From Insights to Impact: Making Data Work for Customers with Peter Laflin of Morrisons

In this episode, Peter Laflin, Director of Data and Analytics at Morrisons, shares how his team uses data to improve the customer experience and drive measurable business results. He discusses aligning data initiatives with business goals, building a diverse team and ensuring data governance in the age of AI.

‍

Key Takeaways:

‍

00:00 Introduction.

02:45 Starting with customers helps solve problems, like finding cranberry sauce at Christmastime.

05:50 Data team members work in stores to build empathy and improve the shopping experience.

10:11 Success is measured by business impact, not tickets or code.

14:39 High-performing teams win or lose together, driving customer satisfaction and growth.

22:20 The right environment helps people thrive and do their best work.

25:05 Different thinking styles spark better ideas and stronger solutions.

30:41 Centralize for speed, then decentralize once data foundations are strong.

38:10 The age of data governance demands trust so AI can enhance human judgment.

‍

Peter: [00:00:00] key thing there is, you know, it'd be so easy for us to build an application that allows you to search by product id. but the customers don't work that way. You know how many customers know the product? Id probably zero. There might be a handful who work in the business, but the point there is you have to, you have to meet the customer where they are, not where we want them to be.

Anthony: Welcome back to Data Masters. Today I'm thrilled to be joined by a true leader in the data and analytics space. Peter Laughlin. Peter is the director of data and Analytics at [00:01:00] Mortenson's, one of the UK's largest supermarket chains. With over two decades of experience helping organizations turn data into effective business strategy, Peter has a wealth of knowledge to share.

‍

He's been recognized as a leader in his field being named one of Global Chief Data Officers, 100 and a member of DataIQ 100. In our conversation today, we're gonna dive deep into the practical realities of building and leading a high performance data function. We'll explore the challenges of governance, the importance of maintaining an independent data team as an arbiter of truth, and how to build a team that's measured not by technology metrics, but by the direct impact it has on the business.

‍

Peter, welcome to the show.

‍

Peter: It's a pleasure to be here. Thanks for inviting me.

‍

Anthony: Awesome. So let's, let's jump right into it. So, you've said in the past that, when you're thinking about a data project, it's [00:02:00] sort of a, a good idea, like a good strategy is to start by thinking from the perspective of a customer, um, and. When we were talking ahead of this, recording, you gave a really a great example of in the supermarket context of how to think about the supermarket experience through the eyes in a very literal sense, of a customer trying to find something in the shop.

‍

so. Maybe share. I thought it was such a great example of thinking, not from a technical perspective, but thinking from a customer perspective. Maybe, if you don't mind, share the example and why It sort of illustrates this dichotomy between, I think our, intuition, which is start with a technology, but rather your view of starting with the customer.

‍

Peter: Yes, everything is better when you start with a customer because, it forces us to solve the right problem rather than the problem that we think we have, The technology can do that [00:03:00] thing. so it's, it's very important to me. It's very important to our business that the data team in, in fact, all of our head office functions, spend time in our shops with our customers, listening, helping, and responding.

‍

And, there was a, a really. the sort of memorable point when, over Christmas, customers came to me and, and asked for, help finding particular products. So at Christmas time, people want to find things like cranberry sauce. They wanna find particular gravy these are great products, but actually can be a little bit tricky to find in the shop if you're not familiar with the store layout. So, I, I sort of, you know, helped a few customers and, you know, by the time the fifth or the sixth customer came across, I thought, actually, there must be a better way of, of doing this rather than relying on my memory, which, sometimes isn't, isn't great. and I, I realized that because our data's in the cloud, I actually had access to this information on my phone. so, during my [00:04:00] break, I, I started, sort of building a, a, a quick query, to, to look up our, product locations in our shop. And eventually that, that process led me to think that we could build this feature into our loyalty app that would help customers, in real time type in the product that they were looking for, and find that product, really quickly and really easily in the supermarket Now. It's a really simple idea, and often that that is, one of the best things to, to try and develop is something that actually is, is really super simple to explain and, and, and to understand. but behind the scenes there's a lot of really interesting data technology, AI that's coming together in that moment, in that millisecond to help help the customer.

