No trough of disillusionment here: big data startup Ngdata helps telecoms lower churn

Big data has been so hyped that some have started to doubt out loud whether it will ever live up to the hype. You can try this at home: a search for the Gartner phrase “trough of disillusionment” turns up mostly blog posts about big data (see also: here and here).

But Belgian based startup Ngdata seems to be unburdened by this disillusionment. In March 2012, it raised 2,5 million in funding from ING . Earlier this year, in January, Ngdata released version 2 of its recommendation engine Lily, a big data meets machine learning tool that can sift through massive amounts of structured and unstructured data in realtime. Last week, it was named ‘one of the 7 big data companies to watch’ by Bank Systems & Technology magazine – and it announced the acquisition of business intelligence consultancy Enqio.

The Ghent based startup (with offices in San Francisco) is clearly picking up steam.

Ngdata was founded by Jurgen Ingels (Clear2Pay) and Luc Burgelman. Burgelman was a co-founder of Porthus, an early player in the cloud which went public in 2006 and was later sold to Descartes. After a short stint as entrepreneur in residence at the Flemish innovation center iMinds, Burgelman launched Ngdata in 2009 with Ingels, because they saw an opportunity in big data and business intelligence.

Burgelman: “Jurgen and I started our first startup around the same time, in 2000. We met from time to time to discuss startups and VC’s over beers. About 4 years ago we started discussing doing something together. I was on the board of a company in the field of business intelligence at the time, and I noticed that a lot of the big players – the banks and the telecoms had huge volumes of data that they weren’t really using because it wasn’t structured.”

“These are intensely competitive markets. Telecom and banking are commodity products, and since everybody is a client somewhere, you can only lose customers. The churn at telcos and banks is very high, and extremely costly. For a telco of some size, you can easily count on 15 to 20 percent churn, for a cost of several hundreds of millions a year.”

But there’s a very good cure for churn, says Burgelman: selling your customer more than one product. “By upselling, you can dramatically decrease the chance of churn, because switching costs become a lot higher. But that means you have to know your clients better, and target them better.”

The problem for the big players like banks and telcos, who might have millions of customers, is that the systems they now have are just not fast enough. “Most systems at this moment are data warehouses, not big data. Which means: data scientists can tell you precisely why a customer left three months ago.”

“So banks spend tens of millions on 360 degrees customer views, but if you speak to channel managers, they will tell you: we don’t know anything more than we used to. We don’t know what they want, what they’re looking at on the website, we can’t see what their latest transactions are, let alone what they’re saying on social media. Because if you want to add that layer, you’re talking about a million times more data points. You want all of this data when the customer is actually before you, so that you can make him a personalised offer in real time. Not three months later.”

This is what Ngdata does with Lily. It combines massive quantities of data from different sources (like QuickView, SAS, SAP Business Objects and Tableau) in real time, using algorithms and a proprietary recommendation engine. According to Burgelman, its engine is twice as good as the one Netflix uses to tell its customers which movies to watch next.

“Ngdata is a big data company, but with a machine learning angle – that’s rare,” says Burgelman, who adds that machine learning was actually one of his first loves. “Back in the early nineties I worked on neural networks and machine learning. But there wasn’t much of a market for it, we were really pioneering stuff. So I went to work on cloud solutions, which looked more promising at the time. Now seemed like a good time to pick it up again.”

I ask to expand a bit on how a realtime recommendation engine  can keep me from leaving a telco. I mean, when I’m sick of a telco, how would Ngdata stop me by looking at my interactions? “You can’t improve a bad product, that’s true. But what does help is seeing that someone starts making less calls, or that they stop using their credit card, or they call the helpdesk more often. You can often see churn coming – if you know which signals to detect.”

“Big data equals weak signals, as the saying goes. You don’t get sick of your telco from one day to the next. There will always be signs, but people are notoriously bad at detecting those weak signals based on more than 10 parameters. Algorithms are good at it.”

“The algorithms also allow us to personalise offerings and promotions. In traditional marketing, you would work with 15 segments maybe. But the ROI of campaigns to those 15 segments is traditionally very low. With our recommendation engine, we can make truly personalised offers and promotions that will always be relevant. That’s incredibly important when you’re trying to upsell.”

It sounds similar to the promise that Captain Dash makes, I say, which says we’re now in the era of synchronising data in realtime. “There’s a few solutions out there, but they’re mostly aimed at the middle market, and on mobile and couponing. Those are solutions that capture and sell data. But we’re more working with what the telcos and banks have internally – those are players that have more information internally than you can find externally. Due to the nature of their operations, they won’t put all that information in the cloud – hence the need to make an enterprise solution that fits in with their existing systems.”

Compared to Porthus, says Burgelman, NGDATA involved a lot less guesswork, and also a lot less of the iterating, validating and A/B testing that modern startup theory preaches.

