Follow Datanami:
May 11, 2020

The New Omnichannel Imperative: AI to the Rescue

Vikram Murthi


Consumers in the United States have been increasingly turning to e-commerce during the COVID-19 outbreak, due to social distance measures and shelter-in-place orders that started rolling out in the mid-late March time frame. This has resulted in online sales increase by 25% overall in mid-March compared to early March, as published in the Adobe Digital Economy Index. In addition, online grocery in the US has seen a 100% boost in daily online sales in the same period. In the UK, e-commerce overall has grown by 33%.

While consumers have been shifting their purchasing more to online from brick-and-mortar over the past few years, the pandemic has accelerated this shift. These elevated levels of online shopping in the U.S. will likely continue as long as shelter-in-place orders remain in effect. As consumers become used to online shopping with various fulfillment options, like pick up at store or home delivery, there is a likely a large portion of shoppers that get used to the convenience and will stick with it, even when social distancing measures are removed.

A Step Change in the E-Commerce Share of Retail

Onlines sales have been growing at a blistering pace for the last decade and was 11.1% of all retail sales in 2019, according to the 4th Quarter 2019 Retail E-Commerce Sales Report. Estimates are that half of all gains in retail spend in Q3 2019 came from revenue generated online, according to Digital Commerce 360.

Given the sustained growth of online sales as well as the recent surge due to COVID-19, retailers will have to get ready for new level of e-commerce – perhaps even as high as 16-20% of all retail sales by 2022, which is a step change over the projection of 13.9% for that year, according to Statista.


The new Omnichannel imperative for retailers is to accelerate the adaption of their supply chains, which were designed for a brick-and-mortar world and now need to address a wide variety of challenges introduced by e-commerce. These challenges range from optimal network design (location and capacity of DCs, ecommerce fulfillment centers, dark stores),  efficient transportation lanes and flows, SKU assortments relevant to consumer demographics, pricing strategies to sustain margin pressures, promotion calendars that span brick-and-mortar and e-commerce, more accurate operational forecasts, robust inventory and replenishment planning and meeting the high service expectations from hard-to-please consumers.

Adapting to the new Omnichannel Landscape

Artificial Intelligence (AI) is an umbrella term that encompasses a whole range of algorithms supporting business processes like optimizing web search, targeting advertisements, approving consumer loans, routing delivery trucks, forecasting consumer demand and allocating inventory. AI includes a wide range of mathematical and modeling techniques like optimization, regression, linear programming, machine learning (ML), neural networks and reinforcement learning.

AI offers one of the most significant opportunities for retailers to respond to this step change in e-commerce. Retailers across the world have been investing in this advanced technology to improve the customer experience, while at the same time increasing operational efficiency. It is estimated that global annual spending on AI by retailers will exceed $7.3 billion by 2022, according to this article in Retail Dive.

We offer some clarity around which use cases are most relevant to address the omnichannel imperative in retail supply chains and unlock AI’s full potential.

1 – Supply Chain Design

A foundational piece to address the coming step change in omnichannel business is the design of the supply chain network. The network of central/regional distribution centers, dark stores and traditional stores needs be able to fulfill orders through methods like ship from DC, ship from store or pickup from store. Traditional AI techniques such as linear programming and mixed  integer programming can be leveraged to design the fulfillment network:


  • Determine the optimal number and location of warehouses and e-fulfillment nodes to meet consumer demand using the existing network or start a greenfield evaluation if necessary
  • Understand optimal SKU-Location mapping with network product flows and stocking levels to meet projected demand
  • Plan for warehouse and transportation capacity including the ability to meet surges in seasonal demand
  • Determine the required capacity and product flows to handle returns
  • Optimal labor planning at fulfillment centers to meet omnichannel demand

2 – Demand Sensing

To succeed and thrive in what is likely to be the new normal with a step change in omnichannel demand, retailers will benefit by being more demand driven and orchestrate their supply chains to fulfill increasingly volatile demand. Retailers need to predict where demand will occur, across brick-and-mortar and online channels, and efficiently fulfill the right quantity of products to thousands and even millions of locations.

Demand Sensing addresses the critical need for the retail supply chain to respond quickly to changing consumer demand patterns, by leveraging newer mathematical techniques that enable pattern recognition with Machine Learning (ML), while overcoming the latency issues associated with traditional time-series statistical methods. Demand Sensing focuses on eliminating supply chain lags by continuously learning and reducing the time between demand signals (order frequency, order size, DC/store inventory, POS, …..) and the response to those signals.

Forecasting with demand sensing  techniques typically leverage actual sell-thru at the point of consumption, whether it’s at a physical store or an e-commerce channel. An accurate and responsive sell-thru forecast enables the complex orchestration of the end-to-end supply chain, so that the right item is at the right location, at the right time and in the right quantity. The more accurate the sell-thru forecast, the more efficient the supply chain can respond to ensure the highest customer service, with the lowest investment in working capital.

Demand Sensing can leverage ML techniques to improve the accuracy of the omnichannel forecast in a number of ways including (a)  the use of newer algorithms like Gradient Boosting and Support Vector Machines (b) leverage internal causals like regular price, placement, offers, digital coupons (c) incorporate a host of external causals like weather, GDP, new housing starts, interest rates, inflation, debt to income ratios, etc …..

3- Demand Shaping and the Merchandising Calendar

Omnichannel demand shaping activities such as placement on the web site, special offers like free shipping, markdowns, email offers, digital coupons, and social media campaigns can drive sales.


