How to Merge Startup Agility with Corporate Stability within your Product Team

Small companies want to be big, and big companies want to act small. I used to work for HSBC in their Global Digital Team, and every month we would have a curated set of fintech startups who would come along, pitch to us what they have. And there kind of two conversations used to come out of that. We and HSBC, with our 25- to 30-year-old mainframe systems—they were built, they would maybe do a release maybe four times a year. That was what their original purpose was. There was us trying to get them to release six times a day.
A few years on, we'd be looking at these guys going, "God, they're quick. Why can't we have their tech? Why can't our stuff look really shiny like their stuff did? They built it in months." And, of course, they'd be coming out of it, the startups going, "Oh my God, they've got like 30 million customers, they've got billions of pounds, they've got all of the security that we want." So, they would—they wanted to become us, and we wanted to become a bit of them.
So, I want to talk to you about some of the challenges that I've noticed as you go through that, whether you're in a small company now or you're in a bigger organization and you're trying to look at how you kind of create that kind of startup agility within it.
As AJ said, I'm Harry. I'm VP of Products at OVO. We are a Great British—that's England, Scotland, and Wales—energy retailer. That means we buy energy from the wholesale market—we don't produce it—and we sell it to about 4 million customers within the UK. Products at OVO is part of our technology organization, and we, within that remit, we look after product management, UX, UX research, product design, content design, and customer experience across the team.
Staying agile at scale is tricky. As I said earlier, there's a constant pull as you get bigger to standardize processes as well. If you imagine the days when Monzo launched within the UK, they launched a prepaid card—there was no debt on it, no possibility of customers losing money. They only had one product, one card, and they could release really quickly. And the financial authorities, they weren’t—they didn’t really see them as a risk.
Now, if Monzo’s login system worked, everyone was probably pretty much all right. At HSBC, if our system broke for more than about 500 milliseconds, we had to report ourselves to the financial regulator because we had one in four customers—one in four adults within the UK—roughly banked with our systems. It’s a different scale, and it brings different challenges as you get bigger and different processes that you have to adapt for.
Energy is a long game. At a system level, it’s a pretty slow-moving industry. The National Grid runs to a 5-nines reliability—so that’s 99.999% uptime. Anyone else got that in their metrics in this room? Hands up if you're running to 5-nines. Good, good, excellent.
If you think about what we've been through recently, I think you’ll recognize, actually, it held up pretty well. There was lots of talk of blackouts, potentially having to prepare, and actually, the grid held up really well. It’s a very well-managed system, but it’s also a relatively slow system. Smart meters were launched in 2011, and there’s still really only about 60% adoption. So, things don’t move very fast at a system level.
OVO was launched in 2009 with a clear mission of making energy better for people and for the planet. As AJ said, today we have around 4 million customers who trust us with a range of services that we provide to them—everything from home energy management, energy flexibility, EV charging, green energy efficiency, air source heat pumps. And then it’s underpinned by our Plan Zero, which sets us out to be a fully net-zero business by 2035. One million pastries a year—a very nice fact there as well.
If you look at the timeline—we’re 15 years old now, we just celebrated our 15th birthday. Back in 2009, OVO was born in a barn. So, our founder Steven Fitzpatrick had a barn in the Cotswolds, and that was the company’s first office. At that time, there was a very limited range of suppliers within the market, and OVO was setting out to be the disruptor—to be the Monzo to the energy sector.
By 2017, we had 1 million customers—pretty rapid growth through that period, and obviously, that brings a lot of pressures within the organization. Energy is a regulated business. By January 2020, we had 4 million customers.
Then my slides are going to break, but then we have lockdown in March 2020. In January 2020, I hadn’t joined OVO. I was still with HSBC, sitting in our Mexico office, having some chats with some colleagues from our Hong Kong office. They were going, "I think you need to pay attention to this virus thing." And we were going, "It'll be all right. It’ll be like the last time—it’ll probably stay over there, won’t it?" Anyway, of course, we know what happens. So, we go straight into lockdown, we’re at sort of peak road prep, things straight into lockdown.
I leave a nice, solid, 150-year-old global bank that pretty much prints money no matter what’s going on in it. My mum was calling me because this happens, and Ukraine happens, and she’s going, "What have you done?" And I was like, "It’s going to be okay, it’s going to be okay. We’ll see what happens to this."
By August of the next year, we see the gas price per therm—and I’m sure most of you who are energy bill payers know what happens. The gas price per therm—that’s the wholesale unit—having had a very, very stable 10 or 15 years with even very small blips, literally went like that. And that created huge consequences. Between that one year, 31 energy suppliers go bust.
At the time, the barriers to entry to be an energy supplier in the market were relatively low. The regulators changed things around this to try and introduce a bit more stability. But effectively, at the time, you could buy and bet against the price of energy not going up. And effectively, there were a lot of companies that were buying at five and selling at two. And I’m sure you all realize that’s not a sustainable strategy.
We go through all of this—we go through lockdown, we go through the impact that the Ukraine war had. The impact on that—it was very difficult for us. It was very difficult for our customers. It was very difficult for our advisors. They also had to pay energy bills. They’re having to deal with the debt that’s mounting up. They’ve got customers on the phone every day.
We managed to launch what was also then a thing called Charge Anytime a year later, which is, I think, one of the best EV charging tariffs within the UK. It was different because what it allows us to do is to actually look at exactly how much energy you’re putting on your EV and charge you a different rate for charging your car than you would for running electricity into your home. So, you can run your car for about 2.5 to 3p a mile if we control the charge.
That, for us, is key because the digitization of the energy sector is about driving flexibility and creating interconnected systems that allow us to move power around the system. So, a pretty tumultuous year.