‍

So, the, the team and, you know, my, my whole team has, has sort of chipped in and sort of helped develop this particular product. And it, it's. It's a fascinating, example of data engineering, machine learning, ai, but at the heart of it is that real desire to, to [00:05:00] help our customers do something that helps them, find those products and make it easier to shop.

‍

Anthony: So be, before we dig into the technology, I'm gonna be a little curious on exactly what it was I, what I appreciate about the example is, the intuitive. Idea from a supermarket shopping perspective would be to let people buy things through a mobile app or through a website. Like, you know, that's the, the business you're in, so to speak, I would think is selling groceries.

‍

but the insight that you've gained, from the experience with the customers. There's a, a really simple first step, which is, and I, I think everybody's had this experience of being in the grocery store in a rush, and you're like, oh, where is the. And then not, you know, not because you wanna buy it online, but because you just wanna go find the thing in the store.

‍

do you literally have folks on the data team, like literally working in the stores? Were you actually in a store?

‍

Peter: Absolutely. It's a really important part of our culture [00:06:00] here at Morrisons. Very important that all of my team, are out in our shops serving our customers. Now, you know, across the course of a year, we might spend two or three days, individually, doing that. That adds up to, to a lot of experience and a lot of understanding,

‍

Anthony: And empathy.

‍

Peter: Exactly. It helps us see things through the customer's eyes, so that we can find, things that we can do better to either improve our availability so that's, you know, where the product's in the right place at the right time or improve that, shopping experience. and you mentioned there about. The, the sort of digital experience, you're absolutely right. Customers can go online, they can buy these products, have them delivered to their home through either our home delivery or through some of our, other partners like Just Eat or Deliveroo. but when you're in the shop, you actually want to find the product there and then, so, we, we we're very clear that we wanted a similar search experience, but just for the customer that was in the store. And when you start to get into the, into the project itself, when you get into the mindset of how do we do [00:07:00] this? there are lots of things that we have to bring together to make that work. So we have to know where the product is. We have to know what the customer is looking for. you know, I've been doing this a long time.

‍

Gone are the days where. We can expect the customer to write exactly what they mean spelled perfectly, because we're so used to autocorrect and we're so used to things being finished off for us that we needed to build an experience that was tolerant of spelling mistakes, was able to

‍

Anthony: So you're saying that none of your customers work in a grocery store and have an intuitive understanding of all the product names and categories? A shocker.

‍

Peter: Absolutely. And the key thing there is, you know, it'd be so easy for us to build an application that allows you to search by product id. but the customers don't work that way. You know how many customers know the product? Id probably zero. There might be a handful who work in the business, but the point there is you have to, you have to meet the customer where they are, not where we want them to be.

‍

So a large part of the engineering we [00:08:00] did on that project was to build an intuitive interface to allow the customer to type what they were thinking and for us to. Interpolate that into a set of product IDs that we could then use to look up against our core data assets. So the way we do that is we have a number of wording embeddings.

‍

We use a number of Google products that help us, with the natural language processing. And so, you know, from a, from an architecture perspective, there's a number of key things that have to come together, within a second. In order to provide that really, you know, fantastic experience for the customer, they're not going to be tolerant if we have to spend 20 seconds about what they've asked, working through all of the data calls. You know, this has to be instant. And again, that was a large part of the, the engineering challenge. but going back to your original question. You know, that's what I so love about doing this job is that there are so many exciting technical solutions that we can, we can imagine, we can build, we can deploy, ultimately it's only useful if it's solving a [00:09:00] problem that the customer has or it, it's addressing a need that's going to lead to a better customer experience or going to enable us to, have better availability or a better understanding of the products that we might not have, that the customers want to buy in the future.

‍

Anthony: Yeah. No, and again, I, and I think the, the insight that comes from spending time. In the actual work of the business gives you those insights about the kinds of solutions that are actually useful to customers as they're in, in your exam. In your example of actually when they're in the shop and having that moment.

‍

And I think this links nicely to something else. So you've asked, the data team to spend time. In the store working with customers, but it also comes to how you think about measuring success as from a team perspective. and I think it's very common to think about measuring data teams or any technical team by the work they do.

