How accurate demand forecasting saves our client R2.6 million a year
12 November 2018
Smart business basics
The success of most businesses depends on accurately forecasting expected daily, monthly and yearly sales: this is basic business 101. This knowledge is underpinned by being able to accurately forecast the expected footfall of any particular store or outlet so that a business can accurately schedule the optimal number of staff to deliver the best service, as well as ensuring that the right amount of stock is on hand to meet that demand. This is particularly important for day-to-day businesses with products or services with a limited shelf-life, or those that depend on seasonal demand, such as hotel rooms or food outlets.
The problem with people
Humans’ inherent biases are well-documented, a fact especially true for forecasting. People mistake random noise for patterns; they overweight recent occurrences relative to longer-term trends; and have difficulty accessing and assessing the large amounts of data which might help inform forecasts. It is easy to see why ‘machines’ – computers with smartly programmed software – outperform people by being able to more accurately forecast and plan, saving money and leaving the people to focus on what they’re good at: managing sales, services teams and co-ordinating staff and orders.
Man vs. machine
At Predictive Insights we decided to pitch man vs. machine and put this concept to the test for one of our clients - a quick-service restaurant chain with over 100 branches in a number of African countries. Using a demand forecasting tool tailored to this specific client - a core product from our suite of machine learning and econometric models, we used the client’s store-level sales data, as well as external data pertinent to their industry, such as weather, local economic conditions or days of the week, to more accurately predict sales per day at an individual store level. And we compared the models accuracy, to the predictions made by experienced branch managers.
Experienced branch managers vs Predictive Insights machine learning model
(% of store days predicted within a specific range)
An early version of our set of demand forecasting models predicted sales within 20% of actual sales on more than three-quarters of individual store-days, whereas less than half of branch managers were able to reach this benchmark. The employees in the competition couldn’t even beat the machine when the target was within 50% of actual sales – only 87% of branch managers reached this level of accuracy whilst the machine bettered it every time.
The machine’s accuracy, with the right tuning from Predictive Insight’s experts, also improved with time as it starts to learn more about the patterns in the data – this model now routinely predicts over 80% of stores within 20% of actual sales. Feedback from branch managers showed that they also liked the tool - instead of expending their time and cognitive energy on forecasting fortnightly sales, they were able to focus on ensuring each store ran smoothly, motivating their team and dealing directly with customers.
But does it save the company any money? And how much?
Yes! Looking at the numbers for one of our fashion retail clients, prior to using our model one of their senior analysts spent a large portion of their time collating forecasting sales using spreadsheets and a model which took a weighted average of recent and longer-term sales trends. Our suite of products was able to automate and outperform this approach, allowing for staff levels to be optimised, saving the company between 2-5% of staff costs (from over-staffing), as well as improving the customer experience and therefore sales by having the right number of staff during periods of high demand. Overall, the company is likely to save over R2.6 million a year if they continue to use this model. They will also be able to shift that senior analyst onto other tasks where the returns to time are higher. For companies who rely on branch managers to make forecasts, savings are even higher than this.
Want to find out how more accurate demand forecasting can save your business money and improve customer experience? Please drop us an email (email@example.com).
First Floor, Rhino House
23 Quantum Street, Technopark