sorry can't get this for you

5 minutes read

March 16, 2023

How Machine Learning is Transforming Customer Insights

Learn best practices for machine learning customer insights within consumer goods. Uncover untapped markets and increase loyalty to drive revenue growth.

Increasingly, firms are leveraging customer insights from machine learning (ML). Nowhere is this more evident than in the consumer goods sector.  This article will provide an overview of how companies can use machine learning to improve:

+ Brand positioning+ Creativity+ Segmentation, targeting and positioning

We also discuss best practices for implementing ML-driven customer insights. This includes the infrastructure needed to drive revenue growth and outpace the competition.

Table of Contents:

  1. From Data to Action: How Machine Learning Drives Customer Insights
  2. Leveraging Machine Learning for Better Brand Positioning
  3. Using Data To Inform Creativity: Segmenting Customers And Refining Targeting With Machine Learning
  4. How Machine Learning Can Drive Revenue Growth
  5. Best Practices for Implementing Machine Learning for Consumer Goods Insight
  6. Challenges Faced When Applying Machine Learning Techniques to Customer Insights
  7. Conclusion

1. From Data to Action: How Machine Learning Drives Customer Insights

By leveraging cutting-edge analytics, firms can better understand customers' habits and tastes. Businesses can collect, interpret and analyse many data sources through machine learning algorithms. These include both structured and unstructured data sourced from a number of sources including social media posts, online reviews, historical sales data, Trends reports (e.g. Google Trends) and multiple customer touchpoints in physical stores and online.

Machine learning can generate insights into customer behaviour that would otherwise remain unnoticed. This can have a transformative effect on marketing campaigns, product development, pricing strategy and supply chain optimisation


Deep Learning (DL) is a type of ML that involves training neural networks with large amounts of data. Executives can then use these data to make predictions or decisions. This could include:

+ Predict equipment failures and maintenance needs. 

+ Forecast consumer demand based on historical sales data and market trends.

+ Optimise production and inventory management.

+ Personalise recommendations based on customer preferences, purchase history, and social listening.

Natural Language Processing (NLP) can be used for user-friendly chatbots providing real-time support. This improves user experience scores and overall satisfaction levels among consumers. NLP can also analyse customer feedback to identify preferences. This can feed into recommendation systems. For example, a manufacturer can recommend a new product line based on social media trends.

Natural language understanding (NLU) is a subset of NLP. It can deconstruct speech and text to understand intent and interests. 

An example use case is text analysis of customer service transcription or social media listening. Firms can track what customers say online and find product/marketing improvement areas. DL, NLP and NLU technology enable businesses to classify customers by their interests. This type of segmentation facilitates personalised marketing and refines targeting and positioning.

2. Leveraging Machine Learning for Better Brand Positioning

Unstructured datasets like blogs or social media can uncover insights about brand perception.

Sentiment analysis is among the most popular ML methods to monitor brand perception. This technique uses text mining and NLU to detect emotions from online conversations about products. Analysing customer sentiment can identify how buyers feel about their products or services. For example, if there is an uptick in online discussion praising “organic ingredients”, companies can adapt their product mix to meet demand.

ML like deep learning and neural networks can identify emerging trends before competitors. This gives brands an edge when planning trend-driven product launches or campaigns. Furthermore, recommender systems allow businesses to gain valuable insights into consumer behaviour patterns. This can help inform decisions about product placement or promotional offers.

3. Using Data To Inform Creativity: Segmenting Customers And Refining Targeting With Machine Learning

ML can help segment customers, refine targeting, and craft product positioning. Algorithms like clustering, decision trees, and social listening can inform segmentation.  Understanding their actions, opinions, motivations, likes/dislikes, and profiles is invaluable. Businesses can then construct an all-encompassing view of their target market.

This knowledge enables companies to refine product positioning. In competitive sectors, positioning is essential for improving marketing campaigns and conversions. Producers can use a clear profile of buyers to craft personalised messages that resonate.