‍

So how much code did you write? How many, you know, JIRA. [00:10:00] Tickets. Did you close? you know, this kind of thing. but that's not how you come at the problem and maybe share the approach you have, which I think is probably slightly controversial.

‍

Peter: So we do measure, Things like you've suggested. So we, we do measure, you know, how many Jira tickets we've got. I dunno whether we measure how many lines of code we write, but, you know, we are interested in the process of taking an idea and shipping a data product. but what really matters is how that gets used in the business.

‍

So I would never want my team to be measured by those, metrics alone because how we do something is not why we do something. And therefore it's really important to frame the measurement of success in terms of how many extra sales did the work we do generate? did we save cost? Were we able to, you know, help our supply chain teams or our logistics teams, save some money on moving products around?

‍

Or did we reduce the amount of waste that we have? because one of the hard things about selling fresh food, for [00:11:00] example, is that. you only have a certain number of days where that product is fit for sale. And so if you order too much at the wrong time, then you have to throw some of it away. And again, we wanna minimize that.

‍

So let's get measured, let's think about success in terms of how much waste we are saving rather than how many tickets that we have open or at any one time. And I think one of the, the things I've tried to bring to the team over the last 18 months. Two years is, is that shift in mentality from, you know, we are less about being project orientated, more about being product orientated? Well, you could fall into the trap of them measuring ourselves as a product business. So, you know, how many new features have we released? How many new, story points have we delivered? how long is our backlog? What's our time to delivery? All of these things are really, really important. However they mean for nothing. those products that we're shipping, if those, developments that we are engaged in, the [00:12:00] capabilities that we are, ultimately making available to our colleagues, if they don't actually drive meaningful change in the organization, then it's a bit pointless. and so it, it can be difficult. To agree on the precise value that we have delivered. because I see our team as an enabling team, it's one that is there to help other teams, other functions, do their best work. So it's, you know, it can be a little bit challenging to say, you know, without us, did you, you know, would you have been 10%, 20%, 30%, less efficient, less effective? At the end of the day, we, you know, it's very important to see this as, as a team game. we are here to make the team so team Morrisons be more effective, and more able to, lean into those, those business KPIs. sometimes find people find the approach a little bit confusing, you know. But it's about shifting your thinking from not measuring the how to, measuring the why. You still have to measure the how, because unless you do that actually, you know, your, your, your scrum [00:13:00] mentality, your, your kind of squads, they can, they can sometimes struggle to understand what good looks like compared with other, other squads and other scrums that might be running at times. and especially when you work with, with offshore teams. You know, development partners is important still to talk in their language. but that's a means to an end. It's, it's so super important to understand the ultimate impact that you are having on the overall organization.

‍

Anthony: I've said this before on the podcast, but I have this mildly controversial point of view that any work we do, any, whether it's software engineering, data engineering work, any work we do can really be boiled down to one of three business impacts. We're either driving revenue. Saving cost or reducing risk.

‍

And if you, if, if you force a team into thinking like, well, what bucket does that fall? Does the work I'm doing fall into? And if it, if you can't fit into one of those buckets, you should one, be questioning it. And if it can fit into those buckets, to your point, mapping that back to the business [00:14:00] metrics that are underlie those, those ideas, that's really a, a kind of critical piece of the planning process.

‍

I think the common objection we get or I hear. Is I did my job, but you know, the company didn't do their job or this other team didn't do their job, which I likened to, the surgery was successful, but the patient died. but when you get a, a new team member and they object, how do you respond?

‍

Peter: I think it's important to understand what success is. You know, if you are a, a soccer team, you know, based in the uk you'd think about soccer a a lot.

‍

Anthony: Football, but fair enough.

‍

Peter: yeah, well, you know, however you wanna refer to it. If, if, you know, if you are that football, soccer team, it doesn't matter whether I've played, an amazing game, if the team loses. because our success is based on whether we win or lose success is whether you win trophies, you get promoted, you win the league. and it should be a similar mentality across a high performing team. I think one of the definitions of a high performing team is [00:15:00] we all win or lose together. and we all have our parts to play.