How’s It Going?

How’s it going? Well, I’d say it’s going. That would be my answer to this. I think sometimes when you’re doing presentations like this, they’re a bit like our LinkedIn profiles—they’re very shiny. We’re only going to show the good stuff.
Within the limit of our media team, who’ve crossed out most of my slides, I’m going to try and give you a little bit more about the gritty reality of what’s going on. It’s a journey, and like most journeys in life, it’s not a straight line. Navigating through those, I think one of the key things is—particularly when you get bigger—you get more complexity. You get more systems, you get more people, and, as you heard yesterday if you listened to Kelly, you get more politics.
You get what I would call—we all find a face. There was a lot of, I think, identification and laughter yesterday with, if any of you were here for Andrew’s shoe telephone. It was funny, wasn’t it? It was funny, and we’re laughing at it in our profession because it’s true. But it’s also—it’s not great. You know, "Would you? It’s got to go live tomorrow." And I say, "Fine, okay."
So, it’s tricky. But it’s also—I just want to come back to this—there’s a lot of good happening. There’s a lot of good happening. If you look at the data, the decarbonization of the UK energy industry is actually a phenomenal success story. We’ve just turned off the last coal-fired power station. These stats show a 98.5% reduction in the use of coal. We’re now down to 100% reduction in coal.
Gas has had to come up to compensate for that, but at the same time that gas has come up, we’ve seen wind three-and-a-half times scale over 12 years, and we’ve seen solar 10x come up to meet that. And as you can see from this, wind is now producing more than gas. For the first time, we’re now seeing non-carbon-free power around 60% in the grid, which is huge—huge.
So, the direction of travel is really, really positive. Now, I’m not talking about where we are in terms of weather systems, climate, and heating. But if you look at the metrics and unit economics in the system, there’s a lot to feel really good about.
At the heart of all of this is digitization. Smartphones and smart meters allow us to connect to our energy, and the key for the decarbonization of the grid, the system, and our homes is about actually connecting to energy. Now, those of you who maybe have the traditional meter where you have to phone in or you give a read every month—you know there isn’t a lot to get from that. You just have to do that read, etc.
But what we really have the opportunity with smart meters and the shift towards decarbonizing our transport, generating and storing power within our home, is to move this thing called half-hourly settlement, which is one of the biggest shifts we’ve seen coming within the industry. Historically, what we would do is buy a bunch of energy from the wholesale market. That would be based on our forecast of how much you’re going to use, how much our customers are going to use, and then we would sell that energy onto our customers. We’d hope we got the forecast right. It’s quite a slow cycle time.
What half-hourly settlement allows us to do is, within a half-hourly period—so 48 times a day—is to buy and sell energy straight through. That also creates 48 price points within the day. The interesting thing, if you look at the way energy flows, is the peak energy demand is between 4:00 and 7:00 p.m. That’s pretty obvious—we all come home, we want to heat our dinners, heat the homes, have a shower, etc. But that’s also peak cost and peak carbon production.
So, as we decarbonize, what we’re trying to do is help our customers connect with their energy, make their homes more efficient, and shift the load into a more balanced profile across the day. That gives us the carbon reduction we need. So, introducing flexibility and connectedness at a really significant scale within the system.

