Data Analytics with Matthew Bernath

On this episode of the podcast series, we are joined by Chris Steyn, Shoprite Technology’s Head of Data Analytics. Serving over 24 million customers in nearly 3 000 stores, with in excess of 25 000 points of sale and 100 000 articles moving through these points, data presents very exciting opportunities for the Shoprite Group and the chance to execute their strategy of becoming smarter. From a data analytics perspective, Chris believes a focus on customer-centricity and the enablement of precision retailing is essential. Customers need to have a rewarding experience when they interact — is the product available, is the quality of the highest standard, is the pricing correct and the in-store experience enjoyable? With precision retailing, it’s all about infusing more science into the decision-making process such as inventory control, wastage and shrinkage, quality control on fresh produce etc. With so many moving parts, there are many opportunities to mine different data assets that currently exist within the organisation.

In order for data to be collected and made useable, many years have been spent on standardising the data pipelines to ensure they are complete, timeous and of a high quality. This led to a data bulk architecture with different channels, exposing more data assets to more of the organisation. ShopriteX is a new digital innovation unit that aims to marry data science, technology and talent with customer-focus at its core; consolidating all the business entities with an emphasis on commerce, rewards, and personalisation.

The focus for retail is the evolution of customers from single to multi-channel and how this activity is tracked and monitored. Location analysis, the introduction of augmented reality in different forms, and the concept of customer offer engines are very important. The old systems targeted the masses, whereas now it’s all about tailoring to the individual customer needs and desires. Shoprite continues to invest more in outside intelligence, and favours an ecosystem where exploration, ideation, research and experimentation is encouraged. They have a saying – productionise your curiosity; create a workforce that is challenging every single norm that exists within the current landscape. For organisations going forward, Chris feels it’s all about investment in human talent and potential and creating an environment where people can be informed, educated and inspired.

On today’s episode of the podcast series, we’re joined by RMB’s Head of Telecommunication, Media & Technology Advisory, Arun Varughese. There is so much happening in the payments landscape, and not many people are aware of the plumbing that exists behind contactless payments. It may happen in mere seconds, but is actually a very complex web of electronic communication both locally and internationally.

Cash is still dominant in Africa and South Africa; 60% of payments are made using physical cash, so the market for card payments and alternative fintechs is massive. The proliferation of smartphones using digital wallet services such as Apple Pay, has enabled consumers to pay for things in a much more convenient manner. Around the world there are payment options such as WeChat, Alipay, PayPal and ever-growing cryptocurrency solutions. With all this innovation and change occurring, there is a trend of moving towards financial inclusion for most of the planet, and the move away from cash is really going to be driving it.

Data is now very much at the fore; within banks there’s a lot of analysis of consumer behaviour, such as where you swiped your card, what you purchased, how often you shop etc. When you tie that up with things such as location data, even from an economic and aggregate perspective, you can see with real-time data which direction the economy is moving.

A new development on South African shores is ‘Buy Now Pay Later’ – a method of payment that’s a direct replacement for credit cards, where everyone benefits – the fintech, the consumer, and the merchant. This system allows one to borrow money on an interest-free basis over a period of time, where you can opt to pay it back in equal installments over a 4-8 week period. It’s a very real threat to traditional banking where interest is made on credit cards and fees. It’s a closed-loop system – the data is extremely valuable; fintechs can build up a much clearer picture of their customers, allowing them to target better products and services.

From an advisory perspective, Arun’s role is to consult players in the value chain on a variety of choices, particularly around capital and how to expand their businesses. As a bank, RMB hopes to grow with their clients in this increasingly active space.

Patrice Rassou, Ashburton Investments’ CIO joins us on this episode of the podcast series. The investment industry is very familiar with the processing and extraction of data in order to make investment decisions. A few decades back this was largely a manual process; now data is available in a digital format with quite a lot of history, allowing for spreadsheets to be built rapidly. There are also alternative sources of data available to triangulate forecasts with what industry experts are saying, thereby improving the quality of these forecasts and also the quality of valuations.

There’s a lot of raw data available, but the real issue is how easy or difficult it is to arrange and process that data in order to be able to extract signals, and how useful these are from a predictive viewpoint. To explore these signals in financial markets, Patrice believes you need to see whether there are relationships that you can observe, and are these persistent and significant enough to make money out of. The volume of data, which is into the billions, really shows that there is a vast field to mine of both structured and unstructured data which requires a lot of computing power to handle the frequency and velocity. The key focus in financial markets now is getting the signals out of this noise. The differentiating factor is making sense of the relationships, and ensuring the data will have a lasting impact that can be explored long-term.

