Why do most data transformation projects fail? And more importantly, how can data leaders buck this trend and write their own success story?
Benny Benford, founder of Datent and former CDO of Jaguar Land Rover, explains to Harbr co-founder Anthony Cosgrove his key lessons about delivering value through data. They discuss:
You can watch their conversation below or on our Youtube channel. An audio-only version is available on Soundcloud.
Anthony Cosgrove: Hey everyone. I’m Anthony Cosgrove, the co-founder of Harbr. And today I’m joined by Benny Benford, who is the founder of Datent, where he’s creating the world’s first support system for data intrapreneurs. We’re going to be talking a lot more on that in just a moment. Prior to Datent, Benny was working as the CDO for Jaguar Land Rover, so heaps of experience.
Benny, thanks for joining us.
Benny Benford: Thank you for inviting me. Looking forward to this.
Anthony Cosgrove: Awesome. Me too. Let’s start with Datent. Tell us more about what you’re doing now.
Benny Benford: So I got really frustrated as a CDO. It’s quite common startup advice that startups are hard. If you want to do a startup, solve a problem that frustrates you in the world.
Daily transformations don’t succeed often enough. I was very lucky to work on one that did and then a bunch of my peers who were working on similar, quite transformative transformations, all said ‘we’re solving the same problems and that just sounds really inefficient’ and the more I thought about it, the more I realized that’s not true elsewhere. If you’re doing something like rolling out a CRM, there’s a playbook for it, if you’re merging two organizations, there’s a playbook for it. If you’re changing an HR function, your member of an HR professional body that tells you what a modern HR function looks like the data world keeps reinventing itself and so I’m creating a organization, a system to support people who want to change their companies to become more data driven, solve all the problems that are common to data transformation.
So they just need to work on what’s special to them, about their organization.
Anthony Cosgrove: Perfect. And within that, you’ve used the term ‘intrapreneur’. Jus break that down a little bit because it’s quite different to helping people with data transformations. It suggests there’s a sort of a value driver or maybe even a commercial component to the work they’re doing and their objectives.
Benny Benford: Yeah, it’s a good point. I think it’s a term that’s not used enough. Intrapreneurs don’t accept the status quo in their organization and they don’t particularly accept that job being what their job title is and what the description is. Like an entrepreneur doesn’t accept the status quo in the market. And they go and build something different.
The difference is they do it inside an organization. So often they’re the people who to start with might set up a new team or function, but then they try and change things within the organization to, to improve the organization in ways that are measurable about what matters to the organization.
And that’s quite cryptic language, because the more I’ve been doing this, the more I’ve stepped back from the phrase ‘deliver returns’, because actually I know data intrapreneurs in the nonprofit world and delivering — improving their organization is around improving impact from a social perspective. So it’s about how people are using data to improve how their organization delivers whatever their organization has been set up to do in a better way.
Anthony Cosgrove: Perfect. And to sort of pull the conversation into data products and what’s going on there, you were actually quoted in the last podcast by Caroline Zimmerman talking about this idea of delivering incremental value. So, delivering a new insight frequently so that people can start to see value as they go on that journey. Probably something I suspect that you’re encouraging those intrapreneurs to do as well. Talk us through what you see as the outcomes of those transformational products.
Is it products? Is it not products? How would you go about this? Like, is it a data product? Is it not? And if not, why not?
Benny Benford: You’ve asked a lot of questions there. So let’s try and do the split. I don’t think everything has to be a product. A product has an ongoing service obligation.
You know, you’ve got a whole product lifecycle. You’ve got to maintain it because there’s a user base for it. They’re expecting it. Things will fall apart if it’s no longer there. That’s not true of everything. There is, and always will be, a continuing role for tactical analysis. Let’s imagine in the most obvious use case, if your company is considering buying another organization, there is lots of data work to do to help understand whether that’s a good thing, help with the process of merger, bring the organizations together. That’s not a product. That’s a one-off piece of data analysis that can add lots of tangible value.
With products, you have an ongoing service agreement. So if you create an algorithm, the most disruptive product that was built whilst I was at JLR from a data product perspective, specified the specification of vehicles that went to retailers around the world. And there’s more ways to permute a JLR vehicle than there are grains of sand! An obscene statistic, but it’s true. You can see color, paint color, sound system, stereo, type of roof, all of this stuff. So actually an algorithm to set based on what we understand on market demand, what is the right stock to hold in each retailer — an algorithm did that.