4. How Machine Learning Can Drive Revenue Growth

With predictive analytics, businesses can optimise pricing strategies and maximise profits. Companies can learn how customers respond to prices or promotions and adjust accordingly. Regression analysis can also be used to identify relationships between different variables. For example age, geography, income, and purchase frequency. This information can be used to develop predictive models to forecast sales demand.

Advertising performance is another area where ML has the potential to increase revenues. Through NLP, companies can analyse customer sentiment towards specific campaigns or products. Finally, predictive analytics helps organisations better understand customer retention and areas for improvement. Additionally, lead scoring capabilities enable them to target prospects more likely to convert into buyers.

5. Best Practices for Implementing Machine Learning for Consumer Goods Insight

a) Improving Customer Value

+ Mackmyra, a Swedish whiskey distillery, produced a limited edition whiskey called "Intelligens". This was the first to be generated by AI and curated by humans. Mackmyra worked with Microsoft and Fourkind to develop a system that combined existing blending recipes, sales data, and customer preferences. The algorithm learned to identify the most optimal flavour combination through multiple iterations. 

+ Diageo has introduced a digital application called "What's Your Whisky?" which uses data and algorithms to recommend personalised whisky recommendations to consumers based on their taste preferences. The program has seen significant success, with over 1.6 million people participating since its launch in 2019. Another use case for ML is to analyse digital data to identify trends and patterns related to positive drinking behaviour. Diageo’s Road Safetycampaigns use insights to identify the most effective messages and channels for promoting responsible driving and reducing alcohol-related road accidents.

+ Carlsberg operated the Beer Fingerprinting Project with Microsoft. The project used ML and sensor technology to identify ‘flavour footprints’ to develop new brewing varieties. Using artificial intelligence (AI) enabled Carlsberg to speed up, improve quality and reduce the cost of new product development.

Working with researchers at Aarhus University, the Danish beer-maker has developed sensors that are able to detect the differences between beer flavours. By utilising this data, Carlsberg has used ML to develop new beers and improve quality control.

b) Improving Carbon Footprint

Consumers are increasingly concerned about climate change. Responsible producers can use ML to identify ways to reduce their carbon footprint. By combining predictive analytics with sensor technology, companies can minimise waste and reduce maintenance and repair times, which decreases plant energy use. 

Not only can this reduce emissions and waste, but it can also improve sales through improved brand perception in the long term.  ML is being used in a multitude of ways to improve sustainability. For example, 

+ Pernod Ricard uses ML to optimise transport and logistics operations.

+ Anheuser-Busch InBev uses AI to optimise its brewing process and reduce energy/water consumption.

+ Diageo uses bots, AI and digital twins to transform its supply chain and improve sustainability initiatives like Grain to Glass.

5. Challenges Faced When Applying Machine Learning Techniques to Customer Insights

The challenges faced can be broadly segmented into

+ ML Science: Data-science and model building.

+ ML Engineering: Infrastructure management or MLOps.

ML Science

A challenge b2c companies face using machine learning is data availability and quality. This can be difficult in global consumer-facing sectors, especially when dealing with multiple languages. Data is often high-volume, fast-moving, noisy, and prone to errors. This makes it challenging to train a model accurately and rely on it. 

Additionally, unstructured datasets like social media monitoring are usually unbalanced and highly volatile. This requires models to be regularly updated with new information. Furthermore, computational complexity can also be an issue. Complex models need large amounts of computing power and can become expensive quickly.

In industries focused on large consumer markets, models may need to adjust frequently to changing market conditions. Organisations may need advanced MLOps architectures to handle this increased model complexity.

Regulatory compliance is another challenge. Businesses must protect their customers’ personal information by storing all data according to industry regulations like GDPR or CCPA (California Consumer Privacy Act).  This may need extra MLOps resources to ensure models are fair, transparent, and explainable.

Another ML challenge faced is model selection bias. This occurs when selecting the wrong type of model for a given task. Choosing the wrong algorithm type could produce inaccurate or unreliable results. Overfitting can also occur. A model may perform well on training data but fails to generalise on unseen data. Data scientists must test models thoroughly before deployment into production environments.

Finally, there are issues related to interpretability. Models are interpretable when humans can understand why a model made certain decisions. This requires careful analysis of how each component works within the system architecture.

