Is there anything more frustrating than walking through the aisles of your local supermarket or convenience store trying to check items off your shopping list only to find empty space where your item should be? You can see it is supposed to be positioned there, where you always find it, but none are available to select. Could it be misplaced on a nearby shelf, maybe some in the back to be restocked, maybe completely sold out? There is no way to tell. That’s because there seems to be a gap between the shopping list and what is actually available in that particular store.
This is a global phenomenon that impacts retailers’ profits and shoppers’ loyalties
Let’s understand how big the problem is.
Assume a retail chain with 500 supermarkets generating an annual revenue of $10 billion, and at any given time, the chain is tracking to a metric of 5% out of stock (OOS). Let’s drill down a bit. If the active assortment for that chain is 50,000 SKUs, 5% OOS means they are not able to accurately replenish around 2,500 SKUs at any point of time. Out of this set of OOS SKUs, maybe 10% of those 2,500 are actively impact the core basket for an average daily customer. We know that customers have brand affinity and while some degree of brand substitution happens often, they will not buy a substitute brand. This is a clear loss of sale for this retailer.
Over time, this constant OOS situation creates a negative perception about the supermarket chain in customers’ minds – why is something always missing from the list. Gradually the trust associated with that chain starts to erode, and at some point, customers start to switch their loyalty to another chain where the OOS problem is relatively less.
Coming back to the example chain of stores, let us assume that this OOS problem leads to a loss of sales of 3% of revenue translating to a staggering $300 million. Supermarkets typically operate around 20% gross margin, including rebates and discounts. This means the supermarket has missed $60 million in gross margin due to its OOS situation.
Let us see what this loss of profit means in the context of the IT budget of this retailer. Average IT spending in retail is between 0.75-1.5% of revenue. For the retailer, let’s assume it is 1%. This means the retailer spends $100 million annually toward technology (hardware, software, manpower etc.) Thus, the $60 million worth of profit lost in this one critical area represents 60% of their annual IT expenditure.
Now let’s examine the key factors contributing to OOS:
1. Lack of operational compliance at store level.
This can be summarized as “human failures.” Automatic replenishment quantity for SKU X for Store Y - even where generated by an algorithm - is deeply impacted by wrong balance on hand (BOH), negative inventory, receiving and updating process issues, manual intervention in system suggested orders, direct-store delivery (DSD) process gaps, manual purchase orders, non-compliance with planogram due to store remodeling, wrong safety stock, wrong minimum display quantity and many such distortions in key parameters. Multiply this inefficiency over hundreds of stores and thousands of SKUs, and one ends up with purchase orders which are not understood by suppliers and not in line with the actual needs of stores driving inaccurate replenishment signals. As most retailers believe in distribution center-based replenishment (supplier-to-DC-to-store) and distribution center allocation algorithms are equally static in nature, OOS can only increase. Thus, data being used by the existing engines have quality issues but there is no mechanism to identify data outliers and highlight cleanup areas.
2. Lack of coordination between retailers and suppliers at both strategic and operational levels.
Many retailers have implemented planning and demand forecasting systems for communicating aggregate demand. But, this forecasting is done only with reference to key suppliers, and at that too, often only at an aggregate level. The forecast is rarely built from the ground up, especially for SKUs under promotion, which typically constitute around 40% of a supermarket’s monthly sales.
Even if the forecast model is built by taking store level intricacies, it rarely feeds the actual replenishment system with that input nor takes data from sales system on a daily basis while also auto correcting weekly forecast for replenishment quantity for SKU X for Store Y.
Moreover, retailers have limited visibility to the incoming inventory from a supplier against a specific purchase order until the supplier’s truck arrives at their distribution center and they go through a discovery session (on time - in full issue – mismatch between line fill and case fill). If such mismatches are known to the retailer early enough, they can react accordingly. Since the distribution center came to know of this imbalance /imperfection when the truck has arrived, distribution center operations will get busy in material handling (conventional flow through). The store manager, in turn, has no visibility of incoming inventory until the material arrives from the distribution center, by which time it is too late for him to react. This “in network” (CPG-retailer) inventory visibility gap is a very important operational inefficiency from the perspective of distribution center managers and individual store managers.
3. Inefficient set of algorithms built many years ago on packaged software implemented by the retailer.
Many retailers still run the packaged software for managing replenishment that they had implemented in the days when brick and mortar was the only form of supermarket retail. Naturally, their algorithms fail to reflect the current omnichannel reality. Since version upgrades are costly, often the retailer’s IT team builds custom enhancements complementing existing packaged software. Appendages are then written on top of these bespoke applications. After some time, the whole plot is lost and few employees in the IT department even know how the custom solutions are intended work much less the lack of extendibility for these point in time solutions. Moreover, such patchwork software, built in-house, does not reflect the agility and changing face of business which has become omni channel. Retailers are both unable and unwilling to pay what it takes to attract and retain advanced IT and data science resources required to support advanced calculation platforms.
4. Zero machine learning input in static algorithms.
The advancement in computing ability, which machine learning provides is ignored and not acted upon by key decision makers. Their dilemma is, “one more IT system that will cost us X million and take Z months to implement and suck away precious operational/domain time.” It is true that “data is the new oil.” But, that data is both structured and unstructured and needs to be “interpreted” in the context of the business. Static algorithms do not provide for the ever-changing reality reflected in the data that is constantly being generated. Interpreting unstructured data through Machine Learning is the only way of consuming this treasure trove of information. The intelligence generated thereby would help in formulating the right algorithm which would then generate near-perfect automatic replenishment order quantity for SKU X for Store Y.
So, how does a managed platform like inventory management-as-a-service (IAAS) help close this gap? To draw an analogy, if Henry Ford had started Ford Motors today, he would have started with electric cars.
Your enterprise and merchandising systems are really good at their transactional duties, but they were never designed to take advantage of, nor keep up with, the latest quantitative modeling and machine learning, which is the brain behind the automatic replenishment quantities IAAS generates. IAAS does not handle workflow and transaction/document management – PO/GRN/return and a multitude of other processes for which the retailer will continue to use a robust ERP/packaged software. Thus, the retailer is augmenting the investment in packaged software with a new brain focused on one thing – calculate the correct replenishment quantity for hundreds of stores covering thousands of SKUs, including promotions, imports and private label SKUs, ultra-perishable produce and regular CPG items. In short, it is designed to handle the entire assortment which the retailer carries.
The algorithms of the IAAS are constantly ingesting unstructured, contextual, relevant data along with structured data such as the Daily T Log and looking for patterns, all the while, learning and adjusting. It is this constant “learning” process which makes IAAS unique and relevant. This learning is industrial and robust and with the software-as-a-service model, the retail team need not worry about constantly adjusting or tuning the algorithms as the active management of the service team and their experts alleviate that need.
As a byproduct of this process, the retailer’s demand planners will get store-wise SKU level forecasts which can be shared with suppliers. IAAS will send predictive alerts to the buying team about assortment issues such as discontinued SKUs on shelves and draw inferences about customer preferences at the store level, thus triggering assortment rationalization. It reduces supply chain and store operations costs currently spent manually managing replenishment targets, reviews and execution. If the retailer has already implemented a central planning tool, then IAAS only complements and does not replace it.
No retailer wants their customers to leave the store with unchecked items on their list and the negative impressions left by those stark empty shelves to the tun of hundreds of millions of dollars. Deploying an actively managed IAAS solution is the best way to fulfill the promise made to those customers—having the right product on the right shelf at the right time to meet their needs.