And it’s a product because there’s an ongoing obligation. If you took that away, all of the processes the organization had prior to that to manage vehicle specification and supply chain were no longer in place and things would fall apart. So a product needs some form of ongoing service to the organization.
And I guess the main question you were trying to ask was around and I’ve just answered all the peripheral ones, was around value. How do you measure value? And it is down to what matters to the organization, but it needs to be quantifiable and measurable, and it needs to be an outcome, not an input or an output. So, what you should be measuring from products are not the number of reports you’ve developed, even the number of users, that’s all outputs.
It should be outcomes like what’s the improvement to profit of the organization, to revenue per vehicle? What’s the improvement to customer satisfaction? Are you tightening your supply chain? In which case, how many fewer days are there in your supply chain? They should be measurable outcomes. And without measurable outcomes, I don’t think you have a product. Which we can go off on, but I’m going to pause there.
Anthony Cosgrove: So when you’re talking to the intrapreneurs trying to try this change, are you thinking in terms of product? Are you thinking tactical analysis? Or are you saying, “No, no, let’s start with the value. Let’s start with the value that we’re looking to achieve and kind of work back from there.” But how does that work?
Benny Benford: Start with the outcome. Don’t worry too much about what the outcome measure is. You can always change your outcome measure. So in some organizations, I’ve seen people waste time debating what outcomes matter most to the organization. Just get started — you can always change your value measures. So choose an outcome. It might be profit — it might be a charity. There are some really good charities in this place. There was a charity I love called the Robin Hood Foundation, which has a tagline, “Fight poverty like a New Yorker.” And they fight poverty with algorithms.
And so their measure of an outcome is alleviation of poverty. So, it’s a measurable thing that your organization goes, “Yes, our purpose is to do X. Our purpose is to make profit, to improve customers’ lives, to improve poverty in these areas.” It’s a measurable outcome. And then once you’ve got that, and you’ve got an idea about it that various people agree on that it’s likely that will improve your outcomes, you then — and this is Caroline’s point — work relentlessly on iterations and MVPs.
And the reason for this, and this is very much similar to the entrepreneur’s problem. Entrepreneurs come up with a business problem that they want to solve for the market. A bad entrepreneur then goes and spends four years building a product and then goes to the market and goes, “Is this what you want?”
A good entrepreneur goes and starts testing with customers. Does this help? Oh no, I didn’t quite understand the problem. And then you deviate, you change your plan. You should be doing the same as an intrapreneur with your products and aiming each sprint to go, “How can I improve that outcome?”
And clearly at the beginning, it’s not an automated algorithm, but at the beginning you should have a customer that’s making decisions in your organization and you should be stretching yourself to every sprint and go, “Can I provide more information to them such that at the end of this sprint, they understand their decision making environment better and I’m able to make a better decision?” And then the reason why that’s so important is not just early value delivery, but you learn. Because if you give information to your stakeholder that, for example, your stakeholder is trying to reduce stock levels in an organization and you give them some information that you all think will help improve stock management. And it turns out your stakeholder does nothing with it. You’ve learned something. You’ve learned that actually this information doesn’t help them to make better decisions. So you then have a discussion about what would and wouldn’t help. And then you actually prove what problem you should be solving quantitatively before you go too far in product development and develop the wrong thing.
Anthony Cosgrove: Perfect. And we spoke about this earlier, but it feels like there’s something inherent in there as well around problem selection. So certain problems are certainly solvable. But the opportunity cost of solving that problem versus solving something else just makes it not worthwhile.
Can you share a little bit more around your thinking into how to work through that? Particularly if you’re iterating through a problem, right? You might think it’s a $100 million problem to solve. And as you get into it and you iterate through and learn all the lessons you just said, yeah, it turns out it’s a $10 million problem or a $1 million problem.
Benny Benford: So, yeah, so I think you’re right. I think one of the reasons people struggle with strategy right now is technology has made so much possible. There’s an abundance of opportunity and if you’re not focused as an organization, you look at things and say, “Well, that’s a good opportunity. Let’s do it.” And you end up working on lots of things that are good opportunities, but then might not be the best opportunity.
If you have a process up front to value how much do we think this problem’s worth for us to solve and how hard do we think it will be to solve. You’ve then got the ability to size. Well, you know, we’ve said this is a £10 million opportunity, a £100 million opportunity. Let’s work on the £100 million one, or we think this is a problem that’s going to alleviate poverty for a hundred people, versus this one will alleviate poverty for a thousand people.
