Decision making in retail is broken! While most retailers want to be a lot more data-driven in their decision making, there are too many system impediments, data accessibility and even cultural obstacles preventing that from happening.
Over the past decade, decision making in retail has been supported by an array of tools including data warehouses, report generators, business intelligence (BI) tools and even Excel. Regrettably, these tools simply aren’t meeting retailer’s needs, especially for small and medium-sized retailers because they are expensive, difficult, complex and don’t provide actionable insights.
Expensive — when one adds up the various costs associated with data warehouse tools, ETL (extract, transform, load) tools, reporting, spreadsheets and visualization tools, along with the labor costs associated with data analysts and information technology resources, the costs can be quite expensive.
Difficult — it is common for the deployment of data warehouses and business intelligence to be severely underestimated. In retail verticals, implementations are always difficult taking at least six months and more likely twelve months to stand up due to the fragmented sources of data (software applications) that need to be integrated, and the resources needed to deploy these tools.
Complex — as many as 70% of data warehousing and business intelligence projects in retail fail. They don’t satisfy their original objectives, and even when a successful project occurs, four out of five business users still prefer to depend on gut instincts rather than using the tools provided. User adoption is hampered by tool complexity, the lack of training, and poor system performance.
No insights — Traditional business intelligence tools are great at churning out a lot of data and feeding that data to users. Most users end up getting lost in voluminous reports, data overload, and charts that aren’t actionable.
The inadequacy of traditional business intelligence tools results in four common challenges for decision makers and their businesses.
- Decision making in retail is slow, laborious, and inaccurate. Despite having access to these tools, decision makers end up developing a dependency on data analysts and technology professionals to extract clean and useful information. Often, getting the right outputs from analysts and technologists can be a time-consuming, iterative process because while an initial question may be answered relatively quickly, there are always additional, contextual questions that emerge when the decision maker sees initial outputs.
- Source data is siloed and not accessible for timely decision making in Retail. Decision makers require synthesized data from multiple systems which today are not well-integrated and usually provided by different software vendors. For example, it is common for retailers to have separate systems in place for point-of-sale, eCommerce and loyalty, though all may have important insights on shopper engagement.
- There is no east or fast way to get good answers. Spreadsheets and traditional reporting tools are not sufficient, insightful or actionable. They typically only provide a complicated look back on past business results that can’t be changed. Spreadsheets and transitional reporting tools fail to explain why results are good or bad and what should be done to improve results.
- Even the largest retailers that have more resources and tools usually have too few data scientists and analysts working with overly complicated tools to support their decision making. Small and medium-sized retailers, with limited budget and resources, are at a huge disadvantage when it comes to having sufficient decision making tools and expert resources to support the tools. Every organization has far more decision makers than they have analysts to support them.
85% of all retailer decisions being made based on gut instincts versus leveraging traditional business intelligence tools. As a result, retailers are leaving significant money on the table from missed revenue opportunities or unrealized cost savings. There is nothing more costly than a BAD or LATE “gut instinct” decision, and nothing better than a data-driven GREAT decision made FAST.
So, if we understand this so clearly, why do we continue to struggle with decision making tools?
Many retail enterprises have been stalled with too much complexity in their environment. The chart below illustrates this complexity starting with various, non-integrated data sources and silos on the left, with the complicated and expensive tools and resources needed in the middle, to finally ending up in the hands of business users on the right.
Beyond the complexity cited above, today’s decision making in retail is further impacted by the digital transformation taking place in retail. Shopper expectations are increasing where they expect personalization at all levels, faster fulfillment options, and they expect unified omni-channel experiences no matter how or when they choose to shop. Competitive pressures are also mounting, and the definition of commerce continues evolving.
Tech-enabled shopping, particularly online and mobile shopping, are creating an exponential growth in data interactions, touch points and velocity. What retailers have historically been doing around decision making in retail wasn’t working, and digital transformation is only exacerbating an impossible situation.
What if data-driven decision making in retail could be done differently?
What if retailers could transition from the current state to a better desired state?
Artificial Intelligence (AI) is Changing Decision Making in Retail Forever
Albert Einstein is credited as saying…
“Insanity is doing the same thing over and over again and expecting different results.”
Artificial Intelligence gives us the opportunity to take a new approach.
What if decision makers could simply ask a system their business question in plain English using their mobile phone, a tablet, a computer or a home speaker (e.g. Google Home, Amazon Alexa), and receive a correct, context-aware answer immediately? This is the world that retail winners will evolve to, an Artificial Intelligence Analytics world. It will obsolete most traditional tools used for decision making in retail with faster, simpler, and cheaper results. What are the next generation components of this new world?