What is alternative data? Instead of using traditional financial information, there has been a whole industry that has sprung up with a swathe of available data. Some examples are satellite imagery that gives data on mobility, traffic outside of shopping malls, ships piling up outside of ports, trucks coming in and out of factories and so forth. There are also social media indicators allowing you to pick up whether your average person is feeling more bullish or more bearish; what their type of buying habits are, what interests them in terms of financial markets chatter, etc. Alternative data used to be quite costly to collect and difficult to integrate, where increasingly there are providers who supply these alternative sources of data, sometimes to the highest bidder to avoid overcrowding.

The world of financial analysis has really evolved from being one driven by humans, to having machines and very sophisticated quantity processes able to handle and analyse huge amounts of data and deliver returns; one Patrice feels is a race between man and machine.

Joining us today on this episode of the podcast series is Mike Grant - Chief Technical Officer at DataProphet, engineer and machine learning specialist. DataProphet is not a dashboarding company. There’s no added value outside of a beautiful looking dashboard where you have to take your production team to look at the dashboard; interpret the graphics and end up having 2 different people looking at the same graph and coming to 2 different conclusions. AI takes out that interpretation component and analyses the data a bit further to a point of action; formulating a plan as to what you should do next. A manufacturing line is a super complicated amalgamation of machines; you do something upstream and it can affect the next process. It’s very hard to say — ‘make this small correction here because you don’t yet have a problem, but if you don’t fix it, you’re going to have a very costly mistake’. It’s that guidance that really affects bottom-line profitability — if you can get it right, you can transform the manufacturing company into something that is hyper competitive.

Mike believes to undertake the AI journey, you need to start with a value proposition; make sure that whatever comes out of the AI system has a tangible ROI. This means operating at a price point where it’s affordable to the manufacturing firm, and the value it adds to the business warrants the expenditure. For example, if scrap is reduced by 40%, what does that mean to my bottom line? Among other things, it means an immediate cost improvement, additional production capacity, lower energy consumption and less CO² emissions. When it comes to having enough data, you don’t need an historical record. It’s like the saying — ‘The best time to plant a tree was 20 years ago; the next best time to plant a tree is now’. So, if you don’t have it, just start collecting it.

As far as the concern around AI replacing jobs, Mike feels this isn’t the case at all. There’s a massive absence of ‘experts’ around the world, so implementing AI is actually creating this input to drive a production system – upskilling existing machine operators to perform better. At the end of the day, it’s effectively going to augment our current processes and make businesses more competitive on the global stage. 

On today’s episode of the podcast series, we’re joined by Property Economist for FNB, John Loos who believes that data analytics in the property industry is of vital importance. Having been in the industry for 20 years, he has seen firsthand how limited data back then led to billions of Rands worth of losses due to banks not fully understanding the housing boom. There was only house price data, no housing market data, which meant it was a speculative market with very little information on what was actually happening. All these things lead to a very high-risk environment for banks. If another housing boom took place today, banks would be inside the mind of the market and the dynamics much better than they were back then.

John feels offensive analytics is not used as much as it should be on the banking side of things. When it comes to housing bubbles, such as in the US at the moment, there’s plenty of data but not enough buyer or seller education, meaning buyer panic still sets in. Data has taken us so far, but he believes it needs to be used more and be used to educate.

More and more diverse data sets are being collected regarding properties and valuations and different metrics; we ask if this as a property economist has made it easier or more difficult to find the signal in all the data noise. The answer is definitely easier. The insights available now compared with 20 years ago are phenomenal. One example was managing to identify that the Namibian market was overheated using the methodologies that had been built up for South Africa.

Do you think AI will become more prominent in the property industry, specifically in banking and asset management, or do you think there will always need to be a human touch? Where there have been mistakes in the past, and there will be big mistakes again, is that machines or models must give you 100% of the answer. What was always drummed into John from early on is that machines/models are there to help you make the decision. If you allow them to actually make the decision, you’re going to get yourself into trouble; as a rule of thumb there should always be humans behind the scenes applying their minds.