Upgrade to ChromeUpgrade to FirefoxUpgrade to Internet ExplorerUpgrade to Safari

tips & insights

Insider knowledge

We’ve plenty of insight and experience to share. Take advantage of it here, and get inspiration for your next move.

Josh meets… Richard Pugh, chief data scientist at Mango Solutions

Posted: 17 October 2017

How has 2017 been for the business?

Great. Over the last two and a half years, we’ve quadrupled our data science business, but this year has seen our strongest growth so far. I’ve also been in the boardrooms of half a dozen FTSE 250 companies advising on data science which has been great. While we’ve got a lot of good projects at business level, having that more strategic level is nice.

I read that Mango had a hand in creating the world’s best coffee? This can’t be true, can it?

This is true! The coffee project we did was with Kraft – one of the challenges they had was to make a consistent coffee blend. Now, if you think about making coffee it’s a varied process —raw ingredients, different procedures, different machines, different people— but what you want in your coffee is consistency because that’s why you buy that brand, you like that taste and it tastes how it did last week. If it tasted different every time, you wouldn’t stick with that brand. Kraft identified that it was an analytics challenge – an optimisation. So we helped them optimise how their coffee tastes: minimising the difference between how it tastes each time you drink it and, at the same time, reducing cost of the beans.

This is actually a really great example and I use it a lot to open people’s minds to what analytics can do. I actually wrote a blog post recently on the fact that people tend to constrain their thinking of what data science can do. For example, people think data science is only about forecasting or customer segmentation, which means they can miss out on solving other really important business challenges. Use cases like that at Kraft are so important because they show what can be achieved with analytics.

It’s a good way of showing that you need to aim high when thinking about the challenges your business has. Then thinking about how the data can solve it, as opposed to thinking: “here’s the data what can we do with it”. We developed some workshops in conjunction with M&S to help people understand the language of analytics that in turn helps them explore some of the analytics problems they have. It’s very much about teaching a common language around analytics. For example, how do you use the word ‘model’ in a business context to qualify whether the challenge they currently have is solvable by analytics.

With that in mind then do you find that the ‘education piece’ is one of the biggest challenges when offering a consultancy service based around analytics?

I think so. The problem with analytics is that it’s an incredibly broad term, so if you knocked on the door of any company in the world and asked, “do you do analytics?” they’d answer “yes” because ‘analytics’ can cover anything from “we have a spreadsheet/dashboard” to “we’re using Machine Learning to drive product selection”. It’s important to educate people around the different types of analytics and where data science sits within that.

It also suffers from the marketing input, in the same way the big data world has suffered. In big data, the idea is that you buy a big Hadoop stack, you chuck all your data in, and you throw an algorithm at it and money pours out the other side. Obviously, that’s a fallacy. So, building on that example of big data I think education is incredibly important because, in the future, people will start to invest in analytics, or data science, not really understanding it, and just expect the results to just come out the other end.

A lot of our work lately has been around advising people in C-level executive roles on “what does analytics actually mean?” Do you need a team, do you need to hire a data scientist etc. Sadly, I know a lot of friends hired as chief data scientists or lead data scientists for companies who have left after a year because the company wasn’t really ready for them – maybe the infrastructure wasn’t set up, or the business wasn’t ready for change.

Do you think that there is a lack of trust from execs towards data, especially when they have to make business critical decisions based on the advice data can offer?

I think so. A lot of the data science ‘advice’ that goes into that executive level comes from technology companies selling big data products, or very large companies looking for ‘land and expand’ opportunities. Often the advice these companies provide can often be bias towards their own ends, whether that is selling a product or doing something way too big – so there’s a real lack of real-world, pragmatic advice out there for people looking to use analytics in their business.

I think there’s also a dumbing down around it as well, where companies go pitching analytics but actually under the hood it’s just BI, and for me that’s where there is a big missed opportunity.

However, the opportunity is there today and a data strategy with analytics at its heart can really add significant value to an organisation. My fear is that companies will invest significantly in data science but, because of the advice they get, they won’t reap the rewards. If that happens, we could be in a situation where, in a few years’ time, we are in a place where the phrase ’data science’ is dead because it was never able to deliver the value promised – I think that’s a real risk.

Moving back to the business, what has been the highlight of 2017 from your point of view?

There have been so many. I think the growth has been a big highlight – the demand for experienced data scientists and real-world advice has meant we’re growing at an aggressive pace, and I don’t see that slowing anytime soon. It’s been great to see our data science and data engineering teams grow so quickly, both in London and in the South West.