Robust modeling of these demand shaping activities can greatly benefit from ML techniques. Omnichannel category managers can run “what-ifs” – look at the impact of changing the timing and duration of promotions, try different product placement strategies on the web site, or discounts, or free shipping and review the impact on expected online orders. The expected demand can be broken out by fulfillment method (ship from store, pick up at store, ship from DC) to drive inventory replenishment needed to delivery high customer service.

Optimization of an entire seasonal promotion calendar can provide inputs to a smart advisor  that recommends the sequence and duration of demand shaping activities, given inventory constraints, total landed costs of goods, and desirable objectives like maximizing revenue or profit.

4 – The Promise of AI for Omnichannel Fulfillment

The best supply chain design and robust demand sensing and shaping would be of limited value, if the item the consumer wants is not available, or if a buy online, pickup in store (BOPIS) order is not ready in time. This is where the application of AI/ML techniques can be very valuable.

  • Diagnostic: ML techniques can identify the root causes for fulfillment failure. There could be one or a chain of root causes that lead to fulfillment failure – items not in store and have to be procured from a regional DC, workforce unavailable for BOPIS orders, missing items at pick locations, limited pickup window duration, excess number/variety of items in the order or inadequate retail store space.
  • Predictive: Once the root causes have been diagnosed with predictive weights for order fulfillment outcomes, the order book can be looked at and ML can then predict which ones are in jeopardy. Alerts can be sent to various roles in the fulfillment supply chain so that this can be addressed.
  • Prescriptive: ML can also recommend an action or a sequence of actions to the appropriate parties so that fulfillment execution is enhanced.

The application of AI for supply chain design, demand sensing, demand shaping and supply chain operations has the potential to materially move the needle on omnichannel fulfillment.

via Shutterstock

Consider a scenario in the complex and highly seasonal fashion retail sector – planning, merchandising, sourcing, allocation and supply chain teams work up to a year in advance, taking into account a plethora of factors, to get the right items to the right locations. Planning is done across an array of tools (often in Excel) with incomplete data and robust  analysis too tedious, and therefore often limited to high-value products, or done at a summary level. Quite often, the business cannot detect shifts in demand signals in time to respond effectively.

As a result, when the product arrives at the DC, stores or e-commerce fulfillment centers, it is often not where the consumer demand is — leading to expensive inventory reallocation or markdowns.

If this process was orchestrated through AI as described earlier, it might look quite different. At the start of the season, merchandisers select products and quantities to order based on recommendations from an AI engine that has knowledge of historical sales, emerging market trends, macroeconomic factors, new product attributes and category strategies. Then the planning teams review suggestions the system has made on how to distribute the products across physical stores and online channels, to optimize sell-through and minimize markdowns.

In fashion retail, the items may be sourced six to 12 months in advance, so the AI-powered system continuously recommends that planners adjust where the inventory is allocated as it arrives, based on constantly updated forecasts, while always aiming to get the items as close as possible to the consumer. Once the items are set to arrive in stores, the system suggests adjusting store-level staffing based actual arrival dates (taking into account shipping delays), and re-optimizes the logistics plan to continue reducing the cost of delivering the inventory to stores.

Finally, as the product begins to sell, slower than originally expected in a specific region, the system informs planners that a localized markdown is the best approach to maximize margins and reduce potential overstocks at the end of the season. Throughout the entire process, the AI-powered system is augmenting the work of the merchandising, planning, and supply chain teams and keeping them informed with exceptions and recommendations on how to resolve issues  with minimal impact.

AI Powers the Monday Morning Jumpstart

A common occurrence at many retailers around the world when Monday morning rolls around is to see sales and operations teams frantically piecing together the previous week’s performance. They scramble to put together the sales revenue, measure the effectiveness of promotions and markdowns, identify products that beat their targets or underperformed, uncover inventory issues and assess labor shortages.

A lot of blood, sweat, and Excel is devoted to root cause analysis of operational issues. This weekly ritual involves teams from every corner of the business, and often takes up the full day (or two), and frequently does not deliver the answers that are needed in time to take action.

Now, imagine that Mondays are driven by AI systems. By the time the teams arrive in their offices, the AI engine has done all that number-crunching, highlighted exceptions and served up recommendations to address sales weaknesses, reduce stockouts and grow margins. Instead of Monday morning debates over conflicting numbers and pointing fingers, retailers get action-oriented recommendations to move forward immediately, and more time to focus on how to delight the consumer.

While the above may sound like a dream that is many years out, many retailers in the grocery, apparel, beauty, electronics and luxury segments are already using AI to drive significant improvements to their businesses. They are able to layer AI systems on top of existing platforms and leverage existing investments to deliver insights and decisioning in a matter of months — without the need for multi-year transformations or science experiments that do not scale beyond pilot efforts.

Retailers that are primarily brick-and-mortar today, will need to prepare for the coming step change in consumer preference for omnichannel, and leverage AI systems for supply chain design, demand sensing, demand shaping and fulfillment.

About the author: Vikram Murthi is the vice president of industry strategy at LLamasoft, where he engages with companies to understand their supply chain challenges and helps shape their investment strategy and transformation roadmaps. As a supply chain thought leader, Vikram paints a compelling vision of the future with an industry specific point of view and presents at conferences and writes white papers, blogs and articles. Vikram has a B.Tech in Electrical Engineering from Indian Institute of Technology (IIT) Kanpur, India and an M.S. in Computer and Systems Engineering from Rensselaer Polytechnic Institute, USA.

Related Items:

How AI Is Helping the Supply Chain Cope With COVID-19

Unlocking Retail’s Big Data Opportunity

How Retailers Use Big Data to Gobble Up Sales