When Dataiku was founded five years ago, the world of data and analytics had yet to hit its tipping point. Big data was at the foot of a hockey stick spike in Google queries and artificial intelligence had yet to move beyond the realm of Terminator films and into everyday chat bots. But it did not take long for all that to change.
As Florian Douetteau, co-founder and CEO, told DataIQ in a recent interview: “When I started the business, people with PhDs were starting to call themselves data scientists, but were not yet calling themselves business analysts. We have been doing machine learning since we started. You don’t set up a data science business without it.”
The company he co-founded with COO Marc Batty, CTO Clément Sténac and CDO Thomas Cabrol had the far-sighted vision of becoming a best-of-breed analytics and data science platform - Data Science Studio - that would take on global leaders in this space through combining data management, machine learning and production features in a single collaborative platform.
In December 2014, Dataiku raised a seed round of €3 million, then more recently went on to raise a $14 million Series A round led by FirstMark Capital in October 2016. This has allowed it to grow beyond 90 staff with a full-time presence in London, where DataIQ met Douetteau, as well as New York, Los Angeles and Paris. It is also working on its internet of things capabilities in DSS to handle sensor data in the next wave of data and analytics expansion.
“Organisations are increasingly thinking globally about data."
For Douetteau, the company’s future lies in following the breakout success of data science. “Organisations are increasingly thinking globally about data, developing solutions that work well in one territory then replicating them in other locations. That is a big challenge for them. CFOs are looking to standardise globally. We have to be able to replicate our solutions, which is why we have built the platform for scale so it can be rolled out in multi-national companies,” he said.
Another driver of its own growth has been the way data and analytics can transform a business model. Parkeon is a French business that supplies parking and transit systems to customers around the glob, with a focus on parking meters, ticket vending machines and mobile payment apps. But it has been faced with a changing business model as consumers switch to paying via their mobile, rather than in cash. It has identified switching to parking-as-a-service through leveraging sensor data on occupied bays, building predictive models using DSS and providing three alternative locations with more than 80% probability of being free to drivers in the “Path to Park” app. “Everybody wins, rather than ‘doing the circuit’ as we say in France,” said Douetteau.
Alongside wholesale business transformation of this sort, Dataiku has enabled clients to manage business-critical challenges on a less strategic scale, such as customer segmentation and personalisation. Boutique holiday retailer Voyage Privé needed to offer relevant travel offers to its customer base. It captured behavioural data, such as click paths and bookmarks, to create customer propensity scores using the machine learning abilities of DSS. Not only did this deliver a 6% increase in the total transactional value per unit per member, but Voyage Privé was also able to bring its entire data team inhouse.
As Douetteau explained: “We are seeing a surge in personalisation by online businesses which is often giving them a 2 to 3% boost in their top line revenues, which they didn’t think they could get. For a business with a turnover in the hundreds of millions, that is significant.”
Growth in the market for data science tools and solutions continues to grow, creating a clear path for Dataiku’s international ambitions. But having enjoyed a five-year growth period, which saw the business becoming profitable three years ago, there are also some potential bumps in the road ahead. One of these is the huge skills shortage around data science and a knowledge gap in machine learning.
“A lot of young graduates with machine learning skills are going to work in start-ups and digital companies which operate in a particular kind of work space,” said Douetteau, noting that its recent office openings in part attempt to replicate this working environment. “The appeal of start-ups is that the problems are more technical. In a career, you can seek those types of challenges or you can focus on things that are less complex, but have real impact. With machine learning today, 80 to 90% of it is not focused on technically hard problems. You don’t need complex algorithms. Only 10% is at the top end working on streaming data or automation”
As he noted, even at Google, many data scientists are being deployed on how to optimise ad placements. Many data science PhDs also struggle with the shift from the perfect data they enjoyed in academia to the world of dirty data inhabited by most commercial organisations.
"Data science needs to avoid becoming a bottleneck in the organisation.”
For Douetteau, the big challenge is finding practitioners who can grasp business challenges as well as understand the data science to solve them. “The other problem with the job market is that you can find people with really excellent skills, but they have also got to have business expertise to understand the parameters, what you can and can not do in a market,” he said. A core element of Dataiku’s proposition is that it offers both types of skills set and is always firmly rooted in fixing a business problem.
“There is a communication problem between the geeks and non-geeks - that is why I set up the company. There has been progress. There are still data scientists who are only into the techniques and are not happy unless they are digging around in big data. But the demand is increasingly from lines of business and data science needs to avoid becoming a bottleneck in the organisation,” he warned.
Like Parkeon, a growing number of companies are looking into how data and analytics can support new propositions, especially in the as-a-service space. Said Douetteau: “The opportunity for us is in helping large enterprises to think about innovation in a meaningful way.” But just as he launched the business just at the right time, given the explosion of interest in data science and machine learning which subsequently happened, so, too, it needs to be aware of rapid changes to its marketplace. The competitive threat now comes from open source and analytical languages like R and Python. But Douetteau is calm in the face of these trends. As he said: “Nobody can work out how to monetise them.”
Related articles: What does it take to be a great data scientist?