‍

And, and yes, there will be days where some of us do better. Some of us do do not so well, our success is do our customers find the great products that they want to buy from us? And are we growing market share? Are we doing that profitably? And yes, there are lots and lots of inputs to that, but across an organization you have to win or lose together.

‍

So, I think it's a mindset piece is that yes, there will be people in my team who are obsessing about a particular new. Algorithm or a particular new implementation of, of Gemini or, you know, whatever it might be. And that's so super important and we, we have to maintain the enthusiasm and the drive and the, the real desire to do that.

‍

But for why, why are you doing that? Why are you investing in that? Ultimately bring it back to. How does that help, our customers find their products quicker, faster, more effectively? How does that help us think about price? How does that help us think about all the other [00:16:00] parts of the, the customer journey that we need to be, be thinking about?

‍

And if, if, if ultimately the work that we're doing doesn't do that, then I would question whether it's valuable or not. So Yes, and people find that quite challenging. I think there's a broader point around high performing teams is it's less about, I more about we.

‍

And I think it's really important that you instill across a team. this is not about a group of, unicorn data scientists or superstar data analysts. You know, at the end of the day we're, we're all one team and, and we all win, all lose together. And therefore, it's great if we have those superstars in the team, but how are they all coming together to ensure that together we win, not individuals win. 

‍

Anthony: Let's talk about the team. 'cause you've grown the team at Morrison's from, as I understand it, a team of just you to a team of 60. That's, that's quite a bit. and that alone I think is, commendable and I'd love to hear your perspective on what it takes to do that. But [00:17:00] one of the things you mentioned is that the team itself is very.

‍

Diverse and probably has a, a range of skills and interests, capabilities, personalities. and I think this is one of these things that people often talk about that they want to have a diverse team with diverse perspectives, but as a practical matter can be quite difficult. maybe walk through what it's taken to grow the team, you know, where you've had missteps, where you've had successes, or been able to build a team that is diverse.

‍

Peter: There's a few really interesting points, that I need to cover to answer your question properly there. let me split it down into two. I think that the first part is, I was data scientist, number one in an organization that had data, but was looking at how to use it, more effectively. the team I. In the first few months, the first year or so is very different to the team I have now because when you have a small team, you do have to have a breadth of experience, you know, storytelling, the ability to interact with stakeholders is as important as your [00:18:00] ability to be awesome coder or a, you know, a really inspirational sort of data scientist that's, that's pushing the boundaries of, of what's possible. So as you grow that team, as you build confidence in that team, you can afford to take greater risks. because in a team of two or three, one individual, so one bad hire is actually a significant drag on the team and on the organization. and so I made a conscious choice in the early days to. some very smart, very bright people. very lucky to work with some amazing people in those early days. But I hired them specifically because I knew that they would interact with the business in a particular way. They would be able to do the work, they might be able to do a bit more data engineering, they might, you might like from a data scientist.

‍

They were able to sort of go to meetings and feel inspired by the opportunity as opposed to feeling frustrated that they had to look up from, from their, their Jupyter notebook. But over time you can take risks, you can bring in different, people with different mindsets. and therefore it, [00:19:00] it's about, I talk a lot about, leaning into the wind. So this idea of you can't be too rigid, I think, in your approach because if you do, if you, you know, if you are a tree and you don't bend with the wind, you will snap. it, it's very important as you're growing a function to be quite agile in your thinking. And I think that plasticity of thought, I think was quite important as we grew the team because, you know, I was very, very happy to, to acknowledge that at times we made a few mistakes.

‍

We went down a few rabbit holes. We, we tried to build the team in a certain way and realized it was a bit too soon. And, you know, this idea of, of data products, it, it's only something that we've been thinking about relatively recently. I tried to do it three or four years ago, but the team wasn't mature enough.

