In March 2020, the demand for most consumer goods had increased almost overnight. Retailers across the globe registered a sharp swell in the sales of daily necessities and personal hygiene products as consumers took to panic buying amidst the fear of a global lockdown due to the exponential rise in Covid-19 cases. At the same time, the pandemic forced industries with low or almost no demand like travel to drop their prices in hopes to reach their projected revenue for the year. Consumer demand and behavior became difficult to predict due to unforeseen and sudden changes. In such scenarios, leveraging data can help retailers to make smart decisions about product offering, inventory, staffing, and marketing expenditure. 

The supply chain of retailers is usually a complex process that thrives on the stability of knowing the quantity, time or location of the products in demand. As a result, it has become imperative for retailers to use quantitative and qualitative data to forecast demand at a granular level. Not embracing and adopting emerging technologies and analytics-driven innovation to gain insight into customer behavior or other factors that help forecast demand puts retailers at the risk of making expensive mistakes.

Traditional methods of forecasting like manually tracking data using spreadsheets are time-consuming, ineffective, and expensive. Outdated demand forecasting techniques don't allow retailers to fully leverage data including historical sales, market research, and stock replenishment data. This results in retailers not reaching their sales targets because of misalignment of supply and demand across the business operations. So, how can retailers around the world rethink and adopt a modernized innovation-led strategy for retail forecasting and demand planning? Retailers can start by asking these questions on their path to be better prepared for the rapidly evolving future. 

• Can advanced statistical forecasting methods help you do more?
• Can emerging technologies help automate and scale demand prediction across your supply chain?
• Can new-age demand forecasting solutions successfully leverage data from disparate sources and help you make profitable decisions?

The answer to these questions lies in the adoption of AI and ML-powered demand forecasting that are infused with tools such as Vertex AI Forecast into retailers’ demand forecasting systems. A survey conducted by the Institute of Business Forecasting (IBF) found that over 70% of the respondents stated AI will be the dominant technological element in demand planning. AI and ML-led innovation have the power to empower global retailers with new ways to innovate and win customers' trust and loyalty. Unlike traditional forecasting techniques where data analysts mostly rely on using historical observations to estimate demand, AI can help manufacturers to generate accurate and highly scalable demand predictions that lead to better inventory decisions. 

Accurate use of data to predict the future fits right into the alley of Vertex AI Forecast—Google Cloud's unified artificial intelligence platform that enables retail and CPG companies to improve forecast accuracy. While technology is at the disposal of many retailers today, the use of artificial intelligence (AI) and machine learning (ML) hasn't been democratized to small-sized retailers. Quantiphi, leveraging Google Cloud’s technology, is one of the first organizations to address this gap. Google Cloud's Vertex AI Forecast is enabling retailers to improve demand forecasting accuracy at the granular level through deep learning, ML algorithms, and a host of GCP services.

How does Vertex AI Forecast work? 

Vertex AI Forecast allows retailers to quickly build forecasting models using advanced AutoML algorithms. In simpler words, it enables retailers to automate demand prediction using a variety of factors specific to their business. Factors for analyzing demand for different industries vary but all demand forecasting models have a common ground. In the case of the retail and CPG industry, Vertex AI Forecast analyzes a copious amount of critical factors such as seasonality, competition, product types, location, historical sales data, and consumer behavior. This helps retailers to ensure that the products are in the right place at the right time and at the right price. 

Let's understand the capabilities of demand forecasting better with a few hypothetical examples:

Accurate Demand Planning to Overcome Competition: A home appliances store sees a pattern in demand for their winter appliances growing every year. Every winter, there’s a peak in demand for products like heaters, rugs, and flasks. To their dismay, a recently opened home appliances store diluted some of the seasonal product sales. However, with the rise in demand for home office furniture due to the pandemic, the demand for desks and office chairs spiked. The older store adapted quickly with increased inventory and used competitor data to predict demand for their seasonal products. With the adapted strategies based on market demand and stocking the right products at the right time, the store’s growth got back on track. 

Overcoming Seasonal Fluctuations: A local sports store has had consistent and predictable sales for over half a decade. However, recently its sales dipped and the owners couldn’t identify the reason. After a few meetings with the in-store sales teams, they found that several sports training facilities around the area had started hosting soccer, volleyball, and tennis events, and their store’s inventory didn’t offer a wide range of products. The store then decided to refresh its product range for all sports they focused on. This unique approach helped the owners to overcome the uncertainty in inventory stocking and over-achieve sales targets.

Let us also look at a real-world example where Quantiphi has leveraged Vertex AI Forecast to deliver improved forecasts for a beauty product conglomerate:

Problem Overview

The customer is an American multinational manufacturer and marketer of skincare, makeup, fragrance, and hair care products distributed internationally through a diverse portfolio of self-owned brands. The customer wanted to increase the accuracy of their product sales predictions over short-time horizons (1 hour, 1 day, 1 week up to 4 weeks) for SKU-level demand forecasts

Solution

Quantiphi built an ML pipeline to predict demand sales and solve more complex use cases for media efficiency and supply chain optimization. These predictions led to immediate benefits by providing dynamic sales targets to e-commerce teams in turn helping them adjust sales drivers and operations to meet sales targets.

Our engineers used Vertex AI Forecast and its versatile platform to experiment and build scalable ML models while streamlining ML modeling, speeding up model optimizations, and automating model selection. This helped Quantiphi utilize SKU-level demand units and product catalog data while incorporating time, seasonality, and promotional trends for more accurate forecasting.

Business Impact

• Reduction in MAPE across all SKUs compared to planning error metrics
• Increased accuracy of predictions by over 2 times for daily forecasts

Below is another customer story where Quantiphi has delivered real business value through demand forecasting for a private label manufacturer:

Problem Overview

The customer is a multinational food processing company specializing in the manufacturing and distribution of private label packaged foods and beverages across North America. The customer wanted to build an enhanced demand forecasting model for a set of 67 SKU in order to capture trend change and seasonality to enable better forecasts, derive marketing insights, and monitor important KPIs. 

Solution

Quantiphi developed a forecasting model with existing historical data and further enhanced using external as well as synthetic data features. This allowed for faster optimization of multivariate forecast models, and automated model selection to help significantly reduce the turnaround time to predict forecasts with the least possible MAPE value. 

Business Impact

• Reduction in MAPE across all SKUs by 33%
• $1.6 Mn annual savings through automated model selection at the click of a button 
• Increased forecast granularity from monthly to weekly-SKU level with seamless scalability for any additional data columns

AI-powered demand forecasting is the next step in shaping the future for retailers around the world. AI and ML-enabled demand forecasting solutions empower retailers to reduce risks, make sound financial decisions that impact profit margins, and achieve increased cash flow. To learn more about how Quantiphi can help retailers like you leverage the full potential of AI-powered demand forecasting with Google Cloud's Vertex AI Forecast for efficient inventory allocation, increased supply chain efficiency, maintaining stock levels, and optimizing production planning, visit www.quantiphi.com or reach out to our experts.