While COVID-19 has had a catastrophic impact on health and the economy, it has also fundamentally changed how people live and work in the foreseeable future. Successful retailers often take great pride in their deep understanding of the consumer’s psyche and buying behaviour. In the last decade or so, they have leveraged troves of data for successful decision-making. However, such institutional knowledge has lost its relevance in a matter of few weeks.

Each part of a retailer’s value chain is impacted. Let’s first talk about customers and demand. The first major shift is the decline in consumer spending owing to uncertainties in the job. A recent survey by London Business School indicates that household spending has declined by almost 40-50% [1] in the UK, and similar trends are emerging across the globe. This will have a significant impact on the Retail industry overall. Similarly, the ‘Work from Home’ model will impact the Fashion and Hospitality business.

The economies of scale that big-box retailers leveraged have been constrained to a large extent. Retailers would have to learn to optimize their operations in order to maintain their bottom-line. Density or footfall used to be the key measure in brick-and-mortar retailing. Most retailers measure the efficiency of stores by sales per sq. ft. This will change to comply with government guidelines and ensure safety. Customer and Patient safety and compliance related to public health are added to day-to-day operations.

Figure: 1 – Impact of COVID on the retail industry

These uncertainties have made demand functions and operations strategies void. Since the past is no longer the best predictor of the future, retailers need to start afresh and learn from the present as if they are starting for the first time. When governments around the world are poised to ease restrictions, retailers have several key questions to answer before they fully operationalize. 

Figure: 2 – Questions to which retailers are looking for answers

Key Challenges that Data & Analytics Organizations Need to Solve for Retailers

Data & Analytics have been embedded across the Retail Value chain. The evolving situation has certain constraints and is challenging the solutions built using Data & Analytics.

• Existing data pipelines do not carry COVID-related information – COVID-related information needs to be brought into the enterprise and it must be integrated with existing store, depot, customer, supplier, and 3^rd-party information.
• Integration of Public Data Sources & Data Quality concerns – Most of the COVID-related information is available on public domains from government or local agencies. This localized information is indispensable to plan operations, but they may not be available in a consumable format for retailers to source and may have data quality issues.
• Current reports and dashboards don’t carry lockdown and pandemic-related insights –Business needs an integrated view of the existing data along with COVID-related information to take the right decisions. Key performance dashboards and reports across the organization must be updated to provide the right context to business.
• Data processing is mostly in Batch than Real-time -  Most of the dashboards and reports are built from the data warehouse, which is populated from daily/weekly batch processes. However, COVID is a rapidly evolving scenario and the business’s need is to have access to near real-time information, often several times a day.
• Existing customer and product segmentations do not factor in the impact of the pandemic – Retailers generally take their merchandizing decisions based on customer segments and personas. However, the propensity to purchase for different segments has changed differently. For e.g. customers who used to buy premium segments might think of buying value-for-money items instead.
• Concept drift in the predictive models – Even the most rigorously stress-tested machine-learning models are not trained to handle an outlier scenario like COVID. Concept drift is common and models may throw erroneous results.
• Not enough data to build new models – One of the key challenges in front of data science teams is the lack of COVID-like data to train predictive models.

New Normal for Retail Analytics

The new normal for Retail Analytics should start WITH identifying relevant use-cases across THE Retail value chain, AND Setting up the appropriate Analytics framework THAT can deliver required insights Using the right data.

Re-Imagine the Retail Business Process

• Research & Segment - Understanding the impact of COVID on various customer segments and identifying the changes in shopping patterns is the first step toward demand forecasting and inventory management.  It is also the time to critically evaluate the relevance of your value proposition. Shopper surveys and social media data can help you identify the right hypothesis. Analytics teams should work with business to develop and prioritize these hypotheses, so that they can be tested in a time-bound manner.
• Plan & Buy - In the globalized world, planning and buying is a long cycle where Retailers are dependent on many factories across the globe. The current situation has disrupted the Buying cycle where orders are canceled, or factories are shut down across geographies. The entire Planning & Buying cycle needs to be restarted in small iterations and ramped up based on Store reopening, demand, and feasibility of factories that are able to operate.
• Move - Governments across the world have imposed lockdowns and restrictions on non-essential goods movement.  Factories and freight forwarders are no longer working the same way they used to. Retailers should leverage public data sources on the global spread and forecasts to identify and monitor supply chain risks. Typically, retailers were dependent on Weather data for supply chain disruptions. Now, they need to integrate COVID-specific information in their procurement and logistics planning.

• Price, Promote, and Mark Down – Pricing is the key to bringing the customer back to the store when it reopens and trying to clear up stocks. Conventional price elasticity functions won’t be relevant, and retailers must strive to find the right marketing mix. Rapid experimentation is the key to identify the right price point with the focus on key product categories. Retailers should leverage analytics to identify products and appropriate customer segments where margin can still be maintained.
• Sell – Retailers should re-imagine the sell process like a new start-up, unlearn the old ways, and rediscover selling at each store locally. Over the years, retailers have decentralized decision-making, giving more freedom to the store/area/category managers. The impact of COVID is extremely localized, even to the extent that two stores within the vicinity may be impacted differently. Localized decisions expedite the decision-making process. The power of integrated analytics and insights should be made accessible to frontline staff to support real-time decision-making.
• Retail Operations - As a part of easing of lockdowns and reopening, governments have defined guidelines e.g. max count of shoppers in the store, count of employees who are tasked with enforcing social distancing, visibility of the signage and queue information etc. This information needs to be collected from stores and reviewed periodically to ensure compliance. Employees who are more prone to COVID can be identified and moved to areas of work where there is less human contact and chance of infection.
• Customer Engagement – Communication on health, safety, and store opening is key to bring customers back to the store. Targeted messages on product availability should be sent to appropriate customer segments.

