Rex is currently Chief Data Scientist of dunnhumby. Based in the USA, he leads a team of almost 500 data scientists across the Americas, Europe and APAC.
dunnhumby is global leading data science company. It specialises in creating unique insights for consumer organisations in order to help them to make better decisions and drive data monetisation opportunities. Its clients include some of the world’s largest retailers and many of the major Consumer Package Goods (CPGs) companies.
Having joined dunnhumby in 2009, Rex has held a number of roles for the company including Managing Director India and spent three years based in London as their Chief Technology Officer. Prior to dunnhumby, Rex was responsible for Loyalty at Coles Supermarkets, the US$27bn, Australian supermarket, liquor and convenience group.
Rex shares with Barracuda Search his thoughts on the current digital transformation landscape, how to manage data science teams and the role that AI-powered technologies will play moving forward.
With increased customer demand and greater internal willingness to embrace new ways of working, the last few months have seen a dramatic acceleration in many organisations’ digital transformations. What are your observations on the way that businesses have been approaching the use of digital during this period?
I see a lot of organisations at the moment realise that they have to do a range of things differently, both in terms of how they face their customers and in terms of how they set up their own organisation. Things that they probably knew were true before COVID have become very true during COVID.
Let me give you a small example of what I mean. For years people have been talking about organisational structure and having the right resources and the right teams in the right places. Suddenly COVID has given us a situation where you’ve been forced to do things remotely across teams in different places and people have had to find means and mechanisms to make this work.
During this time particularly larger organisations have recognised the fact that actually there’s a whole lot of things that they have been doing in a fairly ineffective way and they now need to do it far more effectively. So I see organisations looking internally and trying to clean up a lot of their processes, tidy up a lot of the unnecessary steps and to operate much more sharply.
At the same time over the last few months certain organisations, some retailers for example, have had a huge short-term sales boost due to a whole range of changed shopping behaviours. These retailers are recognising though that if we fast forward a year’s time then they may have comparable sales that look very different, partly as a result of last year’s numbers but also the world will be different. So they’re gearing themselves up to try and sustain that higher advantage or alternatively have a cost base that matches the change in environment that they will see.
So you’ve got those two elements: what they’re doing internally to tidy up their internal processes and do a digital transformation from that perspective, but also what they are doing externally to get the customer offer to where it needs to be.
You’ve spoken previously about the importance of process when approaching these sorts of transformations, particularly when introducing AI-powered technologies. When you are sitting down with your clients, what is it you ask them in order to determine how best to approach their transformation?
One of the very, very first things we do is to encourage people to make sure that their data is structured in some shape or form so that we are able to achieve the sort of things that they want us to do. This relates to pretty much any potential client that dunnhumby works with.
Ultimately one of the biggest mistakes that we see almost every organisation making is that they want to start with, let’s call it the sexy data science part of the transformation, without first doing the hard graft and organising their data.
One of the hardest things that we have found over many, many years in this space is just working with an organisation to get a view of the customer that is consistent across that organisation. It’s one of the hardest pieces to achieve in the first place. Whether it’s a CPG, a retailer, an insurance company, financial services, we still see a lot of pain in their ability to get a proper view of the customer.
A data scientist needs to be working off some degree of order. Let’s just think about the CPGs for a moment. They’ve got a bit of data here, a bit of data there and feeds that come from other parts of the organisation. Stitching all of that together to get a proper viewpoint of the customer is one of the first investments that they need to make and to get right for the rest to be a success. Ultimately, you’re not going to even be able to start this journey unless you’ve got some degree of cleanliness in terms of how the customer data is presented in your organisation.
What I would add is that this needs to be done in a way that is consistent and ready for all the privacy and consumer laws changes that are going on in the world today, so when you are making these investments you need to be clear that your people aren’t running around and storing or using data in ways that you didn’t intend.
For many businesses, being able to provide a single view of the customer may seem like a daunting feat. Do most organisations already have the systems in place that would enable them to do this or are we looking at a significant investment in terms of the re-platforming?
Very good question! We come across a lot of different situations.
We will find that we might hit an organisation that has just made a huge investment and re-platformed for a reporting purpose for example. Putting a customer information warehouse, or something like that, on top of that platform is usually less of a job but what we do find is that because the platform wasn’t designed with this functionality in mind then we need to do quite a lot of work to make it fit for purpose.
