What good is a translator with a limited vocabulary? Or a recipe missing ingredients? Both result in a lack of richness or perhaps even a failure to execute altogether.
The same is true when you try to pour data into traditional merchandising tools with the hope of gaining insight. You might get the general “message”, but something is missing. In other words, your data is only as good as the tools you use to translate it into decisions.
Legacy tools and approaches to merchandising can guide simple decisions based on available data. But relying on them for crucial decisions such as buying and assortment, generally leads to retailers’ hands being tied behind their backs in the face of demand changes. This is where an “agile retail” approach, enabled by advanced technology like AI, comes to the rescue.
Here are 5 ways in which digitally-led models deliver intelligent, flexible, and sustainable merchandising executions that go far beyond simple reporting, when oversimplified tools and methodologies just aren’t enough.
1. Capture the reality of demand: Decision-making from the bottom up.
Retailers have traditionally taken a “top-down” approach to merchandising. This means that, based on available data (i.e. past sales performance), they use traditional tools like WSSI and open-to-buys to estimate how much inventory they need each season.
In a natural response to uncertainty, retailers tend to overcompensate by producing, buying, and allocating more than is really necessary. Later, if demand does not unfold exactly as planned, the result is normally overstocking, heavy discounting, and waste.
But in agile retail business models, decision-making works the other way around. Real-time data on the ground, rather than estimations based on past performance, informs merchandising decisions. With a truer vision of the market, retailers better understand trends, adapt to them faster and introduce new products more frequently. And since this process can be automated through AI, retailers become faster and better at delivering on new needs, effectively reducing consumer to manufacturer (C2M) lead times.
- “No” data, no problem: Managing missing and distorted data
All of the past performance data in the world wasn’t enough to predict how COVID-19 would derail demand in 2020. And related store closures have resulted in periods of time for which there is simply no sales data for items in physical stores. On the other hand, the data may also reflect contextual and temporary changes in demand that can skew future forecasts. Thus, retailers must be able to make decisions based on limited data or to distinguish between regular seasonality patterns and external events.
In these cases, when the past doesn’t reflect a likely future, more flexible forecasting models enabled by AI can treat and contextualize data to protect future calculations. For example, some advanced solutions can omit stores or periods of time from calculations in case of events like closures to avoid periods of no sales appearing in forecasts as periods of low conversion rates. On the other hand, some forecasts will “tag” periods of abnormally high demand in past data to neutralize the effect it may have on future calculations.
3. Take a customer-centric approach: Sharper, hyper-local assortments
Trends and demands change all the time. Relying on broad store and size clusters to determine product allocation often results in overstocking some stores and understocking others. In this approach, real-time customer demands rarely, if ever, enter the equation.
AI lets retailers become more customer-centric by sharpening assortments at the hyper-local level. The ability to collect and process data at the most granular level means they know exactly what products and sizes are in most demand at each point in their networks. Machine learning algorithms also make it possible to group and regroup stores and products dynamically based on new and changing demand patterns.
In other words, retailers can finally decouple the buying and allocation processes. Instead of buying products and allocating to the best of your ability - which is again, an inflexible top-down approach - you can make allocation decisions based on what is selling now.
4. Know what to expect: Predict the impact of your executions
The advantages of agile retail and AI on increasing merchandising efficiency are clear, but an equally important factor is being able to quantify the impact that these executions will have should you take action.
Take predictive analytics in a rebalancing scenario, for example. The calculations they enable make it possible to not only estimate what is likely to happen (e.g. a size S red shirt is 3 times more likely to sell in store A than store B), they also recommend what course to take (e.g. send to store A from store B) and the potential impact each decision will have such as the estimated sales increase you can expect.
This insight provides both an objective business case and increased retailer confidence in taking one course of action over another, something that traditional approaches to merchandising execution cannot.
- Embed sustainability at the core: Use inventory more efficiently
One of the biggest enemies of sustainability in the fashion industry is a lack of efficiency, generally a result of oversimplified, traditional approaches to merchandising based solely on historical data. To manage uncertainty, retailers have customarily overcompensated by overproducing and overallocating, ultimately leading to wastefulness.
But today, AI and data-driven automation allow retailers to shake off the inefficiencies of the past. Sustainability is not limited to sexier-sounding concepts like eco-friendly materials, virtual fashion shows and the like. Sustainability can literally be embedded at the core of merchandising, considered from the very beginning of planning.
Since AI allows retailers to take the data-driven, bottom-up, granular approach they need to better predict hyper-local demand, they can gain the confidence to reduce safety stock, rethink product offerings, and reduce assortments to provide more value over volume. In fact, by applying probabilistic algorithms, retailers can even reduce order volumes by 20% while lowering stock-outs by 60%.