Montreal-based Datacratic is turning challenges into opportunities and helping enterprises turn static data into real-time, actionable intelligence
A Q&A with Datacratic CEO James Prudhomme and CTO Jeremy Barnes. The Montreal–based company was founded in 2010 as Recoset, and raised $2 million in Seed funding earlier this year. Investors include BDC Venture Capital with participation from Real Ventures.
SUB: Please describe Datacratic, and the value proposition you offer to the enterprise.
James Prudhomme/Datacratic CEO: Datacratic is a software company that is focused on applying machine learning and predictive analytics to real-time data flows. Essentially we have built a real-time data platform that can ingest tens-of-thousands of events per second and then run real-time predictive models on that data; the models are then made available as a service to applications which require them. We’ve built a suite of products on our platform which are targeted to the real-time marketing and ad-technology verticals; our products are available on a SaaS model. The applications have tremendous value for the real-time advertising exchanges which are quickly dominating the online ads space.
SUB: Who are your target users?
James: Our target customers are ad technology and marketing platforms who have real-time data streams and need to make decisions based on these streams in real-time.
SUB: Who do you consider to be your competition?
Jeremy Barnes/Datacratic CTO: We can split our competition into two main categories. There are machine learning as a service platforms, from small startups like precog.io to larger established companies who are repurposing their monolithic applications into a more service-oriented architecture. We are currently not competing directly with these companies as they have a platform-oriented go-to-market, whereas our vertical specific go-to-market tends to be a better fit in the digital media space. Then there are other companies in advertising technology who provide optimization, although this is rarely based on a standalone optimization platform.
James: We expect to see more competitors in the near future. The current crop of big data companies tend to mostly be focused on the capture and storage of data. That does not necessarily answer the entire equation. We feel that we have a head start on our competitors given that the team has been developing the platform for more than two years now. As Jeremy pointed out we also feel that our vertical go-to-market strategy provides an advantage over other enterprise software companies.
SUB: What differentiates Datacratic from the competition?
Jeremy: I would say the main differentiator is that we provide genuine causal machine learning models that operate well at scale and in real-time. It is one thing to take a static dataset and generate a model that does well according to some metric given unlimited time and resources; it is another thing to train that model in real-time and generate results with real-time data, which is what our platform has been designed to do.
James: From an ad technology perspective, our ability to calculate a conversion probability or a bid price for an ad impression, in milliseconds, is key—as is the ability to handle vast volumes of real-time data, up to tens of thousands of events per second. Datacratic has a very strong engineering focus and our products are designed to solve very real problems for our customers. We distinguish ourselves from other players in the ad-tech space by focusing on technology. We do not sell advertising or data, we provide software tools for others who then in turn sell advertising.
SUB: When was the company founded and what were the first steps you took in establishing it?
Jeremy: Datacratic was founded as ‘Recoset’ in February 2010 based on ideas we started exploring in late 2009. Originally we planned to enter the ecommerce recommendation engine market. That market proved to be very difficult for a small startup to penetrate—we often found that the engineers who were tasked with integrating our technology were creating internal recommendation projects at the same time, and our technology was seen as a new cost center rather than a competitive advantage. After a near-death experience we took a step back and evaluated the strengths of our technology: real-time machine learning, ability to operate at large scale, an ability to join together structured and semi-structured data, and the economic modeling layer on top of the machine learning. These advantages were very pertinent in the digital media space. We have spent the best part of the last year executing on this pivot.
SUB: What was the inspiration behind the idea for Datacratic? Was there an ‘aha’ moment, or was the idea more gradual in developing?
Jeremy: Our original vision was to bring to market a predictive analytics platform that would make data scientists ten times more productive and lower the bar on putting a useful model in production. In my previous startup in computational linguistics and through data mining competitions I saw models failing again and again due to the same types of problems, especially related to biasing. Even highly accomplished ML researchers were falling into the same traps. I also saw that 90 percent of time was spent on tasks that were highly skilled and subtle, such as bias elimination and building feature generation pipelines, but still mechanical in nature and could eventually be automated. The real-time and highly scalable aspects of the platform developed spontaneously as we applied it to problems where this was important.
SUB: What have the most significant obstacles been so far to building the company?
Jeremy: It was hard to gain traction with some of our prospects who assumed that a small Canadian startup must be out for a talent acquisition and were more interested in our team than our products or technology. We also had to work very hard to gain credibility in the digital media space, something that became a lot easier when we hired James as our CEO last August.
James: We had some fund raising challenges early on. Since the company had pivoted it was difficult to show customer traction to interested VCs. We seem to have overcome that challenge now. Being based in Montreal has been both an advantage and a challenge. It’s an advantage because we have great access to engineering talent and we’ve been able to build an amazing development team. It’s a challenge because our target customer base is in the U.S., mostly New York, and that requires a lot of travel.
SUB: You just closed a $2 million second stage Seed funding round. Why was this a good time to raise this round, and how do you plan to use the new funds? Do you plan to raise additional funds in the near future?
Jeremy: The $2 million is the total Seed funding which the company raised between Feb 2010 and December 2011. We didn’t announce it at the time as we hadn’t yet solidified our go-to-market strategy.
James: We made the decision to invest almost entirely in engineering talent and focus on building great technology and products. We’ve budgeted carefully and are just now starting to make some initial investments in marketing and customer acquisition. We intend to seek additional investment later this year which we’ll use to make larger investments in sales and marketing as well as doubling down on engineering and product development.
SUB: How does the company generate revenue or plan to generate revenue?
James: We have a fairly standard SaaS model which is calculated based on transactional volume. For example, in ad technology we’ll calculate licensing fees based on a CPM rate for each data point we ingest and each calculation we make.
SUB: What are your goals for Datacratic over the next year or so?
James: We intend to remain focused on real-time advertising and marketing as our key go-to-market vertical. We also intend to reestablish a presence in the ecommerce and retail space. Since the demand for real-time predictive analytics is high we are also exploring opportunities in peripheral verticals.
Datacratic – www.datacratic.com