‍

We weren't, stable enough in the business for it to really fly therefore you, you have to adapt. It's, uh, you know, back to the, sporting analogy, I think the coaches that are so successful, are the ones that are clear on the non-negotiables, but able to adapt the things that make sense in the moment. The other [00:20:00] point you made in your question was around diversity. And I think, you know, I, talked a lot about, neurodiversity, but the same is true for any kind of diversity, whether it's, across all the, the different gamuts that, that that brings. I think it's really important to live and breathe it in the team in a way that it, it's less about a box tick exercise. I'm not being critical there. I, I maybe a. Trip myself up with the language sometimes. So please don't see that as a, as a thing for, you know, for people to jump on in the comments. But I, it can, it can turn into that if you're not careful. So this is about ensuring that, everybody in the team has a very, you know, shared goal, but the way in which those people will, will deliver against the goal could be very different.

‍

You might do your best work at seven in the morning. You might do your best work at 10 at night. You might be the sort of person that loves pitch. But equally, you might be somebody who loves diagrams. So rather than having a very fixed way of working it, it's really helpful to focus around the outcomes and make sure that as long as you're delivering to the [00:21:00] outcome, how you do that is down to you and how you work best. And I think, it's been great to see a number of neurodiverse people in it across the team really flourish in an environment where we enable that. To happen. we do that in a couple of ways. You know, things, very, very simple Things like making sure that every meeting has a purpose. that's really helpful for people who are neurodiverse, but it's actually just really helpful for humans in general. let's value our time as opposed to showing up for a meeting that we're not clearer about the objective for. You know, making sure that we, give people who need it time to reflect before making a decision. Because you know, certainly in retail, people are very good at making really quick, very instinctive, insightful decisions. not everybody can do that, and that's okay. So let's, let's build a safe space where if you need a few hours to just reflect and organize your thoughts. That's cool. You know, not everybody wants to write things down, but some people need to. So it's about having those ways of working in the team that mean that people feel safe. A lot [00:22:00] of the ability to flourish is down to the environment, and it's down to removing an element of fear brains don't work brilliantly well when they're frightened, or distressed. And therefore, how can you create that sort of safe space? I'll be honest, we don't always get it right. It's not that every day's a a, a land full of unicorns riding on pink rainbows.

‍

It's, it's not that easy. and it's a journey, you know, it's, it's a journey to try and have more good days than bad days. but yeah, I think there are simple things, but you can do in a team. But again, leading by example. You know, I, talk a lot about my daughter and how proud I am of her. She's, she's autistic, and I've seen her struggle and I've seen her need accommodations, but with the right environment, with the right accommodation, she, she flourishes and I'm so super proud of her and what, what she's able to achieve. It's only right that we're able to do that for our colleagues because at the end of the day, we're all humans. We want. We want to get the best out of people. and so, it's, it's an area I get quite passionate about. 

‍

[00:23:00] 

‍

Anthony: No, no. So I, what I love about that, and just to connect these two ideas, is I think there's often this sense that diversity and performance are at loggerheads. you can either be, you know, you can either foster a community of diverse ideas and opinions, or you can have a high performance team. And I think, if I can put words in your mouth, your point is no actually that by focusing the team as a unit.

‍

On these [00:24:00] business objectives that actually fosters diversity because it allows you to focus the whole team on creating the outcome, the business outcome, the customer experience, and not, inwardly focused on the how you get things done. Is that fair?

‍

Peter: Absolutely. I, I think that the more we can align on why we do things and why it matters, the less the how matters we can trust that we're all professionals, we're all, you know, you know, experienced people or on a, on a journey to learn more. therefore we can trust the how will be done in the right way. You know, my, my job as a, a leader is not to obsess too much about the how, it's to unite around the why. And I think if we can do that and we can align that to the business outcomes, then it gives a safe space for people to do the how slightly differently, so long as the outcome is the thing that, that we value. and I, I, I very much, you know, encourage. People in the team to share their diverse ideas. You know, we don't all have [00:25:00] to think the same. We don't. And, and there were two layers to that. One is having ideas, d you know, having different ideas. So thinking a bit outside the box. The other is the way in which we cognitively process those ideas as well.

‍

You know, I was in a meeting this morning where a colleague was saying, that. she thinks a lot about, high level concepts and she doesn't need to worry too much about, you know, individual yes, no answers. Whereas I had another colleague in the same room who was like, no, my brain works differently.