But the highlights for me are always the customer stories that really bring analytics to light, such as the Kraft example I mentioned earlier. For example, we worked with a finance company this year who were looking to build a ‘Netflix Style’ next best action recommender engine for their staff. So when their customer representatives interact with customers they get recommendations for upsell or cross-sell opportunities. We went from 0 to tools in the hands of users within 3 months, and the impact has been fantastic, increasing overall upsell and cross-sell by over 20%.

We actually built the whole thing in R and Shiny and it’s a great example of R in production. We’ve done a lot of R in production recently and have been really pleased with how well R has scaled. In fact, it’s been great to see the continuing growth of technologies we’re heavily involved in such as R and Python becoming established analytic tools and be able to hold their own in production environments.

The other thing is the more strategic stuff, advising the execs at large companies around the value of analytics and the steps on the data-driven journey. It’s been particularly interesting to see companies wanting to use data proactively from sectors who are less ‘analytically mature’ than traditional industries such as finance, insurance and life sciences. These are companies who have never done analytics at all beyond spreadsheets, or dashboards and advising CIOs and CEOs on what can be done with analytics and so on has been great and a really creative process.

Would you say that’s one of the great advances of Data Science over the last few years? That it’s breaking into new industries as we hear more and more success stories, and it’s also allowing startups get ahead since they’re making more reliable decisions?

I think that’s true. Of course, the use of data to drive decision making is nothing new. I was a statistician back in the late 90s, where my role was really reactive. These days, the data science movement has made data-driven decision making it way more strategic. As analysts we’re empowered to get closer to the business and really make an impact which is exciting.

But I think this is quite a disruptive thing for an established business where changing behaviours might be difficult. I think that’s why startups are really being able to get going quickly, because they can design their businesses around data-driven principles from day one. We’re doing a lot of work helping traditional companies evolve with data too, but it’s more of a cultural change than a technology one, and that’s where challenges can arise.

You mentioned the shift in the meaning of stats and data science over the years, so what drove you towards a career in the data industry?

I always enjoyed maths so I knew I wanted to something in that area. When I did my A Level, I found that Statistics was something that came naturally. Then I did a Maths & Stats degree at Bath. When I got my first job I thought “I’m going to change the world using data”, but I realised the role of a statistician was more reactive back in 1998.The idea that I was going to go up to the CEO and say “what’s your biggest problem? I’m going to solve it with analytics” – that was never going to happen. I was a bit disappointed with that, so I moved around a few jobs trying to find one where I could drive decisions using data.

The other thing is I realised was that I was quite an odd statistician in that I wasn’t just doing analysis. What I would do is write the code that would automate the analysis I was doing. That meant I automated myself out of anything interesting to do. One of the first things I’d do in my job was set up analytics routines, but that narrowed my job down to running the same code everyday with different parameters and so quite quickly I was thinking ‘what’s my next challenge?’ but it felt like people were happy with me just running that script for years to come.

I was trying to get more involved in communicating with the business and I felt I was more of a coder than an analyst. And then, of course, the data science movement came along, and we were all “this is driving business!”, ‘this does make decisions!’, ‘this does support growth!’, so that just felt natural to me, that’s where I wanted to be. I think I was a bit of an oddball at the time, but it makes sense now!

So, I would assume then that you are not surprised by all the new things we are now solving with data, given that always been your natural belief?

Absolutely. My fundamental belief is that every decision can be made using data. I absolutely believe that. Plus, I like that approach because that ethos works well for me. You start out with everything being attainable and then you work backwards. We’ve use this approach a lot with customers where we assume any decision can be made using data and work backwards from there – it’s nice because it doesn’t take anything off the table.

So what is it about data you still enjoy 20 years on?

I enjoy solving challenges. I enjoy supporting people on how they can make better decisions with data and I like the creativity side of it. I also like solving real-life challenges with data and analytics and use those stories to inspire further thinking. When I say we help BP predict when their pipeline is going to crack, or that we help Kraft make better coffee or an accounting company understand the optimum amount of bonus to pay their staff, it opens people’s eyes to what can be achieved with data.

I guess it’s when someone asks you ‘how?’ that it gets more complicated?

True, but that’s where having a great team of data scientists is really useful. But conceptually, analytics is straightforward – you’re either describing what happened with data, exploring structures hidden in the data or using models to understand how best to behave. Whilst prescriptive is the most complex aspect of analytics, it also speaks most directly to what the business is trying to achieve. So when someone says “how” I try to use language that drives a good business conversation – if I’m speaking to a CFO and I start using words like ‘Bayesian’ I know I’ve gone wrong somewhere!

