Turning the Predictive Analytics Model on its Head: A Spotlight Q&A with Simon Arkell of Predixion

Originally published November 5, 2012

This BeyeNETWORK spotlight features Ron Powell's interview with Simon Arkell, CEO and Co-Founder of Predixion Software. Ron and Simon discuss Predixion’s disruptive approach to predictive analytics.
Simon, for those in our audience who are not familiar with Predixion, could you provide us with an overview of your company and your product focus?

Simon Arkell: Predixion is now a few years old, and we set up the company to fill what we felt was a pretty significant gap in the business intelligence industry, specifically around predictive analytics. We spent a lot of time rethinking how predictive analytics is done, how predictive analytics solutions are developed, and how they're consumed. We came up with a great team, many of whom are former data mining or predictive analytics experts from Microsoft. We also have a great team of people who've been very successful previously with starting up enterprise software companies, filling a very strategic gap in the market, and then providing extreme value for that market.

As it relates to predictive analytics, we looked at the incumbent vendors – companies like SAS and SPSS, which is now a part of IBM. These are the technologies that have been around for quite a long time. What's broken about the model is that these are products that need highly trained experts, specifically in data science and, in many cases, people with PhDs.

Although they have tried to develop easy-to-use tools, it doesn't really change the fact that you need very extreme training in order to be able to use this technology. So we came up with a development environment for predictive analytics models and a platform, our server platform, that allows the end user to use a very familiar tool, which is Microsoft Excel. We have a complete workbench within Excel. It’s wizard driven, and it allows a business analyst to learn predictive analytics much more efficiently than if they were to use another third party and very complex tool.

Through this workflow built into our Excel client, they are able to cleanse their data, prepare their data, and create predictive models using any number of different algorithms, test those models for accuracy, and then deploy them in an automated fashion.

We're all about making predictive analytics available to many more people – democratizing it – and then making it easier for people to consume the output of those predictive analytics results in environments that they're accustomed to using, the workflows that they use on a daily basis. As an example, we could be optimizing a marketing campaign where we just want to find out which of our prospects are most likely to purchase a product and then market to them. The end result may be a new campaign that's created automatically inside of a CRM tool.

Another may be within a hospital setting, the output of risk models or risk profiles of patients into a case management tool that allows the case manager or the nurse to intervene with a patient in a very specific way based on the risk profile of that particular patient.

We provide an end-to-end solution. We’ve made it very easy to create the model, and we also provide a collaboration engine to make the sharing of the models and the datasets much more social. We’ve also made it easy to deploy these models into production, and they are much easier to consume by end users who've never heard the word algorithm before.

Extending predictive analytics with Excel, which has been the de facto standard for many, certainly opens up a large audience to predictive analytics. Are you seeing predictive analytics becoming more mission-critical in the C-suite?

Simon Arkell: We absolutely are. Our vision is that all business intelligence will be predictive. As you point out, Excel has been the de facto client for business intelligence, whether people like to admit it or not. But the problem with BI is that it has historically been retrospective. It's fantastic to get dashboards to executives to give them a bird's-eye view into real-time happenings in the organization that may be spread out globally. They can pull this information in and look at historical and real-time trends. At that point, the problem is that you’re expecting executives to use intuition to make predictions of their own. Should we rely on the intuition of a number of different executives who may have their own biases? I think the answer is no. We want to use real data science to provide the ability to take action on this information.

As an example, do we really want to forecast our inventory levels by looking at our historical sales numbers and then just make a guess as to what they might be in the future? How can you do that in these days of big data when you have hundreds of millions or billions of transactions, and hundreds or thousands of data points about each one of those transactions? Is it really possible for a human being to assimilate such huge amounts of data and then make a decision as what to do next just based on their individual intuition? I think, of course, the answer again is no. Our vision is to enable all business intelligence with predictive capabilities so that you're using real mathematical algorithms and data mining to assist in making a decision. Think about what that would mean for a doctor making a life or death decision based on what is increasingly an unmanageable amount of data. Being able to provide actionable predictions in an interface that is easy to understand for a less technical person is what it's all about. We see the benefit of that perspective, and I think the industry now is absolutely catching up to our vision.

Well, it's always nice when your vision gets accepted. Can you tell our readers a little bit about Predixion Insight and how it works?

