What is machine learning and how can my organisation benefit?

There has been a huge amount of attention on artificial intelligence (AI) given the rise of large language models and the everyday use of chat-GPT and other tools. Whilst these tools have their place in enabling change in organisations, business leaders can also learn much from their data with machine learning, a subfield of AI, to understand new insights on how their organisation is working and do something about it. Starting out can be relatively low-cost and high return.

What is machine learning?

Machine learning is a subfield of AI that focuses on the development of algorithms and models that enable computers to learn and make predictions or decisions without being explicitly programmed. These algorithms learn from experience or historical data. The learning process involves identifying patterns, relationships, and statistical regularities in the data, which are then used to make predictions or take actions on new, unseen data.

How well are organisations extracting value from data for these models?

Despite rapid growth in the amount of data created, consumed, and stored over the past decade, little of it is being retained, let alone used in machine learning models.

A September 2022 report from Statista mentions that the projected increase in data volume will rise from 64.2 zettabytes in 2020 to over 180 zettabytes by 2025, which roughly equates to 21 billion continuous years of standard definition streaming video.

Despite this surge in data creation and replication, only a small fraction, about 2%, is actually stored and retained. We assume that only a fraction of that is being used in machine learning applications.

The volume of data being generated is simply too large for people to comprehend, increasing the need for–and the and the use cases of–machine learning.

On a micro-scale, we have seen many organisations overwrite certain datasets in their databases as new data is generated–missing some opportunities to learn and improve from historical data.

What are some sustainable wins for organisations?

Whilst there are myriad opportunities for companies’ manufacturing operations, the main focus at Re:Adapt Data Science is on extracting patterns and signals from customer behaviour–and how the organisation responds–to generate themes for improvement. This is an intuitive way to focus on innovations and more rapidly increase competitiveness.

Some of the following use cases require models to be “put into a production environment” with the help of MLOps professionals and Data Engineers, but a proof of concept can still provide business leaders the initial valuable business decisions.?We list a selection of 10 machine learning use cases to consider:

1. Customer Interaction Analysis: Deeper that looking only at customer sentiment, organisations can look for the high-frequency demand from customers to understand why customers are getting in contact, which in turn generates better business objectives.

2. Demand Cause Analysis: This takes the view that a portion of customer contacts (and cost to the organisation in servicing it) can be obviated with new approaches and innovations to customer process and procedures.

3. Ancillary Revenue Generation: Patterns from how customers use the product and service are analysed to inform which additional features or services are most likely to generate the most value for customers–and profit for organisations.

4. Product-Market-Fit Acceleration: Understanding frequent, but potentially subtle, use patterns of a current product or service to make quality enhancements and improve product-market-fit.

5. Churn Prediction and Prevention: Not only predicting customers who are likely to churn or discontinue their services so that intervention can be targeted, but also discovering patterns in how customers are treated to propose changes to how the business operates to prevent even the possibility of churn.

6. Personalised Recommendations: Providing tailored product or content recommendations based on customer preferences and behaviour.

7. Social Media Analysis: Monitoring and analysing social media platforms to understand customer trends, brand reputation, and customer engagement.

8. Price Optimisation: Analysing market dynamics and customer behaviour to determine optimal pricing strategies.

9. Contact Centre Optimisation: Analysing historical data, such as customer profiles, past interactions, and organisational performance, to determine the most appropriate action for customers from their point of view.

10. Customer Flow Analysis: Discovering signals in data that denote some type of friction for the customer, along with the probability that it will recur. This helps teams know when to deal with customers needs personally, and when to redesign customer journeys to prevent problems.

Our use cases are related to improvements that improve reputation, make for happier customers, and involve teams in solving problems alongside the data scientist leading the analysis. This ensures that learning is retained in the organisation, and improvements are built to last.

How does one get started with minimal cost outlay?

Getting started does not require expensive data engineering or architecture–in fact, many data initiatives have failed due to large capital expenditures but little up-front validation. A data scientist can augment a business team–even with effort of one day per week–to identify what is possible and of value before making large investments in data architecture. Data-driven insights can be generated using machine learning methods at relatively low cost to prove out a particular use case. This lends to a better decision on how to invest in software and hardware tools that enable storing, working with, and using data in a more automated manner. A simple but effective way to get started follows:

1. Iterate the definition of business objectives with data discovery: Gathering insights from data generates questions about how work is being done, why things are happening, and why people are doing what they are. This is a first step of translating data into knowledge and how more meaningful business objectives can be generated. A good team structure usually comprises a business leader, a data scientist, and a person who carries out the work. The trio can meld together a contextually powerful set of objectives to learn, which can be answered with the data analysed. This is an iterative process where better questions are asked as the previous ones are answered.
2. Ensure data can be shared with the data scientist: It will be necessary to coordinate with a person in the organisation to obtain the data from various sources and set the data scientist up for success, but they do not need to be a member of the core team above, saving time and effort. The benefit of this approach is that no time-consuming–and sometimes fragile–pipelines or dashboards need to be created at this stage.
3. Start to bring in other team members as the questions and data requirements widen: In many cases there are people in different parts of the organisation who can benefit from the initial insights discovered. These should be shared early on as this process often spurs collaboration on a more complex and valuable business problem to solve. In many cases, the new stakeholders that are brought into this learning process have data of their own that they might be able to contribute to further analysis, building a more integrated understanding of how work is being done, its effect on customers, and potential next steps that would not have been discovered in isolation.
4. Model potential changes and set up tests to confirm value: With a more holistic understanding across interconnected teams, predictive models and experiments can be proposed to generate robust understanding of changes made in one area of the organisation on others, with the customer’s point of view always in consideration. This systemic approach makes data-driven improvement more sustainable and very often larger in total return on investment.

With this approach, the value of data to the organisation can be experienced first-hand by its people (as opposed to presented in a slide deck which does not garner the understanding necessary for strong support). This also focuses minds on what should be done next with data science and machine learning methods to learn and improve the organisation. This in turn makes for better informed questions about how to deploy technology spend, where business leaders have more confidence in what data–and how it is stored, delivered and used–will be fit-for-purpose.

#data #machinelearning #customerrelations #businessintelligence