Why is demand forecasting a
Featuring Chief Commercial Officer Peter Denby and Chief Customer Officer Thomas Hill.
Category:
Podcast
Series:
#2
Episode:
#3
Date:
February 2023
Stream on Spotify
Easy as 1, 2, 3?
At its core, retail is all about having the right products, in the right place, at the right time.
It sounds as simple as 1, 2, 3; but accurately predicting product demand is a huge challenge. It’s also widely regarded as imperative for retail success.
Demand forecasting isn’t a trivial matter.
Demand forecasting estimates and predicts customer demand over time, using machine learning based on historical data.
Businesses like Amazon are leading the way with AI-driven predictive forecasting. This capability means they can swiftly respond to unexpected changes in demand – like consumers panic buying toilet paper in 2020.
Ineffective demand forecasting is the single biggest impact on retailers’ profitability. But what does this look like? It could be too much product availability taking up valuable warehouse space. On the flipside, it could be not enough stock on the shelves, meaning you’re losing out on sales.
Research by IHL Group in 2015 found forecasting issues, like overstocks or out of stocks, cost retailers $1.1 trillion globally in lost revenue. Another survey reported 30% of consumers felt stockouts created a poor customer experience.
Out with the old…
So where can retailers start with demand forecasting? Traditionally, the focus was on four key areas:
Time series analysis.
Projections based on previous years’ sales. This approach is common for continual lines, like milk and other consumer packaged goods.
Customer-led forecasting.
Predictions based on customer segments and loyalty, such as historical sales, to project future demand.
Marketing activity.
Amplifying forecasts based on promotional plans – such as Easter, Christmas or seasonal sales.
Customer behaviour.
Forecasting based on customers’ interactions with products, including regular purchase cycles and gifting moments.
Forecasting has changed in recent years – as a result of supply chain challenges caused by macro factors like COVID-19, Brexit and the war in Ukraine.
Micro factors are also affecting forecasting. For example, influencers recommending products on social media can cause huge spikes in demand. It’s challenging for retailers to understand the signals, predict short term sales spikes and manage margins and product availability accordingly.
…and in with the new.
Successful demand forecasting relies on customer data. Understanding customer behaviour is critical for effective supply chain forecasts – and related retail decisions, such as pricing, assortment and marketing personalisation.
Historically, many retailers have had limited access to customer data. However, browsing activity for multi-channel retailers and tokenised bank card data in physical retail is the basis for calculating product affinity and substitutability, instantly enhancing supply chain forecasts. Putting customer data at the heart of decision making is how retailers will make better decisions.
Customer need states.
It’s tempting to start with products, but retailers should understand and forecast customer need states. Retailers typically stock sets of products meeting the same customer need – for example, women’s smart trousers in black, or four-star hotels with swimming pools in Turkey. Any changes to one product (e.g. price) impacts the others in the group.
Analysing customer needs also helps retailers incorporate demand transference in their forecasts. For instance, which products are substitutable or affined, which items do customers buy together, and which products compete for the same need.
Product availability.
Understanding product availability is critical for demand forecasting. Let’s take fashion as an example: products should be tagged at size level under a grouped product ID, so that retailers can identify problems with a product i.e. it’s fragmented, overstocked, understocked, or needs clearing through and promoting. Having products in sizes that customers don’t want to buy does not mean the product is ‘available’ and is understating the impact of lost demand.
Decision intelligence helps diagnose availability issues by marrying up the power of AI and the visualisation elements of BI, with recommended next best actions.
Data in the hands of decision makers.
Once you’ve collected the right data and tagged it accurately, the resulting insight should be served to decision makers in an easily consumable format. The intelligence guides decisions; by providing more detailed information to power smarter decisions.
Let’s look at fashion buyers as an example. They’re responsible for P&L and selling through stock, so need to answer questions like: am I above or behind forecast? Why? What needs to be done? Customer-centric insight helps them make intelligent decisions around pricing, stock levels and promotions.
Demand forecasting in action.
For great demand forecasting, retailers need to understand customer need states and product availability, then put customer insights in the hands of decision makers. But what does this look like in action?
Shopping habits.
Understanding customer need states is crucial for shops with different buying patterns e.g. a grocery store in a holiday location. Holiday grocery shopping missions tend to be very different to everyday food shop missions. In this case, decision makers need to understand the difference between infrequent, low loyalty shoppers and frequent, high value customers when forecasting demand.
This is a great example of where customer data/need states can be used in conjunction with time series analysis to fix the challenge of disruptive spikes in demand.
Demand transference.
Linking the promotional calendar at line-level to forecasts, and understanding the inter-relationship between products, is crucial for protecting availability and managing stock. Ensuring trading and marketing activity is linked to supply chain forecasts through centralised data and capability ensures more efficient supply chains.
True availability.
Availability is too often measured by products added to basket – and not by the items left behind. If a product is available in sizes 8-12, but 90% of demand wants it in sizes 14-16, then it’s not really available. Combining stock data and identifying failure states, such as low conversion on site, helps identify true availability and fragmented lines that need to be cleared rather than held. High traffic and low conversion is often the first sign of an availability challenge.
Media spend.
Retailers need to quickly understand why media isn’t performing well. If an advert is getting lots of traffic but low conversion, then there’s a problem – again, it could be down to availability of size (or a lack thereof). This example highlights the importance of understanding product availability to avoid wasted media spend.
Influencers and the importance of real time.
Let’s circle back to influencers recommending products on social media. This edge case can cause products to sell out exceptionally quickly, so real-time demand and sell-through curves need to be on an hourly basis, rather than daily or weekly. If products have sold out, retailers need to understand how demand can transfer to other products in the same customer need state. Decision makers also need access to data, so they can make margin decisions to delay the demand curve and adjust any promotions in line with demand signals.
Demand forecasting is a big opportunity for retailers. When it’s proactive, rather than reactive, it can help retailers step closer towards that perfect Elysium: 100% availability, zero waste and a frictionless customer experience. If you’d like to discuss how decision intelligence can power more intelligent demand forecasting for your retail business, please contact us.
Want to know how we can help?
Find out how far
HyperFinity can
take you.
Decision intelligence
is seeing HyperFinity
in action.