Telling the Story of a Connected Home

What we’re doing within product and design is asking: how do we tell this story? How do we tell the story of what a connected home looks like? How do we connect the home to the grid and to the wider environmental challenges we have? The reason why I think we find that challenging is that most of our customers want their energy to work. Their relationship with their energy supplier is like their relationship with their light switch. They don’t really think about it.
Is the light switch going to go on? If you come back to that 5-nines reliability, it’s just going to work. So, the big challenge we have is to kind of wake up our customers, connect them to their energy, and connect them to a value proposition that resonates with them. Inevitably, it’s going to be grounded in price. But we also know, although price is a primary driver for our customers, they’re also really concerned about climate. They see what’s happening, they read the news, and they want to understand: what can they do? Sometimes, they feel really powerless—I know probably people in this room do too—what can we actually do about what’s going on?

Tips and Insights for Scaling

I want to share with you a few tips—things we do, my experience of this, and how you manage through that process at scale. We’re not a small startup with one product. We have a lot of customers, and when you have a lot of variability in your customer type, that starts to come in.
So, you need to have a plan. We do—we have a plan. One of the things we built is a vision type of what we want our future experience to look like for our customers. We talk about that a lot within the organization.
Now, I’m sure you’ll all realize this, but humans are reality-making machines. Our reality is largely subjective. There’s very little objectivity in how we perceive what’s going on around us. The reality is people listen to what they want to listen to—I know those of you in the design profession know this all too well, particularly the researchers among you. They see what they want to see and hear what they want to hear.
You have to keep telling them where you’re trying to go. So, we use this internally to really try and drive the conversation as much as we can about where we are going within the organization. Where are we going in terms of our proposition? How do we want it to mature? This is where we’re going, and any opportunity we get within the wider company is to say, "This is where we are now. Remember, this is where we’re trying to get to, and this is where it will go forward."
The actual reality of this, in terms of what we produced, is it will never look like this—I hope it never looks like this. I mean, bits of it I quite like. But this was never meant to be a finished prototype. This is a vision type, and that gives us a lot of flexibility. It also gives us a lot of efficiency because we don’t have to test this, we don’t have to see how it works. But it allows us to, as you can see with things like the house in here—yeah, that’s not a typical customer’s home—but it’s a very lightweight and relatively inexpensive way of creating a shared understanding about what’s going on within the organization. What do we think is going on for our customers? And where are we trying to move our teams and our investment?

Investing in Systems

The other thing we’ve learned to do is invest in our systems. We have a design system because it saves us money, and it saves us time. That’s it. That’s why it gets investment. What we’ve learned to do is not to rely on a design system as being a pet project—someone’s baby in their 120% time. It is run like a product, and it has a P&L. That means we understand how much it saves us, and we understand how much it costs. It saves us more than it costs, and that warrants the investment and ringfenced capacity around it.
We use the same principles with other systems because if we don’t treat them like products, we don’t get the efficiency we need. Now, again, when you’re small, and you’ve got maybe two, three, four, five teams, it’s fine. But when you start having 50 teams, those efficiency gains start to add up to a lot. You have to work very hard to, one, get adoption of your systems, but also to actually keep the systems working at scale to drive the efficiencies you want.
Now, it also looks really pretty, so we like that as well. But when we come out and talk about systems and talk about the value of them, again, the lead is always within the commercial context primarily. We get the knock-on effects as well. We also get aspects of things like accessibility.
What we’re trying to do overall is reduce the cost of front-end code deployment. The cheaper it is to deploy, the more often we can, and the more we can test and learn. I’m sure you all understand this—if it costs £10 to release something, that’s going to be a big investment. That’s probably going to be your one shot.
What we’re trying to do is get it down to costing £2 to release something so we can try five versions and throw away the four that don’t work without worrying about it.

Speed and Scale Must Come Together

The other thing as well is that speed and scale have to come together. You can’t have one or the other—you can, but it won’t get you where you need to be. We’ve had to work a lot on understanding how to break down some of the paradigms you might hear within your business. “You can’t be at scale and move fast” is a common one.
In order to do this, I’d say the three things we probably focus hardest on are modularization, componentizing our system. I used to work for a food-tech delivery business in the UK, and when I got there, we were looking at how we could optimize the flow, conversion rate optimization, checkout optimization. The basket they put in was awful. I was like, "Wow, this is really easy—look, I’ve just started. Quick win here. I’m going to get loads of brownie points."
I said, “You can’t touch the basket.”
I asked, “Why not?”
“Well, it’s pretty much stapled to the backend. The team who built it don’t work here anymore. No one knows how it works. If it goes down, we’re screwed.”
So there I was, thinking I was going to look like a real hero really quickly. “Look at all the money we’re going to make!” But again, it’s a good example of when you’re starting small and moving fast, you’ve often got to cut corners. You’re going to create technical debt, and you’re not going to think about the architecture of your system.
Unfortunately, what I’ve found is that when you start to scale, those architectural shortcuts—the lack of thinking about the separation of concerns within your system, be it within your design systems, within your frontend, your relationship in terms of your intermediate API layer, and your backend systems—they really start to bite.