We have other clients who just simply haven’t got this viewpoint. The organisation may be very siloed functionally. The finance team has what it needs, the marketing team has what it wants and the merchants have what they need to operate but none of them are really talking to each other.
What we often do first off in these situations is a data consulting piece to stitch together a proper customer viewpoint. This is usually possible if we think about it carefully because re-platforming for some of these organisations is just horrendous right. Even today, if you’re trying to do a transformation it’s a brave CTO to turn around and say that they need a complete change of platform. Often it’s just a case of being able to get this feed to match up to that feed and so the investment doesn’t need to be as much.
AI-powered personalisation and recommendation tools have seen some great development in recent years. When we look at a transformation, which other areas can we expect AI-powered technologies to be deployed in that may have a more far reaching impact on the organisation?
Well we have built some incredibly advanced personalisation, recommendation and ‘have you forgotten’ algorithms and it has come a long way from when I started in this field. I do feel though for many retailers there is still a long way to go with this.
When you think of all the years of development and the amount of investment that the market leaders have put into this and even a lot of that is still quite simple and doesn’t really factor things like seasonality, which means you can get recommendations that are not very relevant.
With personalisation and recommenders, people like to get their hands off the wheels with these things very quickly. They think, “Ah brilliant I’ve got an algorithm and it’s going to be a money printing machine, fantastic”. But if you actually sit back in your day to day life and think about the range of irrelevant communications and advertisements that people are trying to put in front of your eyes then you realise that a lot of tweaking and refinement needs to be done to make it effective.
If we just talk about seasonality for a moment, Asia is a great example of opportunity because of the amount of holidays and festivals that lead to specific buying behaviour at a specific point in time. We’re coming up to Diwali in India as an example. Well there will be a whole series of things that are bought in October and November which should be promoted at Diwali, not any time else and certainly not in December. In the USA, I’m always amazed that I have major retailers recommending that I buy a Halloween costume in April!
Beyond personalisation, without question one of the more portable areas that we see organisations applying AI is in supply chain. It’s an area where there is obvious benefits and we are seeing a lot of AI-related investment in the supply chain world. If you follow this through and think about where businesses are focussed at the moment from a digital transformation point of view, there is more focus on ‘Back of House’ areas. That sort of a robotic process automation focus is where I am seeing the quickest use of some of the AI techniques and may even be a better place to start for some organisations.
As the power and importance of AI increases, we have seen consumer organisations around the world building their own teams of internal data scientists in recent years. What are the challenges that you have experienced in terms of integrating these teams into the wider organisation and what can be done to overcome this?
Let me address a couple of areas with this.
First off, I can’t think of a single major retailer or CPG who isn’t trying to set up their own data science organisation. Everyone is sitting there realising that they need to do something with their data and they’re trying to work out what that something is. Eventually someone will say, “We need a data science team. We need a role that’s specifically focussed on taking this forward and generating value”. When they get to this stage there are a few things that organisations need to think through.
The first one is about getting the right people. You are dealing with a world in which there is a huge demand for these skillsets. There is rarely a list that comes out now that doesn’t have data science as either number one or number two in the “high demand jobs of today” of either this year or this decade or even this century. So you’re dealing in a world where there is a lot of demand and you’ve got to go out to attract and hire people that you probably haven’t hired before.
You need to think about retention. To keep data scientists engaged, you’ve got to have interesting problems for them to solve. Data scientists are absolutely drawn to a problem and are noted and famous for the fact that they will worry less about the money and are much more focussed on what problems they are trying to solve. So you have to have your ecosystem set up right to allow for them to have an impact or you’re going to struggle to retain the talent.
You need a range of leadership and management skills to be able to inspire and manage them. Unless they are being guided in the right way, they’re probably not going to come up with great recommendations. That’s usually simply because they haven’t been given the proper guidance and therefore not setup for success.
One really big nugget that I would give to people, especially larger organisations, is that you’re going to have to share. Let me say what I mean by that. If you’re a retailer, don’t think that you, the merchant team can have your own data science team and you, the marketing team can have your own data science team. If you’re starting this up, you’re going to have to have some form of central team. You’re going to have to use your economies of scale and share.