Simon Arkell: Predixion Insight is a software platform that allows for the full end-to-end deployment of predictive analytics in an organization. It starts with a developer, possibly a business analyst or someone with domain experience, who works with data to create the predictive models by pulling in the data, creating the models, testing the models, and then deploying them. The developer benefits from a collaboration platform that allows regular analysts or web groups, maybe even marketing analysts, to collaborate and drill in on a particular model to see if it is really what they're trying to predict, to determine if they’re looking at the right data in order to create these models, and then determine if it is going to be effective for the organization. Collaboration is something that's taking off across the software industry in general, but it's never been done before with predictive analytics. We've come out with the first ever collaboration engine for predictive analytics.

Then, once the team is happy with the model that's being developed, we can push it very quickly into production, which means that model is now sitting out on our server, or in the cloud, and the data is basically updating or querying the model in real time automatically, and then outputting the results to an iPad, or some sort of tablet, or thin client, or line-of-business application that the non-technical people can use and take action with.

We see this from a holistic perspective, whereas our competition has always looked at this as a data science black art where PhDs are sitting in a back room creating great models, but then not being able to do anything with them – for example, not being able to have a nurse take an action as a result of something that they recently analyzed. The holistic approach is something that is resonating in a big way in the industry with our customers right now.

Predictive analytics has always been the domain of quants and statisticians. Are they involved also with Predixion Insight or who is the general end user for your product?

Simon Arkell: We're not replacing quants and statisticians. The problem is there just aren’t enough of them.

And if every university around the world churned out only data scientists, then we still wouldn't have enough given the big data explosion that’s happening across the world. We're a very disruptive vendor to SAS and SPSS. In fact, we beat them head-to-head fairly regularly because this is not just a data science tool. This is a platform that provides the benefits of predictive analytics to many more types of users in addition to the quants and statisticians. We're seeing an explosion of the number of users that can use this technology, and we're selling seats that are in the hundreds or thousands, whereas typical vendors in this space have historically sold very complex and expensive software to very few users. We're flipping that model on its head, and that's why we're becoming so disruptive to the industry.

Obviously, you compete with SPSS and SAS. What do you feel is your biggest competitive advantage against them?

Simon Arkell: Well, I think there are a number of areas where we're much more attractive to our customers. First, our end-to-end solution is very attractive because so many more people can use our platform than those who can use SAS or SPSS – not just the data scientists, which is really where SAS and SPSS are targeting. The flexibility of our platform means that we can deploy these solutions very quickly. We also provide a very attractive consumer interface. We provide a web-based application that allows these end users to start interacting with predictive analytics output and make decisions very, very quickly as a result.

There are some new features that we're getting ready to release. One is a concept around semantic modeling, and I'll leave the details for the big marketing release when we announce it, but it’s a semantic model that means that for the first time ever predictive models are now portable from one organization to another. As an example, if a hospital in Michigan creates a great predictive model for readmissions, they could effectively send that model to a hospital in Hawaii. That Hawaiian hospital could very quickly get that exact same model into production without having to write it from scratch. Every other vendor in the industry has the problem of non-portability of models, and it really has inhibited the growth of the industry.

We've cracked that nut, and I think we have created something extremely disruptive. As a result, we will be launching a yet-to-be-released model marketplace that will allow for the submission of predictive models and the consumption of predictive models in an exchange or an app store type environment. We're very excited about that but also really feel that these innovations are completely differentiating us from the competition.

A company that is using SAS and SPSS could bring your product in to help extend what they're already doing. This could greatly help to enhance what they're doing with predictive analytics. Is that correct?

Simon Arkell: Absolutely. For anyone reading this who has a license with SAS or SPSS, we can import any predictive model that they've developed with those tools and it looks in our environment as though it was created by Predixion. You can use Predixion to create a similar model to test the accuracy against the one you've already developed with those other tools, but then it becomes a collaborative, social and consumable piece of technology or intellectual property that has so many more potential users than if it was still being used only by a quant inside of SAS.

Based on your experience, are there any specific industries that are most interested in predictive analytics today?

Simon Arkell: Yes, there are. The first one that comes to mind is healthcare where we're spending a lot of our time right now because the industry has so much to gain from creating wellness versus just being a fee-for-service industry. That change is going on right now, and we're finding that predictive analytics is a perfect fit for many of their problems. Other industries are life sciences, pharma, insurance, marketing, and high-tech. We've seen customers in a number of verticals, but we are seeing by far the most interest right now in the healthcare industry.

