language Processing Market by Type Technologies by Deployment Type, Vertical & by Region – Global Forecast to 2021,” July 2017

AI in the Supply Chain

Machine learning has an important place in the supply chain, particularly when it comes to demand forecasting. With traditional planning methods, demand forecasts are not always accurate. This leads to out of stocks, overstocks and products being returned to vendors. It also creates unhappy customers and makes retailers unable to attain financial goals.

Machine learning helps forecast inventory, demand and supply in that predictions are not based solely on historic data. Rather, the technology predicts what will sell, driving enhanced forecasts based on real-time data using demographics, weather, performance of similar items and even online reviews and social media. Predictions can be made by store, SKU, size, color and other criteria.

Machine learning even helps identify and correct data errors and risks in the supply chain, elevates insights from the Internet of Things devices in the field and plans logistics. This optimizes delivery of merchandise while balancing supply and demand, making human analysis unnecessary.

Issues stemming from inventory management or, more broadly, supply chain planning can have a profound effect on logistics operations. For example, a retail company that is developing an omnichannel strategy struggles to manage the complex trade-offs between demand forecasting, inventory orders, channel allocation and logistics costs and capacities as it tries to respond to customer demand. Retailers may also be facing a multitude of challenges related to warehouse and DC management. In order to ensure successful execution, companies need to take an end-to-end view of the supply chain, managing the

relevant trade-offs and synchronizing planning and logistics to drive value. Adoption of machine learning in mapping varied demand patterns and scenarios for more effective inventory optimization and channel allocation could become an important step in timely replenishment and efficient logistics.

Usage of AI in supply chain management

Currently use
Plan to incorporate in next 12 to 24 months

Demand forecasting

Inventory planning


Source: 2018 EIQ Retail Innovation Survey

AI in Assortment Planning

Historically, planning a retail assortment involved looking at the previous year’s sales data to see what performed well and what did not, then factoring in new fads and trends to come up with the right mix. The result would be a combination of brand new items that followed the latest fad, a few timeless perennials and some of last year’s mix for those customers who were not quite ready for a change.

The problem is, the consumer population and its tastes and habits are a constantly moving target. The teens who shopped this retailer last year may have gone off to college. Or, if the target was working professionals, changes in the job market may have impacted their spending habits. Whatever the reason, historic data talks about yesterday’s customers.

AI-influenced algorithms can predict the most relevant items to add to a retailer’s inventory by analyzing the product assortments of competing retailers and brands, then comparing those products to the demographics and shopping history of that retailer’s customers—in

real- time. Some tools can even predict the ebb and flow for each particular product over the next 30 days, including demand changes by both percentage and item count.

Machine learning can also be used to “read” customer reviews on social media or e-commerce sites. A machine learning algorithm can be taught to categorize posts or look for text patterns, and AI can even detect foul language and fraudulent reviews.

This is particularly valuable for retailers. They can determine how the same or similar items are performing elsewhere, give them a ranking and decide if they want to order them, how many to order, how long to feature certain items, what stores to offer them in and other criteria.

As a measurement tool, online reviews are particularly valuable. Unlike a focus group or other study, they voluntarily come from individuals who have actually purchased a product. And consumers take them very seriously:

If two similar products have the same rating, shoppers will purchase the one with more reviews13.

97% of shoppers say reviews influence their buying decisions; 92% hesitate to buy anything if no customer has reviewed it14.

73% of shoppers say written reviews impress them more than star or number ratings15.

Usage of AI in assortment planning

Currently use
Plan to incorporate in next 12 to 24 months

Source: 2018 EIQ Retail Innovation Survey

Merchandise management

Customer/consumer insights

13 Psychological Science study | 14 Fan & Fuel Digital | 15 Deloitte


AI is still in its infancy. By 2020, however, 85% of customer interactions will be managed by AI16.

Thanks to Amazon and other cutting-edge retailers, AI has already made major inroads in e-commerce, particularly when it comes to more pinpointed product recommendations. This online personalization trend will only intensify as e-commerce continues to grow, customers become even smarter and more demanding and AI applications like visual search and NLP digital assistants become more widely understood and applied.

In some other areas mentioned in this report, AI has a long way to go in terms of uniform and consistent adoption. But with the cost of bringing products to market and the high failure rates, retail and CPG companies in particular have much to gain by applying the technology to trade promotions management and personalized marketing. AI is also gaining ground in assortment planning, supply chain management and product development where an endless loop of forecasting continually adjusts inventory levels. This alleviates inconsistent inventory buys, overstocking, understocking and consequent margin erosion. It also creates happy, loyal customers who keep returning due to more relevant assortments and new products, thus ultimately driving overall shopper satisfaction.

16 Gartner

Was the report interesting?

Register to download the whitepaper as a PDF and keep up to date with the latest research as it’s published.

Symphony RetailAI is the leading global provider of Artificial Intelligence-enabled decision platforms, solutions and customer-centric insights that drive validated growth for retailers and CPG manufacturers, from customer intelligence to personalized marketing, and merchandising and category management, to supply chain and retail operations.

EnsembleIQ is a premier business intelligence resource that exists to help people and their organizations succeed. We empower retailers, consumer goods manufacturers, technology vendors, marketing agencies and a vast ecosystem of service providers by leveraging an integrated network of media and information resources that inform, connect and provide actionable marketplace intelligence.