Beyond traditional: Grocers need to use new forecasting methods to stock the digital shelf
Anush Viswanathan is an account manager at Ugam, a global data and analytics company. To learn more about the application of data and analytics to make confident business decisions, visit www.ugamsolutions.com
This year has been unkind to many traditional retailers. Those that have struggled to adapt to the shifting consumer behaviors and preferences have been forced to close their doors, unable to meet shoppers’ demands. Unfortunately, many traditional grocery stores are in danger of following suit.
With newer non-traditional channels like delivery apps and meal-kit services fueling the growth of omnichannel grocery, traditional stores must adapt or be left behind. For grocers to best prepare for the disruption, it’s critical they update the way they forecast demand so they can carry the right supply for each channel. Grocery is a hyper competitive industry, and given the complexity of coordinating multiple channels, we’ve seen that low forecast accuracy can shave 10% to 15% off of the bottom line. Getting an accurate, detailed forecast at scale is table stakes today in order to compete with the Amazons and Instacarts of the world.
Loyalty in grocery is dying, so a bad omnichannel experience means losing customers. Inaccurate forecasts can lead to unfulfilled orders and missing items, hampering customer experience. To avoid this, store owners should be aware of several hiccups their stores might face when forecasting omnichannel demand.
As omnichannel grocery takes off, retailers’ biggest issue is trying to get a handle on localized online demand. For example, it’s not uncommon for a customer to order online from his office but want the delivery sent to his home. Thus the demand is generated in a different place than the fulfillment location. Grocers need to figure this out and are trying various techniques to supplement in-store forecasts with online demand.
An additional challenge is the granularity required to be flexible in the supply chain. In the offline world, a daily level forecast was enough in terms of sophistication. However in an omnichannel world, intra-day forecasts to hourly forecasts are required, since logistics and staffing need to be planned around fulfilling online demand through the stores. The more granular it gets, the tougher it becomes to get very accurate.
Grocers deal with the risk of wasting food, leading to markdowns and margin loss due to incorrect forecasting. While this was always a risk, it has magnified due to the erratic demand patterns associated with online. We’ve seen customers primarily tend to use online as a channel for convenience and urgency when it comes to their grocery needs. Both come together to create demand patterns that are not suitable for forecasting using off-the-shelf algorithms.
To combat these obstacles, grocers in particular will need to think beyond off-the-shelf algorithms and in-store demand forecasts, and become more sophisticated in their machine learning capabilities to compete with the tech-led disruption happening in online grocery.
One leading grocer we’ve worked with is launching a hybrid forecasting approach, combining off-the-shelf algorithms with custom machine learning-based reinforcements. With this approach, a store manager has access to the volume of tomatoes required to be placed on the shelf vs. how many are required to pack for online orders, allowing for better labor planning. The retailer was able to improve in-store availability for some categories by up to 30% as a result.
We’ve seen such programs require a strong cultural change in the organization and can take two to three years to perfect. Given the longer lead times, grocers need to accelerate their journey now. Grocers that get accurate, localized omnichannel demand right now will come out ahead.