“I think we went about it a lot more systematic than the first time. We said: we want to go to companies that have the budget for this, in a few verticals that are very similar to each other, with customers who feel a very urgent need to solve their churn problem, and where we can have an immediate impact on the product.”

“Very quickly, we ended up in telecom and banks, also because we knew those markets well from our earlier experiences. You could sell it to government and healthcare too, I think, but those are verticals that we just didn’t know well enough at the time. In telecom and banking, I would say that we already knew most CIO’s. Next, we started focusing on product development.”

Selling to the enterprise is about selling a roadmap

After a year of development, Ngdata sold an early version of its solution to Adeo, a large French retail chain (which owns Brico and Leroy Merlin).

“Selling to the enterprise is almost always selling a roadmap too,” says Burgelman. Adeo will work with NGDATA for the next four years. “These companies say: yes, we have that problem, and we want to work with you to arrive at a solution. Selling to the enterprise also means that your product roadmap is a joint exercise between you and the customer. It’s not just about what exactly you will do, but also about the priorities, about when you will do it. But fundamentally, our vision remained intact since starting Ngdata.”

One thing that Ngdata does have in common with modern startups is the agile development. “I would say we work reasonably agile, in sprints, yes.”

Recently, a Silicon Valley banker predicted that the $ 1 trillion valuation of SAP and others would be eaten away, attacked by startups who can offer lower prices. I ask if NGDATA is one of those startups disrupting the enterprise world with lower prices.

Brugelman: “Partly, yes. But we don’t really compete on price, rather on a functionality that the other players haven’t developed yet. When enterprises notice that their 360 degree vision is not realtime, it’s not easy for the bigger players to develop a new product. We can go in, take three months’ worth of data and show them a simulation of what our software can do. We can show them that with our product, they might have avoided 10% of their churn – which could mean a difference of 100 or 500 million € per year.”

I ask him why VC’s proclaim the enterprise to be “the next big space”, and yet we still see so many consumer startups pop up. Because selling to the enterprise is not easy, thinks Burgelman. “It’s complex. You need strategy and experience. Once you sell to them, our customers are very sticky. Our solution supports a number of important processes, so it’s difficult to let you go from one day to the next from one day to the next.”

“But they also expect an entire ecosystem to take care of them, including consultants. You can’t just put some people in a garage and hope that something will come out of it. So the costs of starting up are usually a lot higher, and the treshold is higher. It also requires you to be out there – you definitely need the right sales persons to sell to the enterprise.”

That’s where the acquisition of Enqio fits in, says Burgelman. “Big data projects at enterprises necessitate not only knowledge of big data and machine learning, but also a knowledge of business processes. We lacked the business skills when it came to people with deep knowledge of that. I think with the acquisition, we now have the critical mass that we needed.”

For the time being Ngdata will concentrate mainly on Europe, the US and Australia, because that’s where most of the traction is right now. Ngdata opened offices both on the West coast and East coast. “The US is a market of early adopters, and having a (415) phone number (for San Francisco, ed.) does help us there. Australia is interesting for the banks, because they have a relatively “clean” system. They have largely grown organically, not through M&A activity like here in Europe. That makes it easier for them to deploy new software initiatives.”

European entrepreneurs: get better at product marketing

Finally, I want to know what he learned as an entrepreneur in residence at iMinds. “Oh, it was very instructive. There are a lot of great people with creative ideas here. The challenge is to get it off the ground. We just aren’t as good as the Americans when it comes to positioning, product marketing and strategy. Maybe we don’t have the guts to go all in. It might have to do with funding or the type of VC’s we have here.”

“In Silicon Valley, most VC’s are former entrepreneurs – like Greylock. They make companies succeed. Here, we still lack the environment. I think iMinds is taking steps. I think we also need more of an attitude that celebrates success. In Silicon Valley, a success story makes people sell their house and give it a go too. Here, we think someone who made it probably did some shady stuff to get where he is. It’s a matter of mentality.”

How about him, I want to know? Burgelman did an exit before – is he helping young startups succeed? “I’m in a few boards, but I really have to limit myself to a few. Just being a business angel without actively supporting the company isn’t enough I think. It’s not just a matter of giving money, you have to execute on a mission. If you can’t do that, you shouldn’t bother.”

[Photo: Eric Fisher, Flickr]

Powered by Facebook Comments

About the author

Raf Weverbergh

Editor of whiteboard. Raf Weverbergh was a magazine journalist whose work appeared in magazines like Rolling Stone, Playboy, Mail on Sunday, Publico and South China Morning Post. He is the co-founder of FINN, a corporate communications agency where he advises startups and multinationals on their PR and Mustr, the easiest media database for PR professionals. You can contact him on Twitter, Linkedin or Skype (rafweverbergh).

Related Posts