Or we think this one’s going to help improve stock control in one country, versus this in 10 countries. So you work out the level of impact your problem is going to have. And then you focus on the ones that are going to have the largest impact on what you care about. And in order to do that you want to move fast. You don’t want to move into this world of needing to do a huge amount of upfront work. So you really just need three people to agree on a value estimation. Someone from the business area that’s going to make a decision; someone from the data team that’s responsible for the data product; and someone independent to hold you both to account. Otherwise those two people could just decide to work on a hobby project.
So someone independent from somewhere like an audit team, finance, or a transformation team to go, “Yes, your assumptions are reasonable.” And then you all admit to humbly go into a learning environment to go, “Based on what we know now, let’s see what we know in two sprints’ time, in four sprints’ time. Do we know any more? Do we still think it’s a £100 million problem? Do we still think it’s relatively easy to work on and be ready to kill projects quickly if it turns out that value isn’t as high as you thought, or if the complexity is much higher than you realize?” So you’re regularly reassessing your portfolio to work on the things that have the highest likelihood of delivering impact.
Anthony Cosgrove: Awesome. And this next question might just be a bit moot because like the whole business of Datent is really you sharing critical lessons that you’ve learned, and you’ve already shared some of those with us just in the last 10 minutes. But if you were to kind of give advice to your younger self, like starting out in data transformation, data product management — what would be the critical things that you would share?
Benny Benford: Build a network.
Anthony Cosgrove: Huh. That’s not what I was expecting!
Benny Benford: Don’t try and solve everything. Data people like solving problems. It’s one of the things that attracts people to the field. As a result, one of the most common problems I see in data people is they love reinventing the wheel.
So find out what’s already been solved, use it, and work on what’s unique to your organization, build a network, learn from as many people as you can, and get started. And don’t reinvent the wheel.
Anthony Cosgrove: Awesome. And for those listening who are maybe in that position and kind of starting out, where have you found the best places to do that learning, to build that network?
Benny Benford: This sounds self-serving. This is why I’ve created Datent — because I didn’t think there was a good network. And to explain what I mean by that, as a CDO, I was flattered with how many networks I was invited to join that were beneficial. Almost exclusively, all of them said, “This is a network for CDOs.”
To which my response was, “But I’m not the transformation leader. I have about 15 to 20 transformation leaders. How do they benefit and share exchange knowledge?”
[They said,] “Oh, that’s someone else’s problem.” So I think trying to change a business is such a big problem. You need a community that involves much more than the top role. I hate the hierarchy concept.
So that said, there are obviously good networks. Go to events like Big Data London. There are free events. Whatever your largest tech vendors are, get involved in their communities, whether you’re a Snowflake house, an AWS house, you know, you’ve got the AWS conferences on at the moment.
Get involved in communities and learn from the outside world. Yeah, there’s enough problems to work on to not waste your time solving problems that have already been solved.
Anthony Cosgrove: Awesome. Great, great advice and quite different to stuff we’ve heard before which is very, very cool. So, within Datent, you are at a bit of a vanguard in terms of intrapreneurs trying to solve hopefully new problems and not reinvent the wheel.
Technology continues to grind forward — cloud, gen AI — focusing more and more on a very broad church of data consumers using data in their everyday roles, whereas 10 years ago, they wouldn’t necessarily have done that or really had any affinity for it. So, it’s sort of in an interesting position at the moment. You’re in a very interesting position to actually see what’s coming next. So what do you think that’s going to look like? Where are data products heading? Where is data generally heading?
Benny Benford: So from a tech space, don’t worry about tech. Look at some of the most inspirational stories that have hung around for a long time in the data space, the Amazon recommendation algorithm, the Netflix algorithm — both of those stories are 20 years old, and they’ve probably achieved a lot more for their organizations than most companies, than the highest impact data product is achieving today, and they achieved it with the technology that was available 20 years ago.
So don’t worry too much about tech to start with. First, prove that you’ve identified the problems that are worth solving for your organization. And when you’ve got as far as you can get on solving those problems and driving impact in those areas with your existing tech stack, then you understand the problem you’re solving well enough to look at new tech. And then you bring in new tech. Technology is a distraction. So that’s the first thing I would say. And it’s reasonably common advice. I can’t remember who it was when I picked up data governance. There was a really good data governance book I read that said, in almost every chapter in the book, “Do not buy data governance technology until you’ve had a data governance program for at least 12 months,” because you just don’t understand what you’re doing enough.