‍

I need to enumerate all the possibilities in order for me to move onto the next step, whereas. the lady that I talked about in the, in the first instance there, she was happy to take things at face value and not have to prove everything before you can move on. great that we are now in a way, in a, in a world where we can embrace that in a meeting and accept that we, we don't all think the same, but that actually gives us amazing and beautiful solutions that just move us forwards.

‍

And and I think very much comes back to, worry a bit less about the how and focus more on the why and the how. Walk on.

‍

Anthony: Got it. And so, [00:26:00] let me shift gears slightly to, at least one area. Of how that you mentioned before, but I wanted to come back to and 'cause I think you have a unique perspective on it, which is, this idea of data products, and in a way the challenge associated. just to connect this back to something you'd said before, you talked about this idea of.

‍

Thinking about data products early, you've returned to it as a concept. and maybe, I think this is one of these areas, we've had people on the podcast a lot talk about data products. Nobody has any clear definition of what a data product is. So one, it would be interested in your definition and secondly, how it connects to.

‍

yeah, to the data work that you're doing, how you think about ownership of data products, how you think about connecting it into the governance systems that you have, but ju maybe just start with a definition and then also how do you think about data products from a, from a technical perspective.

‍

Peter: Yeah, so I think data product, it needs to be solving a business problem. It [00:27:00] you need to be able to find it. So that's where we talk about discoverability. It needs to be governed so we can trust, it needs to be secure. There's a whole list of things that we, we need the data product to have, but it boils down to, it's probably a set of data tables that are exposed out into a dashboard through an API. There there's a range of things there, and I think it's also important to recognize. That data products can feed other data products. So it becomes a little bit of a recursive process where you, you build on top of them as you, as you go. I think that the reason it's important for, for us as an organization right now is, I get probably three or four emails a day from people asking for a dashboard.

‍

You don't need a dashboard. What you need is access to the information, access to the insight. and often,you know, a a, a trap that we can fall into if you like, is, if a data team just services the request, it gets, Blindly and, and it's never blind, but it's, if you just do exactly what you're asked, you will end up repeating a lot of the same work, [00:28:00] and you are in danger of going back and reinventing the wheel every time, as opposed to embracing the, the wheel as the most important thing that humans have ever invented, because it, it changed, changed so much in the history of civilization. and my point here is that, you've gotta interpret what's the business asking for, and what do you need? It's that whole Henry Ford, do you need a horse or do you need a car? and that needs a bit of skill because there are times, and there are many times where you do have to do what you're asked because in the moment it's the right thing. But over time it's important to become more strategic in saying, actually, I keep getting a request for this thing, so therefore I'm gonna build this capability. That means hopefully in three or four months time when that development is finished, you never need to ask me for that ever again because you can selfer. and, it's not. Particularly easy process. It's not a particular linear process and there is as much art in it as there is a science in that. It's about understanding, it's showing the empathy to your colleagues to say, [00:29:00] I've heard what you've asked for, but I understand why you need it, therefore let's try and work together to kind of bridge the gap between what you think you need and what we can provide.

‍

That Will will do a great job 

‍

I was gonna say the, the advice I would give to people is you can't switch to just doing data products exclusively, because there will always be need for a, a data scientist to do a bit of consultancy or to just give a number or, or just do something quickly. It's more around trying to get the team thinking more strategically that says, why do I rebuild every time? Why can't I build once and reuse many times?

‍

Anthony: So one, version of that question that it comes up a lot. Is whether the data team should be embedded as part of the business unit that it's working with or centrally located. And the argument for embedding it with the business, goes very much to I think some of the themes you've talked about, which as well then they're connected closely to the business needs and requirements that that organization has.

‍

To your point, when they get asked questions, they can [00:30:00] just, just do it, right? They have to, think about doing things in a repeatable way. You've very much chosen a, an opposite approach of, of centralizing the team. And to be clear, I don't have a strong, point of view. I think both can work really well.

‍

But what I appreciate about the, the way you've come at the problem is you, you have thought about it and you've, you've thought that, no, actually we want to have this. Central team,maybe talk about that both organizationally, but as it relates to the data products question we were just talking about.