A Key Lesson: Data Model > API > UI

One of the biggest changes that I’ve seen us go through, taking forward from my previous work, is this: a lot of teams work UI > UX, then think about the API they need, and then wonder where the data is going to come from. You’ve got to do it the other way around.
We teach our product managers to think data model > API > UI. Data model > API > UI. And the reason being is, if you don’t understand the data model—and most software is just moving data around—and you don’t understand the services, then what you’ll build is a service constrained to the user experience and the flow you’ve got.
What happens when someone comes along and wants to plug it into AI, for example? They realize it doesn’t work. If you’re not careful about how you manage that, and you have your service catalog internally, what do they do? They build another one. It pretty much does the same as the first one, but now they’ve duplicated it.
At HSBC, we did a whole program around this. If you think about a banking app, one of the things you need to do is get a list of the customer’s cards. Guess how many APIs we had in the UK for getting a list of customers’ cards? We had 14—14 APIs doing exactly the same thing.
So, we had to rationalize that all down. Start small, create your feedback loops, and foster more of an iterative development model—agile as it is—because it reduces risk. You don’t know if software is going to work until you put it live, so you want to put it live in small chunks.

Iterative Development and Portfolio Management

Waterfall is great if you know what you’re doing. If you’re going through a repeated, known process, you can really standardize every step of it. But most of the software I think probably most of us work on is trying to see what’s going to work and looking for the wins.
The hardest thing I’ve found is throwing away the stuff that doesn’t work. You’ve probably all heard the cliché: “Fall in love with the problem, not the product.” I think it’s hard—you’ve got to fall in love with the product a little bit. But actually, taking the time to pare things down is important.
So, we think about our portfolio. We think about what we’re doing in terms of exploratory work, what we’re doing in terms of scaling—seeing how it works at scale—what we’re doing in sustaining, and what we’re doing in demise. If we don’t see an investment in the demise side, that’s a red flag for us. We go, “Hold on—if we’re not taking stuff out, we’re not going to create the funding to put stuff in at the beginning.”

Metrics and Business Alignment

Kelly yesterday talked about learning the language of business. I think the lingua franca—the common language of business—is metrics. You need to understand your KPIs, and they’re not just about development metrics.
We use a common set of metrics across all of our teams. It’s about understanding how the business works and how you can take that within your organization. One of the things we did within our plan at OVO is to have a very simple algorithm that shows how we make money as a business. Because a business that doesn’t make money—and you go back to the timeline I showed—is not a sustainable business. If you can’t understand that, it doesn’t work.
We understand how we make money in terms of the energy we sell, the debt we hold against it, the cost to serve the customer, etc. We use that a lot in our conversations with all parts of the organization. We try and break down the idea that there’s a bunch of folks called “the business” who are the only ones who can talk about that.
I don’t know if any of you have “the business” as a term within your organization, where they’re separate from everyone else, but we try and break those barriers down as much as we can.

Watch Out for Traps

And the last thing I’ll just talk about: watch out for the traps as you scale. I’m going to try and explain this concept within the renewable energy vibe, as a metaphor.
Biomass has emerged a lot, and one of the things in that chart along the bottom was biomass—that’s gone up. Now, if you think about fossil fuels, pretty much all of the energy we have on our planet comes from the sun. The sun’s energy comes in, creates heat, and that heat is not evenly balanced. It creates different air pressure, which creates wind. So, most of it is about moving photons around within the system.
Fossil fuel—coal, oil, gas—is biomass that’s taken from putting a photon through a leaf. The leaf converts it, turns it into carbon, builds a tree, the tree dies, and 500 million years later, we pull it out of the ground and burn it. Biomass is the same thing, but with trees we cut down yesterday.
Now, biomass looks like a really good idea because oil and gas are dirty—we don’t want those. Trees are green and lovely—let’s burn them instead. It makes a lot of sense if your business is burning stuff to produce electricity—you want more of that. Just like the people who say that hydrogen works in cars and home heating tend to be the same people who like piping combustible stuff down pipes and burning it in a car or a boiler.
But it’s a delusion. It doesn’t work. It’s the most grossly inefficient system when you compare it to taking a wind turbine at 40% efficiency that takes sun heat, air, moves it, puts it through a turbine, and creates an electron. What we’re trying to do at OVO is to understand: are we going through a biomass process—somebody over here has an idea, goes through 15 different hops, gets filtered out—or are we trying to reduce the cycle time and the hops and start to have a wind model?
The thing I think you’ll find within business is that there are people who like burning stuff to produce electrons, and they get defensive when people say you could just have a wind farm. Similarly, within organizations, you’ll meet that resistance—people who say you need to be biomass. You have to constantly challenge that.