So, first get started with your problem. Technology is a distraction and then look at the technology. And then you’ll be much better placed to understand what technology you need. But that doesn’t entirely answer your question of what’s coming next. I think — I hope, that the data world starts to become more professionalized, because [at the moment] it’s not. The best analogy I can come up with is that I think the data profession now is very similar to where software development was in the late nineties. So, in the late nineties, there weren’t common standards for your software development life cycle, for how you did code management, all of those great things. If you’re standing up a software team today, you wouldn’t invent any of your processes around code management from scratch. You’d copy from a whole range of different frameworks that are available out there. A lot of data teams today start from scratch. So what I hope is happening in the next five to ten years — and data will play a part in — is producing frameworks so that there are common ways of doing data that you then modify to your organization. Then data becomes professionalized, because at the moment it’s not a profession.
Anthony Cosgrove: And do you see those two things as linked? Because I think in your first point, you’re talking about not getting swept up in technology and kind of almost data being… I don’t want to say necessarily its own distinct discipline, but just being very tightly coupled with business value as opposed to being tightly coupled with a technology.
And you mentioned earlier, you know, there’s so much opportunity now because there is so much technology that can be quite challenging if you don’t have discipline around “Where’s my value? And how would I use this to actually deliver the value that is most important to me right now?” So that’s sort of being your first point.
And the second point being this sort of professionalization of the industry and kind of industry standards and starting to leverage the increasingly large foundation of work that’s already been done, things that haven’t been reinvented. So do you see those two things as linked?
Benny Benford: I think they are. I think the data profession overly defines itself around technology and therefore struggles to stabilize because it’s always chasing its tail to go, “Oh, my God, a new thing’s arrived. Everything’s changed.” And other people don’t do that. And other areas are being impacted by technology as much. I mean, look at a warehouse.
Warehouses are being largely automated, but that doesn’t mean that the profession of warehouse management and logistics has not defined itself. It has defined itself. It’s a well-established profession. There are professional bodies for logistics. Or take supply chain in a similar way. Graph technology is going to completely revolutionize supply chain and enable digital twins to happen in a different way.
But they’re not sat there going, “We’re a technology profession, and we can’t come up with standards because there’s more technology coming in.” They’re going, “No, we’re a supply chain profession.” So I think data needs to stay. And this is, I very, very strongly believe, CDOs and data teams cannot report to technology leads for this reason.
It’s confusing the problem. It’s really, really confusing the problem and getting it, getting it hooked up — “Data is tech.” Data is completely different to tech. Everyone is dependent on technology, but no one else has made the mistake of thinking they are a technology area as a result.
Anthony Cosgrove: That’s super interesting. Here’s one very quick final question because this came up in the second podcast with Patrick.
So, if data is not reporting into technology, where would you see it reporting into?
Benny Benford: At the moment for most organizations, my preference would be Chief Transformation Officer first, because the way in which data works needs to change radically and it needs to link to the organization’s priorities.
So if there is a Chief Transformation Officer, it’s ideal. Second would be either the CFO or COO because they actually have an enterprise-wide remit and most roles don’t. So when you end up with the CDO reporting to a Chief Commercial Officer, for instance, or Chief Customer Officers, as happens in some instances… it’s a weird thing because their remit is functional, whereas a CDO’s remit should be enterprise-wide. So I’d say Chief Transformation Officer, followed by CFO or COO.
Anthony Cosgrove: Which chimes pretty nicely with what Patrick said as well. And having observed different reporting lines leading to different outcomes, he landed on the CFO being a great change agent for data. And also I think looking at it in the way that you described as a value driver, and “What can I do with this asset in order to change kind of valuable outcomes for our organization?”
Well, look, we’re going to have to stop there. But this has been jam-packed and super interesting. Thank you so much for joining us. For those who are listening and are now hopefully super intrigued by data intrapreneurship, visit datent.com and you can learn more there. And if you’re interested in setting up a commercial or enterprise data marketplace, check out harbrdata.com. Thank you for listening.
Benny Benford: Thanks for inviting me.
Data commerce insights. Delivered monthly.
© Harbr Group Limited 2024. All rights reserved. Privacy policy | Cookies policy | Terms