- It starts with the new platform being able to automatically ingest any amount of data from various, multiple sources leveraging automated machine learning (AutoML). In retail, it is common for there to be eight to twelve primary technology systems ranging from point-of-sale, eCommerce, loyalty, merchandising, marketing, labor, finance and supply chain systems.
- AutoML also helps to “stitch” disparate data together into a cohesive unified data fabric. This unified data fabric essentially replaces the necessity for a traditional, pre-structured data warehouse. Data ingested can be automatically discovered and defined by the system and AI can help identify relationships that exist across different data sources. No need for expensive ETL tools or laborious data mapping.
- Once the unified data fabric is created, automated reflective, predictive and prescriptive insights can be derived from signals that exist in the data. Think of data signals as exceptions, correlations, causations, clusters, cross-effects and predictions. These signals are the fuel for AI-based intelligence engines and data science capabilities such as Price, Promotion, Product and Placement analytics (a.k.a. 4Ps), Customer Acquisition, Activation, Retention, Referral and Revenue analytics (AARRR), and personalized time and location-aware shopper/channel analytics.
- A modern AI solution will leverage the latest in cloud-based in-memory data management and elastic performance workload management designed to provide instant answers and insights to users…no waiting!
- A well-designed AI-Analytics solution will insulate the business user from all of the complexity and sophistication that is running in the background. A business user should be able to ask questions in simple English through text or voice interactions, with a simple “Google-like” experience. This is accomplished through the Natural Language Processing (NLP) and Natural Language Understanding (NLU). The system should offer an integrated virtual assistant to support the business user in obtaining relevant results and insights quickly without the need for a data analyst.
- A good litmus test for a well-designed AI-analytics interface would be that the business user can access the system anywhere, anytime and through a variety of touch points. Beyond the newer search and conversational text/voice capabilities, the analytics system should also support traditional dashboards, alerts, and export capabilities.
A next generation AI-Analytics decisioning solution should be able to tell retailers precisely how their business is doing across various departments and metrics far better than traditional business intelligence solutions can do. More importantly, the AI platform can explain why something happened (causation), and what you should do about it. Traditional business intelligence can’t do this!
What are examples of what can be done better with predictive and prescriptive AI-analytics capabilities?
AI Demand Forecasting
Accurately predicting sales, item movement and profits with awareness of pricing, promotions, assortment, seasonality, weather, and other consumer demand-driving factors. AI Demand Forecasting can provide highly accurate day-of-week and time-of-day forecasting at an item/store level.
AI Production Planning
Based on AI Demand Forecasting, understanding when to prepare exactly the right quantity of food items in advance of consumer demand, i.e. to satisfy the breakfast, lunch or dinner crowd. Preparing enough to satisfy consumer demand avoiding lost sales attributed to out-of-stock, but not too much either to avoid shrink/spoilage.
AI Pricing and Promotions
Based on AI Demand Forecasting, evaluating pricing and promotion effectiveness with awareness of competitor intelligence, item cost changes, and anticipated shopper sensitivity to price or promotional changes.
AI Assortment Evaluation
Based on AI Demand Forecasting, evaluating item additions and deletions based on their contribution and impact to a category or subcategory. Determine what items should be added or deleted from the assortment with advanced awareness of how consumers will react to the changes.
AI Basket Analytics
Understand the correlation/relationship that items have in shopper’s baskets, i.e. when a shopper buys a given item, they always certain other items, or when they buy a given item, it is at the expense of a similar item.
AI Test and Control
Helping retailers to evaluate the risk of a given initiative by testing the concept in advance with a subset of stores, or shoppers, measuring and tuning the impact the initiative before rolling it out to other stores or shoppers.
It is no longer sufficient for retailers to use outdated tools to make critical decisions. Time-starved decision makers no longer want to be overloaded with reports that merely look backwards. They aren’t interested in getting contradictory findings from disparate best-of-breed solutions that are poorly integrated. Instead, retail decision makers are demanding succinct, visual, actionable insights delivered in an easy-to-use, self-service fashion. The expect comprehensive, predictive (forward looking) insights with prescriptive recommendations.
Retail is a target-rich environment for AI Analytics. Retail leaders will increasingly begin to use these next generation capabilities to be more data-driven across merchandising, marketing, operations, finance and supply chain. AI Analytics will help retailers to drive cost savings, profitable revenue growth, improved customer experiences and loyalty. AI-based analytics are available now. They are very affordable and quickly implementable.
✎ by Todd Michaud