Pipeline AI - a platform to empower data scientists

“Our platform abstracts the complexity of managing thousands of models in production, whether on-premises or in the cloud, and provides data-scientist with out-of-the-box strategies to optimise their infrastructure management. 

Mystic’s hardware and software experts have delivered a novel serverless GPU API that currently helps thousands of data scientists and ML engineers get instant access to a GPU for their ML pipelines. We are now giving access to the platform that powers this public-facing API to help consumer-facing organisations supercharge their own ML systems on their own infrastructure.”

Oscar Rovira - Chief Product Officer, Mystic AI

ML Operations

Once the model has been trained and is ready to be in production, the data scientist hands over to the engineering team all the code required (ML pipeline) to run their model. The engineers then take care of packaging and optimising the pipeline to run at scale inside their infrastructure.

Existing solutions like Kubernetes help manage some of these problems. But engineering expertise is required to build, customise and maintain these Kubernetes-based systems. Furthermore, Kubernetes' strengths lie in CPU-based workflows, whereas ML workflows run better and faster with GPU-based hardware. Managing GPU-based hardware or hybrid hardware (CPU and GPU) can greatly improve consumer AI models' performance. Still, more technical expertise is required to optimise and manage these systems. The considerations for good ML infrastructure management are therefore:

+ Fast and scalable architecture design and;

+ Simple but powerful model management.

A robust architecture should be designed to scale with increased demand and ensure data security and privacy. This requires a team with specialised ML engineering, data science, and cloud computing skills. The internal and external communication to this system is handled via distributed APIs. To achieve speed and scalability of these APIs, their total overhead and throughput need to be optimised.

Pipeline AI abstracts the complexity of managing thousands of models in production, whether on-premises or in the cloud, and provides data-scientist with out-of-the-box strategies to optimise their infrastructure management.

API overhead refers to the additional computational resources and time required to manage the the ML pipeline in production. And API throughput refers to the rate at which the system in place can process concurrent utilisation of the ML pipeline, which is limited by the available hardware resources and the system’s design.The machine learning tech stack typically includes programming languages, machine learning libraries and frameworks, data storage and processing tools, cloud computing platforms, DevOps tools, and visualization tools. The specific tools and technologies used will depend on the specific machine learning application and the preferences of the developers and data scientists involved. The key is that the tools are compatible, and that the choice of partner does not tie the enterprise into expensive proprietary deals.

8. Conclusion

Machine learning customer insights can revolutionise how businesses approach marketing and sales. By leveraging data-driven insights, firms can better understand their customer base and adjust strategies to attract and retain them.  ML can:

+ Monitor brand perception and track customer sentiment.

+ Make predictive analytics more accurate.

+ Analyse historical data and inform product development, pricing and marketing strategy.

+ Manage risk and supply chain optimisation better.

To leverage machine learning, companies will need to solve both the science and the engineering challenge that comes with it.  The science challenge is characterised by acquiring the right data and building the model that solves the business requirement. The engineering challenge is characterised by providing access to the infrastructure required to train, deploy and monitor the model at scale. 

While AI may have challenges, the potential benefits outweigh any drawbacks. Consumer-facing businesses can leverage machine learning to gain advanced customer insights and increase revenue and market share.

About Mystic AI

Mystic AI is a venture-backed enterprise delivering solutions for managing machine learning (ML) in the enterprise. We provide tailored solutions and expert guidance to companies looking to build and scale a robust machine learning infrastructure.

Having developed and launched our flagship MLOps automation toolkit and cloud solution, Pipeline AI, we have first-hand experience with the ML challenge for enterprise. We are experts in the end-to-end ML hardware and software stack, and in supporting complex, high-speed, distributed and real-time deployments.

Pipeline AI - a platform to empower data scientists

Pipeline AI makes it easy to work with ML models and to deploy AI at scale. The self-serve platform provides a fast pay-as-you-go API to run pretrained or proprietory models in production. If you are looking to deploy a large product and would like to sign up as an Enterprise customer please get in touch.

Follow us on Twitter and Linkedin.