‍

Peter: think success is actually that we decentralize. So I think the journey is to centralize, to then decentralize, but I think there is a time and a place. so one of the things that strikes me is that the central approach enables us to execute quickly because we've got a tight group of people who are used to working in a particular way, and therefore the, the kind of playbook, if you like, is relatively well-defined. and we can execute at pace. However, the domain ownership I [00:31:00] think is really important, and therefore, in the end, you do want to decentralize, but the, the risk of decentralizing is you lose pace and you lose a little bit of control. but ultimately if you've got the right foundational, data products that go across the whole business, then that's less of a risk.

‍

So this is my point around sort of bending into the wind, if you will, that. you know, I, I've read many things about data mesh and data products, and I've spoken to a number of consultancies and they've all got amazing points of view that have been really helpful in kind of helping us evolve our processes. at the end of the day, you've gotta take the, sort of theory and you've gotta work out how you embed that into your organization. And ultimately it comes down to empathy. down to how can you work with your colleagues and your, you know, whether they're senior leaders, whether they are, members of, of, of kind of the, the team spread across the business. It's different depending on the audience, but at the end of the day, you've gotta think about how is this gonna work best? So I often say that the success of our team is, is, is less [00:32:00] about. The algorithms that we write, or the code, or the engineering frameworks, the engineering standards, it's more about how we, how we operate within the business and how we, how we work together with our colleagues.

‍

And as I said before, there's a humility in recognizing that not everything we've tried has worked. and that you do have to change tack and, and that's okay. So long as you learn and you understand why things need to change and that you are constantly putting yourself in the eyes of the customer.

‍

Now, often that is the customer that's in our shops buying our products, but sometimes that's the customer in our organization that is consuming the data that we're providing.

‍

Anthony: so let me, challenge a little bit here. yes, we want the team to work with the business and be embedded and very responsive, but also you have this strong point you make around, truth and having this kind of central authority and independent authority about what reality is.

‍

Because I think one of the risks of, Having a data team that works really closely with the businesses, that they can almost become [00:33:00] enablers for whatever, bad idea or psychosis is coming out of, whatever team is working on. I'm trying to make the extreme example. instead, what you want is a, a data team to be able to, especially in service of the.

‍

Broader mission of the organization, to say, look, no, actually, the data says this. And yes, you think you, you wanted to do this other thing, or you think this was, resulting in these kinds of KPIs, but actually here's the truth. So how do you balance that sense of like, you know, to use your term customer centricity with being an independent arbiter of truth?

‍

Peter: it's very hard. 

‍

Anthony: Fair enough.

‍

Peter: it's very, very important that somebody is, I'm gonna use the phrase, marking somebody's homework. It is not as, as, as kind of bureaucratic as that. It's, this is not about trying to find problems, it is about having a curious mind that says. I've just heard you use this statistic and you've used this other statistic, they [00:34:00] can't both be true.

‍

So which one is it? And being, in an environment where it's safe to challenge, I think that's, that's the important thing. So it's about en enabling curious minds to be able to speak up, not because as we're being disrespectful or we're trying to challenge authority or any of those things far from it, we're trying to enable. Robust debate and to ensure, you know, the, the thing I'd love the most is when our colleagues, we, we can have a debate with them, but actually we, we conclude that everybody knows their subject areas so well. just that we've got a different perspective. You know, I've chosen to ignore, you know, a certain number of stores for a very legitimate reason, or

‍

Anthony: Hmm.

‍

Peter: not included some products because, they're gonna skew the average at the end of the day.

‍

It's the art of debate that. Makes this so fascinating and so empowering, but it's really important that you have that, that function to make sure that we can, you know, in that debate, somebody is, is trying to just keep score a little bit to make sure that we are holding ourselves to account in, in absolutely the [00:35:00] right way. But it's really, really difficult because, you know, senior leaders can be, quite, forceful in their arguments and it can be quite difficult to challenge, and I think there's a certain amount of bravery that's needed. but I'm, I'm thankful that I work in a team where that bravery, you know, the, the biggest gift that we're given is, is, is bravery.

‍

to be challenging, for the right reasons, not because we're trying to be disruptive.

