It’s midseason, and a local celebrity has just been photographed wearing one of your bright yellow dresses. Demand for that item skyrockets. It’s a huge win for the brand, but there’s a problem. No one expected the dress to be a top-selling item, so it wasn’t allocated to all the stores in the area. How do you adjust the supply chain to maximize sales and ensure that the right amount dresses are available in stores where they’ve never been before? Traditionally, replenishment algorithms require some store level data in order to be accurate. Now, however, the latest iteration of Evo Replenish can use alternative data to provide accurate sales and demand forecasts for new items introduced to stores after the first allocation.
Evo has just released this latest version of its Replenish tool, significantly expanding forecast capability for clients. Until now, Evo had two primary mechanisms to allocate inventory to stores. Allocation allowed retailers to distribute new products to stores at the start of a season based on where each new product would be most needed. Replenish would then take over, restocking that same inventory according to sales forecasts. There was no way to decide if or when to introduce new items to stores that had previously been sold only in other locations or to accurately forecast how much each store would need of those new products.
That has changed. The latest release of Replenish gives clients the option to predict when a particular product has a high likelihood of selling well in another store and to accurately allocate the correct number of SKUs based on Evo forecasts.
Replenish can now better adjust to changing local realities while still providing the accuracy and simple automation that Evo clients depend on.
Elena Marocco, Evo Supply Chain Team Lead
How the new Replenish tool works
Essentially, this new feature of Replenish works much in the same way that the original Replenish tool worked— but with a few tweaks. Replenish uses a large amount of internal and external data to forecast demand, including previous sales of that same item in the same store. Accurately predicting demand for products not previously sold in a particular store creates a challenge because there are gaps in the data.
To overcome this obstacle, the algorithm uses alternative data to forecast sales in the new function. The AI identifies similar stores and similar products and uses their historical sales as additional points to calculate the forecast. This alternative data makes it possible to forecast sales of these new offerings as accurately as the model can forecast sales of products with extensive historical sales data.
Of course, defining what makes a store or product similar to new proposals isn’t easy. Similar stores could be located in the same geographic areas, have comparable sales volume, or even offer a related mix of products. Similar products could fall in the same category, have the same seasonality, or even the same colour. Many factors play into the calculation, ensuring that the resulting forecast is as accurate as possible. Moreover, the model learns over time, discovering patterns of similarity that may not be obvious to human logic.
Released from previous assortment constraints by using similarity values to make forecasts, we can determine whether or not a proposed item will do well in a new location. We can also recommend what items have ‘potential’ or what items are good to have in the store before clients suggest an addition themselves.
Simone De Rosa, Data Scientist on the Supply Chain team
Keeping shipments in line with display requirements
In addition to extending Evo Replenish’s forecasting abilities to new products, the latest release includes a couple of additional capabilities. Clients can now ensure that all shipments will conform to store and display requirements. For example, a store may only have the capacity to handle 1,000 SKUs at a time. A clothing store in a particular location may not stock swimsuits or accessories. Any systemwide or store-specific requirements can now be fully automated rather than addressed by shipment rules on a case-by-case basis.
Replenish can also follow display requirements. Every brand has a unique way to display items on shelves. For example, a fashion brand may always show sweaters in sets of three colours, or they may always place two tie options alongside every suit. Replenish can now ensure that every shipment of new and existing products includes enough companion pieces to comply with display requirements, as well as achieve the desired size distribution.
How to benefit from the new Replenish features
The new Replenish features are optional, but almost any business can benefit from them. These upgrades give clients more control over allocation, both in terms of the mix and distribution of products. New display requirement functionalities can ensure a more pleasing aesthetic and, in turn, a better customer experience in stores. Clients are empowered to hold higher standards for shipments without increasing supply chain complexity.
Most importantly, this latest version increases the predictive capabilities of the Replenish tool. Since the algorithm can now adjust the mix of products offered at each store over an entire season, businesses can react to disruptions with increased agility. No matter how store-level strategy changes, Evo Replenish can adapt and continue to provide accurate forecasts and automated replenishment. This flexibility is always helpful, but during a crisis, such adaptability becomes vital. Evo can now better serve emerging client needs with Replenish.
About the author
Kaitlin Goodrich is Evo’s main storyteller who helps communicate Evo’s message to the world.
Kaitlin received her BS in International Affairs and Modern Languages at Georgia Tech and then an LLM in International Trade Law from the University of Turin. She worked in Latin America doing education outreach for U.S. binational centers and has since worked as a content writer for international clients.
In her free time, she likes to travel or curl up with a good book.