Healthcare continues to be one of the biggest areas today. Why would you say there is such a demand for predictive analytics in healthcare?

Simon Arkell: I think it's the perfect storm where you have government mandates and financial incentives to bring electronic medical records into being, and that's happening now. Many of these providers are deploying this through what's called Meaningful Use. But we're also seeing an explosion of big data as a result of these electronic medical records. Far and away, the biggest problem for the healthcare providers and even for the payers, the insurance companies themselves, is what to do with this explosion of data. So how can you start to get insight into the data and how can you benefit and improve your organization as a result?

We have seen use cases across the board, and we're selling our software to a lot of hospitals right now because they want to get ahead of some of the problems that they're being penalized for. As an example, readmissions are a big problem. There are nearly a million preventable readmissions every year in the U.S., and it's a $25 billion problem for the industry. In other words, if you're one of the bottom 25% of hospitals in this particular metric and your readmission rates are too high – in other words you discharge a patient for something like pneumonia or congestive heart failure and they return to that hospital within 30 days for the same problem – not only is Medicare or Medicaid not going to reimburse the hospital for that particular readmission, but also they're going to penalize the hospital across the board for all reimbursements. Some of these hospitals are operating at one or two percent profit margins, and that penalty could effectively take out the entire profit of a hospital for a whole year if they don't solve their readmissions problem.

October 1 was the first day that these penalties went into effect. As a result, you're going to see this not only accelerate, but also you're going to see new penalties around things like hospital-acquired infections. If you could predict which patients are most likely to pick up an infection in a hospital, and you can get ahead of that and stop that from happening, then there are billions and billions of dollars in savings just for being partially accurate in those predictions. What seems to me to be an unacceptable problem in the U.S. around hospital-acquired infections, known as HAI, is that 100,000 Americans die each year because they picked up something in the hospital. This is a massive problem that's costing too many lives. We really think that we're helping to save lives as well as helping solve the multi-billion-dollar problems in the industry. The time is right, and we’ve created a great solution that allows nurses and case managers to take action and prevent problems from occurring in the first place.

Simon, Another area that we see great interest in the BI world is mobile and we're seeing a lot, obviously, done with iPads. Is that also helping to accelerate predictive analytics?


Simon Arkell: Yes it is. Today people want their information at their fingertips. In fact, many doctors, nurses and case managers are using tablets now and really forcing the IT departments of hospitals to support those devices. We've created a very visual interface – a thin client application of sorts – that could be used on any PC but also a tablet. In the case of readmissions, for example, a nurse could be sitting bedside with a patient and could basically go through various scenarios with the patient as to what will reduce their risk of infection or their risk of being readmitted once they're discharged. If you can go through those scenarios and see with the patient what sorts of behaviors should be followed in order to reduce their risk, then the patient is much more likely to respond and take ownership of their own health. We're seeing a significant amount of adoption because of our mobile strategy. It's what was responsible for a very large customer win that we recently enjoyed, and we're seeing that type of adoption accelerate.

Simon, if someone was interested in learning more about your solution, is there something to download or is there some way that they can look at it prior to engaging you?


Simon Arkell: We are pleased to announce that last week we released Predixion Enterprise Insight Developer Edition which is a free version of our entire platform for anyone wanting to develop models. They don't pay until they decide to use Predixion in production. Our server platform can be downloaded free of charge and then install the Excel client, or for those comfortable with the cloud, they can use our cloud service, again free, by just downloading the Excel client and registering at our website. Simply go to www.predixionsoftware.com and have fun! We have taken the friction out of "try before you buy" and are getting amazing feedback about it.

Simon, thank you for telling our readers about your company and why Predixion Insight is such a disruptive product in the predictive analytics marketplace.


  • Ron PowellRon Powell
    Ron, an independent analyst and consultant, has an extensive technology background in business intelligence, analytics and data warehousing. In 2005, Ron founded the BeyeNETWORK, which was acquired by Tech Target in 2010.  Prior to the founding of the BeyeNETWORK, Ron was cofounder, publisher and editorial director of DM Review (now Information Management). Ron also has a wealth of consulting expertise in business intelligence, business management and marketing. He may be contacted by email at rpowell@wi.rr.com.

    More articles and Ron's blog can be found in his BeyeNETWORK expert channel. Be sure to visit today!

Recent articles by Ron Powell



 

Comments

Want to post a comment? Login or become a member today!

Be the first to comment!