An Analytics Framework to Deliver

The decision of reopening should start with a clear set of objectives. The objective can vary from getting back to your most valuable customers as soon as possible to clearing the stocks before the start of the next season. We advocate granularity as a key trait for successful decisioning at times like these. In a rapidly evolving environment like this, it may not be realistic to set targets for months and quarters ahead. The time horizon for decisioning needs to be daily or weekly. 

Figure: 3 Decisioning framework

Once the objectives are agreed upon and communicated, simulations can be developed for different likely scenarios and the outcomes of the decisions on the objectives. This can be made visible to executives and decision-makers through interactive dashboards to enable efficient decision-making.

Each decision should be treated as an experiment. The data related to outcomes should be collated centrally and the insights should be shared with relevant stakeholders. This is indispensable in the context of decentralized decision-making as the learning from the decisions from one area / product category are made available to others in a consumable fashion.

Sourcing the Right Data

The foremost challenge that the Data & Analytics organization is confronted with is to source the data related to the spread of the pandemic. There are several third parties, government agencies, research organizations, and technology providers that have already collated the data related to COVID, which can be sourced and used. The covid dashboard[2] prepared by Johns Hopkins University contains a wealth of information. The United States government[3] and National Institutes of Health[4] have made several data sources public, which can be leveraged to understand the spread of the pandemic.

Figure: 4 - Sources of data that can be leveraged

Amazon (AWS) has created a data lake specifically for COVID[5], consolidating information from multiple sources and made it ready for computation. Similarly, Tableau has created multiple COVID dashboards[6], which are accessible to the public. In addition, several organizations are making COVID-specific data available through APIs.

In times like these, when historical data alone does not serve the purpose, it is a good idea to reach out to customers directly to seek their opinion. Surveys are a very potent way to sense customer demand and buying preferences. Customer surveys can be conducted to get a sense of their expectations on assortment and price points, and potential disposable spending. Frontline employees often have an intuitive understanding of the customer’s pulse for a product category in a region. Employee surveys can be conducted to consolidate this knowledge in a centralized way and present it for the entire organization. Employee surveys are also a great way to understand the potential health risks because associated with COVID.

Data from social media feeds is another channel to understand customer sentiments and product preferences. The response to COVID needs to be dynamic with access to real-time information.  Retailers, over the years, have built data platforms e.g. Operational Data Stores for low latency and granular data. Such platforms can be augmented with additional COVID and survey data to make data available in real time.

Accelerating Data & Analytics Strategy during COVID

The data & analytics organization is critical to the reopening and revival of the retailer. Therefore, it is paramount that the business objectives be aligned with technology objectives. In addition, technology initiatives must deliver outcomes within resource and time constraints.

1. Leverage existing resource pool
2. Leverage existing technology investments
3. Shift your cost from capex to opex
4. Take the opportunity to shape your future as well - tactical vs. strategic
5. Agile delivery. MVP methodology. 

The knowledge contained within the existing resources and teams should be leveraged to find creative solutions to data and technology challenges. While ideas can be sourced from each, separate cross-functional teams can be formed and entrusted to deliver on specific objectives. Securing any additional funding for new technology investments at this stage can be extremely challenging. The team should leverage the existing technology stack to the maximum extent possible. As the longevity of the solutions are not known, the developer’s time should be spent on actual development rather than documentation, review, and approvals.  As the situation is extremely volatile, with new information coming up every day, all developments must go through an MVP-based iteration. The cost should be moved from capex to opex. It is often said that crisis brings the best out of human beings. So while trying to respond to COVID, data and analytics organizations should not lose sight of the future. In fact, they can use the time to resolve operational silos that prevented sourcing of data, thereby improve the efficiency within the team, and innovate.

References

1. https://www.ft.com/content/3180a1b3-ad68-4d7f-ac81-b762162bab98
2. https://coronavirus.jhu.edu/map.html
3. https://www.coronavirus.gov/
4. https://datascience.nih.gov/covid-19-open-access-resources
5. https://dj2taa9i652rf.cloudfront.net/
6. https://www.tableau.com/covid-19-coronavirus-data-resources

Karthikeyan Subramaniam

Managing Consultant in Wipro’s Data, Analytics & AI (DAAI) Practice.

He is currently based out of Wipro’s Reading office in the UK. He can be reached at skarthikeyan.subramaniam@wipro.com.

Deepak Kumar Dash

Managing Consultant in Wipro’s Data, Analytics & AI (DAAI) Practice.

He currently works out of Wipro’s Reading office in the UK. He can be reached at Deepak.dash1@wipro.com.