‍

Anthony: Sure. And so, so let's up the ante. I'm, I'm very proud to say we've, we've had a great conversation to this point, without mentioning ai. but here we are. we're obligated to at least, bring it up. But I wanna do it in the context of what you just said, which is. you want the team to be the arbiter of truth.

‍

You want the team to challenge and work in that space. and, and maybe to kind of bring this all together in a kind of final idea, one challenge in this new era of AI is that people can make up real facts. you know, AI has this. Slightly annoying tendency to [00:36:00] sound, very accurate and very true.

‍

And, and the, the opportunity to, I mean, I don't want to go too far down a,a tangent, but like, literally creating videos that look like they're real or creating data that actually appears to have, you know, that sort of fits the shape and it feels right. how have you thought about, AI in the context of some of those challenges? Again, not generally, but just very specifically.

‍

Peter: I think one of the most important human features, as we evolve our relationship with AI is gonna be the ability to question and the ability to be curious and the ability to spot a hole in an argument. use Gemini quite a lot. I think it's a really helpful tool. but there are days when it, will be potentially misleading, whether on our own data or whether just generally in terms of its, its perspective. and therefore more, you know, now more than ever. we have to question and we have to challenge our assumptions and ask people to prove things from base principles. I'm a mathematician by [00:37:00] training. you know, I, I see myself more as a mathematician than a data scientist, you know? and a lot of what being a mathematician teaches you is that there comes a point where you can prove things from the beginning, you know? We have a, you know, a, a construct that enables us to quickly add one and one and two, uh, to make two. And, and that's fine. We all accept that, but ultimately, as a mathematician, we could go and prove that, you know, that's a whole topic for a different, podcast. Let's not get over excited. But my point here is, we should never lose the ability to challenge the information we've been given.

‍

And that brings us full circle back to data governance because in a, in a business world, unless we can be confident that we understand what makes up those data points and what assumptions we've taken to get here, we will ultimately fail. You know, I've seen too often data gets shared. the data is accurate, it's technically correct, but there's an assumption that people are missing, in terms of how to interpret that and how to make good decisions from it, and how quickly the house of cards can fall down if you're not clear on that.

‍

So I think the answer to your question is [00:38:00] actually that this is the age of data governance, because unless we can trust the raw. Source material and understand how our processes have added one and one to make to, then we do run the risk of, increasing the risk in our decisions. and therefore, I, I think, you know, AI should be there to, to help us do more of the things that we need help with. and less of, less there as a, as a sort of intellectual crutch. You know, I, I, I can't remember the exact quote. Now I'm sort of kicking myself. But there's this concept of, I want to use AI to free myself up to do more of the things that as a human I have to do because the AI won't do it for me. I think we're on that journey.

‍

I think there's a danger that as, as you start adopting new technology, that you get carried away with certain things, and we should always, as leaders, be bringing it back to, let's use AI where we need to, where it's going to be. Quick and smart. Back to our product finder. Let's use AI in that, interpretation layer to understand the customer. Brilliant. Put some guardrails around it and enable us to do [00:39:00] more. but let's not lose the ability to challenge each other, and to keep spreading fake news. You know, I think the, the final point I'd make is that that has to be the crux of where we go as, as a, as a society is how do we unpick what's true from what's made up?

‍

And that's getting harder and harder in this world of observations through large language models. so yeah, it's fascinating and scary at the same time.

‍

Anthony: Yeah. No, but look, I think the, the core idea there, which is. Is grounding these AI tools in high quality, trusted, governed truth data that you can actually trace back, to, you know, to source to where it was, built. You know, and that is at its core, almost certainly. gives you the path by which you can start questioning the results, questioning the data, and getting back to that core idea that, you know, and find where there's been hallucination or where there's been stuff made up.

‍

Well, Peter, thank you for the time, the insights and, really appreciate the ideas that [00:40:00] you brought about, connecting back to the customer need. How that aligns to the way you measure the team, how you organize the team and build the team and bring in diverse perspectives. And then even to, how that shows up in the data work you do with data products and ultimately how you can trust and verify that even in the new world of AI that we're jumping into.

‍

But I really appreciate the time today.

‍

Peter: been a brilliant conversation. Thank you so much for having me on the podcast.