Analytics has been around for decades. Many businesses are already using data and analytics to get to a certain outcome and have been in various formats for some time. That said, we’re still yet to fully understand the methodology and implications of using machine learning and artificial intelligence (AI) at scale.
When you combine analytics and machine learning you get to the complex capabilities today’s business require in order to solve scalable problems. It has the power to underpin every KPI, enrich any task or solve any problem.
There are quite a lot of ways that can happen. Every few years there is a new buzzword or a new trend, usually to do with technology. It was the internet, then mobile. Today, many people are talking about artificial intelligence. You only have to look at the government’s Industrial Strategy – which advocates the use of AI to support everything from schooling to decisions – to see how invested we are in it as a country. Wider still and its playing a crucial part in the United Nations’ (UN) 17 sustainable goals.
Some people are likening it to the coming of electricity.
Looking back, electricity fundamentally changed our society. If you go back to the beginning of the century, the industrial revolution had already started changing society, with people moving from the villages and rural areas to the cities. Without electricity, life was very different. Sleep patterns, mealtimes and behaviours were all different.
Before electricity, at the beginning of the industrial revolution, all power was centralised around the factories. Outside the factories, there was no power because it couldn’t be transported. Electricity brought the ability to decentralise the power. You could get that power to different factories, different towns, even the home eventually. It changed society as we know it.
Unlike many other technologies, machine learning also has that capability. It can both automate and optimise most business processes using localised data – something deemed impossible until now. But I think we need to question the amount of hype surrounding it and we need to understand it for what it is. It is an innovation (if used right) that can transform many things and many business tasks.
Even today, most algorithms tend to be quite generic, meaning you’ll take a lot of data through linear models, take some sort of average and the machine will work. On the current models, it will make some average decisions. With machine learning, you don’t need to do that. You can specifically look at the set of data and make a decision that is more accurate. There are very few processes in your enterprise that you cannot change with machine learning.
As I said, back in the day, electricity managed to decentralise power consumption, it gave electricity to all. Similarly, machine learning is decentralising data. There are very few processes around all your enterprises that you cannot change with machine learning. Computing is getting cheaper and cheaper, in terms of storage and computing power. Then there’s the data: 90% of the world’s data has been generated in the last five years and it keeps growing exponentially. Also, the people skills are there today, when before they were not. There’s the perfect storm.
Why is machine learning so different?
Machine learning can look at localised data in a specific point in time and point of data. Why is that different from what we were doing before? In common with artificial intelligence products, a product using machine learning creates what we call AI virtuous circles. This is basically a locking loop, or a self-feeding mechanism.
Many of the worlds biggest technology companies have invested huge amounts of money on this technology. If you think of a traditional IT system, you install it and then live with it. You might make enhancements – such as automaton or augmenting it as your needs arise. But, the version you install, at large, remains there, unchanged, for years. Machine learning isn’t like that. You deploy the model, and the model will adapt and learn. The more data you get, the better the product is. More users produce more data, and if used intelligently data analytics can start producing signals of information. Perhaps signals that you weren’t looking for originally. I’ll touch on that more later.
Effectively, machine learning allows you to start doing platform economics. Companies like Google, and Amazon all started with a very specific, narrow problem. Google started just doing search, and now they do everything. Why? Because, every time they get the data, they use it to create more data. And that’s how platform economics works.
This is where we need to be cautious. Data shouldn’t be exploited, it should be used with the customer, their needs and wants, at the heart of it. Every time you do anything with data there needs to be a clear customer centric purpose. (At Experian anything we do, particularly using AI, is done so aligned to our mantra: FACT. Read more about what FACT is here).
Alchemy versus science
In terms of the challenges we have, think about alchemy versus science. Early in the 12th century, alchemists managed to combine elements to produce metals using proto-science; but they didn’t understand what they were doing. They predicted a lot of things that didn’t happen. They wanted to transmute any material into gold. A challenge with machine learning and data science is that science isn’t the difficult part – implementation is. You have to do it right; you must use the right modelling and it must be scientific, you need practitioners who know what they’re doing. The benefits of doing this translate into a number of areas:
Improving customer experience
How can you use machine learning? The first area to put forward is improving customer experience. With machine learning and advanced analytics, you can make fine tuned decisions on the spot. You can respond quickly. More importantly, you can switch the question around. Rather than pushing each product individually, you can say, ‘let’s identify the customer, understand the customer and then make the decision.’ We’re moving from a brand or product approach to a customer centric approach.
When machine learning allows you to do that, you can start having proper, relevant conversations. More informed conversations.
If you’ve ever applied for a financial product – a credit card or a mortgage for example – you know that the current process can be painful and takes time; often a lot of time. In some cases, the lender will ask you to bring a paper statement to their branch as validation of income and expenses.
At Experian, we’ve digitalised the whole journey. More importantly, we’ve created an engine that takes any bank statement and automatically categorises transactions. It can determine what is income, what’s an expense for rent, what’s an expense for gym membership, what’s spent on going out. When you do that at scale, you can calculate an individual’s affordability much more precisely.
In terms of the benefits, you go from a process that might take days or weeks to one that takes seconds. It’s also very accurate. That makes it a win-win for both the lender and the customer; the lender because they get a better overview of each customer’s affordability and risks, and the customer because they’re more likely to get a more accurate decision, sooner.
The second purpose is to for fraud prevention. With machine learning, you can start analysing the transactions through the ecommerce site. Fraud is quite a specific use case, because it is very unbalanced. You will get peaks in the models. But unlike a traditional, rules based system, the machine – eventually – will pick out the fraud. It will adapt, and become more and more sophisticated. It will also change, through self-learning, to stop the fraud.
Let’s return to the concept of the virtuous circle. Once you start getting all these transactions and doing analytics at scale, you’ll see things you weren’t looking for. You could, for example, do clustering on the segment, so you can understand the type of customer; how they shop; how they behave; who prefers to shop in the morning or afternoon. And much more.
One challenge with rules based solutions is that they’re based on what companies assume about their customers. And sometimes it’s not right. When you let the machine do it sometimes you’ll be surprised at the reality. It takes out any subjectivity and helps you reach a correct decision.
What’s important about machine learning is that it benefits both businesses and society.
At Experian, we’ve invested heavily in Data Science and Advanced Analytics. This investment isn’t a new one, but an investment we have maintained for years – we keep building on our capabilities, our services, as the need and technology allows. In our DataLabs we have Data Scientists working on research and development projects globally, solving varied and large scale problems faced by our clients. At the other end, we incorporate analytics into all our propositions and models – making the output ‘intelligent’. In the future we see this as even more important and believe that utilising the now advanced techniques of self-learning and deep-learning will bring even more potential and even more accuracy and accessibility to the market to help make better decisions. Better customer decisions.