Knowing what your customers bought last month does not cut it on the UK high street anymore. Traditional business intelligence tells you what already happened. Now, some independent retailers are using prescriptive analytics to figure out what to do next.
The spectrum of intelligence: From hindsight to prescriptive action
Most businesses understand descriptive analytics, which shows what happened, and diagnostic analytics, which explains why. Many also use predictive analytics to forecast demand. Prescriptive analytics goes a step further. If a predictive model spots an upcoming spike in demand, a prescriptive system suggests exactly how to prepare for it to meet your business goals. It changes the question from "What will happen?" to "What should we do?".
Hyper-local demand: The SME advantage
Independent UK retailers can compete with larger chains by focusing on their specific locations. A boutique in central Manchester sees different customers than a shop in the Cotswolds. Localised demand forecasting predicts what people will buy at a specific store by looking at hyper-local data. These models track regional weather forecasts to adjust stock for umbrellas or ice cream. They also factor in local events like football matches or festivals that drive foot traffic, alongside neighbourhood demographics.
Revolutionising assortment planning
Prescriptive analytics changes how retailers plan their inventory. Instead of guessing the right mix of styles and price points, businesses use mathematical optimisation to find the exact stock levels needed for each location. The system recommends replenishment schedules to prevent stockouts and reduce markdowns. For instance, the software might suggest a 15 percent discount on a slow-moving item at three specific stores right before the season changes, helping clear space while protecting the profit margin.
The algorithmic engine: Making machine learning accessible
You do not need an in-house data science team to use these tools. Platforms now package machine learning algorithms like Random Forest and XGBoost into interfaces that let managers run "what-if" scenarios. These models calculate substitution behaviours, estimating what a shopper will buy if their first choice is out of stock. They also use advanced choice models to predict demand for new items based on attributes like brand or size, even without historical sales data. Some platforms let store managers type questions in plain English, asking the system which products to mark down to hit a specific target.
Building a Culture of Quality
Prescriptive analytics only works if the underlying data is accurate. Retailers need to combine information from their point-of-sale systems, loyalty programmes, and external sources into a single database. If the input data is flawed, the system's recommendations will be wrong. Getting this infrastructure right lets managers base their inventory decisions on algorithmic forecasts instead of just looking at last month's sales reports.
Sources include:
Assortment Optimization Over Time, Cornell University
Constrained Assortment Optimisation, Institute for Operations Research and Management Sciences
Algorithm and Demand Estimation Procedure for Retail Assortment Optimisation, University of Pennsylvannia