How important is structure, how these teams are physically structured and placed within an organisation?
Vital, just vital! I have had many colleagues and friends who have taken on newly established leadership roles for Data and Data Science within different organisations and nothing has been set up for success because of how the team is structured. It’s so important for organisations to get this right.
The team needs to be in a part of the organisation that has some form of clout. Different organisations are led by whole different ranges of people so where you place the team will depend on the nature of the business. If it is a merchant-led organisation then you place the data science team in there. If it is an operations-led organisation, then you place the team in there. They need to be in a part of the organisation that has clout because otherwise it’s not going to have a real impact. It will come up with some interesting curiosities and future learning pieces but inevitably you won’t see meaningful impact and the whole thing will just die a natural death.
Is it a challenge that there may be a lack industry knowledge when putting together a team of data scientists? For example they may not have ever worked in a retail or consumer organisation previously?
For sure but data scientists don’t necessarily need to have a deep domain knowledge. You’re not necessarily going to expect them to have that, but they certainly need to be guided.
We get a lot of requests for help because an organisation finds that that their data science team is telling them things that are either unactionable or so mind-boggling obvious that they can’t comprehend why the team spent so much time and effort coming up with the answer. This is down to lack of proper guidance.
The team will need to have a lot of guidance around what is relevant and what is not. What information will enhance the business and what information will just confuse or frustrate people. This guidance needs to come from someone who can be an interpreter between the data science team and the broader business. They are going to be the bridge to the business and they are the one that really does need to have the knowledge and understanding of the commercial business and the industry in general.
I want to pick up on one of the things that you mentioned earlier around the regulatory and legal landscape when managing data. What considerations or risks do you think that businesses are facing up to if they are not being mindful of their responsibilities when it comes to their customer’s data?
If people are not putting serious consideration right now into the regulatory, privacy and security elements that relate to this level of work then they’re making mistakes.
This landscape is changing rapidly. I am lucky because I have data scientists and clients working both on CCPA (California Consumer Privacy Act) in the US and on GDPR (General Data Protection Regulation) in Europe. Since I’ve got clients in these jurisdictions we have had to think very seriously about data regulation from the moment these legislations were proposed.
I am convinced that most countries around the world are going to be imposing some version of these GDPR, CCPA-style regulations and that if people are not keeping an eye on this, they’re setting themselves up for a lot of difficulties in the future. All the trends point to a higher degree of transparency and accountability for businesses.
There is a whole range of basic activities that organisations need to be doing now to make sure that they are operating properly. Even ignoring regulations, just think about the reputational damage that is done by businesses not having their act together and having a bridge that leads to names, addresses, credit cards or other elements. If that kind of system is compromised, then the damage can be huge.
Some companies are making big bets with their investments around data at the moment and if you haven’t given this area enough thought then you’re going to be in for a set of very, very awkward situations down the line.
Many organisations have used this year to reset their business and adopt a ‘day one’ mindset. How do you think that organisations should be considering the role of AI-powered technologies in order to build themselves moving forward?
You can look at AI in two parts. You can look at it in terms of things that will make you leaner and you can look at it in terms of things that will make you smarter. Those are the two elements that you break down into from an AI or Machine learning perspective.
When you consider things that make you leaner, there is a lot of AI that simply removes labour, introduces automation and makes you more effective and sharper. Basically it will improve what you are currently doing.
The other side is about making you smarter. You can implement tools to give you better answers, to answer things you didn’t know, the ability to understand your supply chain better, to understand your customer better and to deliver smarter promotions. The things that you couldn’t see before but now with the combination of data, analysis and advanced techniques you can.
If you haven’t got a bit of weight on both of these areas then I would say that we are starting to get to a stage where you will get left behind. The number of organisations that are getting beyond just playing with these areas and are now really implementing means that the gap will open up in the next five years. It’s not immediate and you can get by for a bit longer but then at some point it will get harder to catch up.
Now is the time for organisations to get really serious about this and at the very least be putting the building blocks in place in terms of people, data, systems and processes. All the while, you have also got to remain mindful of the security and regulatory considerations.
Interview conducted by Barracuda’s Head of Hong Kong, Max Holdsworth.