Optimism Wins

When you’re challenging the status quo, remember this: pessimists always sound smart in the short term, but in the long term, it’s the optimists who win. This narrative around renewables—called “blackout”—is a good example. Has anyone here got an EV? Electric vehicle? Put your hand up. Great. Has it blown up recently? Just spontaneously caught fire? Collapsed a car park? Anything like that? No.
They don’t blow up. In fact, they explode less than internal combustion engine cars. They don’t collapse car parks. And the reality is, we’re now seeing the battery pack on your EV will outlast the chassis. But the fear, uncertainty, and doubt industry will tell you they’re just going to be awful.
So, I’ll just finish on this: have a plan. Invest in your systems. Think about speed and scale. Measure the same things. If you can’t measure the same things, you can’t talk commonly across the business. Watch out for the traps—they’re coming. You need to see them and counteract them. And above all: stay optimistic.
Thank you.

Post-Talk Q&A Highlights

Audience Question: “Can you give some examples of how you make profit and loss of design systems?”
One of the things we look at is the cost of code deployment. Even a very simple thing—if you take a button or a form element and think about how many variations you’ve got going throughout the system and how many states each has, you multiply that by the number of teams working on it. Then you calculate: what would be the actual cost if everyone was designing their own? What we’re really looking at is a reduction in time.
We also use it to reduce the cycle time to production. For example, if you’re launching something new and it entirely uses the design system, we know it’s accessible, so you don’t have to go through an accessibility loop. We don’t have to go through extra checks.
Audience Follow-Up: “Does your design system have a product manager?”
It does, but in a very light-touch way. There is a PM that looks over a number of global components, frameworks, and systems, but they’re not dedicated to the design system. It’s largely left to engineering and UX practitioners to develop it themselves.
Audience Question: “How do you find investors for a startup to launch and scale up?”
This is slightly off-topic, but I think it’s relevant. Talk to a lot of people. The one thing I would say, if anyone is thinking about launching an idea, is this: if the first thing you want to do is pay for an NDA—don’t bother. It’s a complete waste of your money and everyone’s time. The biggest problem I’ve found in launching a startup is not people stealing your idea—it’s people caring about your idea. So, don’t put any barriers in the way. Don’t spend money on lawyers. Go out and talk to a lot of people. And generally, I’ve found that if you ask for help, most people want to help you.
Audience Question: “A bit of a challenge to your data-API-UI approach—could you explain how it works in practice?”
I totally love this challenge. Think about it like Lego. If you have a really robust, well-documented service catalog of all your services within the system, you can move so fast. What you’re saying is, “I can take that from here, that from there, pull it together, and test this with my customers.”
But if it doesn’t exist, you’ve got to go and write the compute, put it over there, and integrate it. So, actually, I think it speeds up our time to production. It reduces the cycle time, which reduces the cost. When the cost comes down, it allows us to try more things. But you have to understand the data that sits underneath it.
In today’s world, particularly with the changes we’ve seen around AI and large language models, having a good solid data foundation becomes even more important. Customers don’t want to think—that’s a fairly standard principle in design. The fewer things we can make them think about, the better. And that relies on having solid data.
The nice thing about APIs is, if anyone here doesn’t know much about APIs and is thinking, “Isn’t that an engineering thing?” you can go and have a conversation with ChatGPT and say, “I’d like to build a RESTful API that will do X, Y, and Z. Can you help me design it?” It will pretty much build the entire thing for you. Then you can sit there and ask it loads of awkward questions: “What about this? What about that? How do I do this?” It will just keep going for you.
Also, there are systems like Gravitee, which is an API management platform, that have really simple drag-and-drop interfaces. You can effectively say, “This is the data I need in this system for it to work,” and it will pretty much build the whole thing for you.