Recent advances in machine learning and artificial intelligence have led to new approaches that are transforming industries, whether that be fraud detection in financial services, disease diagnosis in healthcare or driver assistance in the automotive industry.
Sectors that are seeing a huge impact from artificial intelligence are the Consumer Goods & Retail industries. Retail companies that were born in the digital era, such as Amazon and Alibaba, have leveraged machine learning across all their eCommerce platforms, whereas traditional retailers are playing catch-up. Consumer goods companies which often have less of their own first-party data, as they often don’t sell directly to consumers like retailers, rely on third-party data, including social data, search data and product review data. Artificial intelligence is helping consumer goods companies make sense of this unstructured data with natural language processing.
In this article we look at some of the top use-cases for artificial intelligence/machine learning in the Consumer Goods& Retail industries, and how to identify use-cases within an organisation.
Analysing the Value Chain
When identifying AI and machine learning use-cases that can derive genuine value within an organisation it is useful to understand the company’s value chain, and where in the value chain the use-case will have a positive impact.
Fig: Porter’s Value Chain Framework
Porter’s value chain is a framework for creating an analytical structure that follows activities from sourcing ideas or materials, through to production and finally, into the hands of consumers.
The value chain model involves five primary activities: inbound logistics, operations, outbound logistics, marketing and sales, and service. Support activities are illustrated in a vertical column over all of the primary activities. These are procurement, human resources, technology development, and firm infrastructure.
When using this value chain model, you must identify whether you are trying to increase revenue, reduce costs, improve operational efficiency or mitigate risks, and consider how changes from developing your use-case will benefit the entire organization.
AI/ML Use-Cases across the Consumer-Packaged Goods Value Chain
In 2022, consumers are leaving a bigger digital footprint than ever before, and CPG companies that are leveraging this data to better understand and cater to their consumers have a competitive advantage.
At the same time, supply chain disruptions are leading to price inflation, and CPG companies are using advances in AI and machine learning to better optimise their supply chain, forecast demand and set prices.
Below we will discuss some of the top AI and Machine Learning uses within the Consumer-Packaged Goods industry.
Product Development
Consumer Trend Identification
Overview – Consumer preferences are constantly changing, so identification of emerging consumer trends, whether that be new product categories, ingredients or claims can be identified using data sources such as eCommerce data, social media data, search data, sales data, and survey data.
AI/MLApproach – Time series data showing consumer trends (e.g., sales trends) could be forecast with models such as ARIMA or LSTMs.
Value – The identification of emerging consumer trends, can support new product development and changes to existing products that will resonate with consumers and lead to increased revenue.
Product Feedback Analytics
Overview – In order to better understand consumer opinions on a brand’s products, product review data from eCommerce stores can be analysed to measure overall sentiment and opinions on specific product attributes (e.g., packaging, flavour/taste, etc).
AI/MLApproach – Techniques such as Sentiment Analysis, Text Classification and Named Entity Recognition can be carried out on product review data using language models such as BERT and XLNet.
Value – Analysis of product reviews can support R&D in making changes to product to better cater to the needs of customers.
Marketing & Sales
Price Optimisation
Overview – By combining sales data, price data and additional data sources, such as marketing data can help brands determine the optimal price of specific SKUs.
AI/MLApproach – Multivariate forecasting of the price impact on demand with ARIMA or LSTMs.
Value – Price optimisation can help brands set the optimum prices to maximise profitability.
Consumer Segmentation
Overview – Consumer segments can be identified by clustering consumer data from sources such as surveys or CRMs which can support marketing strategy and targeting.
AI/ML Approach – Clustering techniques such k-Means Clustering, Density-Based Spatial Clustering of Applications with Noise (DBSCAN), and Agglomerative Hierarchical Clustering can be used on for example survey data to identify consumer segments.
Value – By segmenting consumers, brands can identify their target consumer and better understand how to tailor their marketing and products to the consumers’ needs.
Marketing Content Generation
Overview – Marketing content, such as newsletters and social media posts can be automatically generated using advanced Natural Language Generation (NLG) models.
AI/ML Approach – NLG models such as GPT-3, BART or LeakGAN can be used to generate text copy/content.
Value – By segmenting consumers, brands can identify their target consumer and better understand how to tailor their marketing and products to the consumers’ needs.
AI/ML Use-Cases across the Retail Value Chain
AI is transforming the retail landscape, to compete in this environment, retailers need to leverage their valuable customer data.
Below are some of the use-cases for AI/ML in the Retail industry.
Marketing & Sales
Product Recommendations
Overview – Recommending products to customers based on historical purchases and similar products.
AI/ML Approach – Recommender systems such as Collaborative Filtering, Content Based Filtering and Two-Tower Model.
Value – By recommending relevant products to customers the total spend per customer can be increased.
Customer Churn Prediction
Overview – Which customers are likely to churn can be predicted and customer retention initiatives can be recommended.
AI/ML Approach – Classification models such as XGBoost can be used to predict whether a customer is likely or not to churn.
Value – By identifying customers likely to churn, customer retention initiatives, such as personalised promotions or marketing, can be carried out to retain those customers, increasing total revenue.
Customer Service
Customer Issue Detection & Categorisation
Overview – Customer issues can be identified and categorised from customer feedback sources (e.g., call centres, social media, website forms) which can be used to prioritise customer support resources or product quality improvements.
AI/ML Approach – Customer feedback data from call centres, social media, etc. can be categorised using text classification via language models such as BERT and XLNet.
Value – Common customer issues can be identified by categorising feedback responses, so that corrective actions can be focussed on the areas that require them most, reducing risk of damage to brand reputation or loss of customers.
Supply Chain Optimisation
Demand Forecasting
Overview – Machine learning can be used to provide more accurate demand forecasting to improve product availability and reduce holding stock costs.
AI/ML Approach – Multivariate forecasting of demand with LSTMs or ARNet can provide more accurate forecasting of demand.
Value – Revenue can be increased with improved product availability, and costs can be reduced be reducing holding stock.
Operations
Workforce optimisation
Overview – Workloads can be predicted to optimise work shift planning for reduced personnel costs.
AI/ML Approach – Multivariate forecasting of workload with ARIMA or LSTMs.
Value – Personnel costs can be reduced by increasing efficiency through better work shift planning.
Store Location Optimisation
Overview – Optimal locations for new stores can be identified based on geospatial data.
AI/ML Approach – Geospatial analytics and spatial machine learning techniques such as Space-Time Pattern Mining.
Value – By identifying the optimal location for stores, revenue from those stores can be increased.
These are only some of the use-cases for Artificial Intelligence within the Consumer Goods & Retail industries and there is considerable overlap between the two. What is important when evaluating these use-cases within your organisation is that there is a clear value proposition of the use-case and it is expected to deliver a tangible return on investment. Too often companies attempt to "do AI" or "do machine learning" without a clear understanding of the benefits and total cost of development and maintenance of the entire solution, leading to budget overruns, delayed projects and solutions that aren't adopted by the business. Artificial intelligence has the potential to transform many aspects of these industries but initiating projects with a clear understanding of the costs and benefits will lead to more successful outcomes.