So you’ve mentioned your journey through your analytics career so far, who were the key influencers or inspirations for you throughout that career?

I suppose my main inspiration came from my University lecturers like Professor Chris Jennison, Professor David Draper and Dr Chris Chatfield. They gave me a real grounding in analytics without which I couldn’t do my job.

Beyond that, I’m mostly inspired by the people actually doing data science every day, and many of them are our customers. For example, I’ve been really impressed with the work Harry Powell and his team has done at Barclays, or the way Ben Downe has embedded analytics into BCA’s business.

I always try to look at someone doing analytics in real life as opposed to some of the media spin that’s going on. There’s a huge marketing machine around data science right now, so cutting through the noise and hearing about what people are doing on the ground is really important.

Do you share that view that the application of Machine Learning and AI can go too far?

Yes it’s a concern. For example, I was working with a FTSE 250 and the CTO told me that data science is ‘easy’ because you can “just get some algorithms, throw it at a database and then we can make decisions faster”. That’s literally what he said. And of course you can, but if you want to make the ‘right’ decisions, it’s an entirely different thing.

As a statistician there’s always that fear that people will use analytics without understanding the assumptions involved. If you put analytics in the hands of someone who doesn’t understand its true function, they can make the wrong decisions. We are there right now. People are investing loads of money in this technology and it all becomes smoke and mirrors and then they do not get what they were hoping for. That’s why we always start with the question, ‘what are you trying to achieve?’ and work back from there.

As we edge towards 2018 what are the objectives heading into the new year?

I think that’s a difficult one because if you look at Mango, I think we are being driven to increase our data science business very quickly and we will continue to increase our head count around that team. But, a lot of the growth is around three other areas, areas we get asked to do a lot within but have never really tried to structurally grow internally. These are data engineering, data architecture and strategic analytic advice.

The Data Engineering team is smaller than the data science team right now, but we are seeing increased demand there too. Data Architecture is also something we’re being asked about more and more. I think the growth in these teams is driven by companies realising they need a sophisticated IT infrastructure to run a data-driven business, so we’re increasingly being asked to design and create fit for purpose environments and data pipelines.

The Strategic Analytic Advice work has come about mostly in the last 2 years. I think it’s based on our real world data science experience over the last 15 years, and our ability as an independent company to offer pragmatic advice. We’ve just made a couple of strategic hires into that team. We’ve also just signed a strategic partnership with Moorhouse Consulting who are specialists in change management. It’s proving really useful because there’s no use doing analytics for analytics’ sake. You are doing it to try and change your business in some way. They are experts in change, we are experts in analytics so that partnership is working really well. For example, we’ve just executed a successful project together in the UK government. So, I think that there will be a lot of strategic consulting growth in the next year around things like that partnership.

So what do you see as being the positive trends coming through?

So for me, I’m really pleased to see the uptake in adoption of DevOps in analytics, or ‘DataOps’ as it’s increasingly known. Because of our background in engineering we’ve been advocating the use of ‘analytic DevOps’ for years. If you look at the way our data scientists work, they are sitting there working on version control systems against a continuous integration platform, with a test framework, and that’s the way to do data science and I think it’s really good to see that catching on in the industry.

The other thing, is the openness of a business to change around data. One of the things you need to consider when prioritising a data science project is: you can write the best code, get the best algorithms, but at the end of the day if the business person isn’t willing to change behaviour you won’t get the value back. We did a great piece of work a while back for a company to optimise the way a store manager behaves in retail. So we did all the tests, all the research, everything was fantastic but the biggest challenge was the person at the other end. Their attitude was, ‘I’ve been doing this for 20 years I know how to run my store’. So that’s one of the trends I’m really excited about now is the increasing openness of business people to actually adopt this.

My last question would be how do you find being based in the South West?

I think it’s great. We’ve been in the South West for 12 years, plus I went to the University of Bath, which is a great place to study. When we interview around this area we get a great calibre of candidates from the universities like Bristol and Bath. We are really proud to be in the South West and we have some great clients in the area.

One last question on the business, I have to ask, is the cat story true?

[Laughs] Yes, it is certainly true! It comes right back to the weekends 15 year ago when me and my business partner, Matt Aldridge, were trying to think of a great name for the business and failing badly – I remember one of Matt’s ideas for a company name was ‘Stats Entertainment’ to give you a picture of how badly we were struggling. In the end we decided to name the business after Matt’s cat, Mango. But naming the company Mango is probably the best thing we’ve ever done, even if we’re now stuck with having to give away toy cats at every conference!