Moving from BI to Big Data Analytics in Pharma: Build It or Buy It?
Introduction Our world is undergoing a dramatic cross-industry revolution, yielding a digital universe that is doubling in size every two years, and is expected to reach 44 trillion gigabytes by 2020. 1 What's truly fascinating is how the business community has turned this massive overload of information into an advantage, rather than a paralyzing obstacle, by using Big Data to identify correlations, understand causation and allow for a better decision-making process. This explains why CIO's top investment priority for 2014-2017 is in BI and analytics. 2 But the question still remains as to what would be the best system to manage all that data, and produce the best actionable insights in the most cost-efficient way. This article aims to present and compare between two models for a Big Data analytics infrastructure; an in-house development and an outsourced solution. By underlying their benefits and disadvantages, we hope to identify the best choice available for best results in the Pharma industry. Traditional BI vs. Big Data Analytics To really understand the importance of Big Data analytics in an organization, is to realize that "while many IT initiatives can help organizations do things better, big data helps organizations know what to do better. It is one thing to know that bookings decreased when compared with the same quarter last year (an example of BI), but it is better to know why they decreased (analytics), and it is best to model and predict how they might increase going forward (Big Data analytics)." [3] 1 IDC 2 Gartner 2
In other words, BI solutions no longer suffice to handle the overwhelming amount of data, and produce not only insights of past performances, but also predictions of future business strategies. That's why there's an increasing investment in moving to Big Data analytics infrastructures. So, in considering both options of analytics infrastructures, we need to keep in mind that data is only a means to an end, and not an end by itself; in other words, a great analytics solution should free our hands of managing the data, to truly focus on its benefits and implement them in our business strategies. That, ultimately, is what we should look for in a Big Data solution. Big Data is Science Each of the elements in Big Data analytics solutions, from collecting data, to storing, displaying and analyzing it, is a complex process that requires specialists capable of bridging different functions within the organization and effectively communicating between them. 3 Unlike traditional BI systems, Big Data infrastructure is made of diverse technologies for each step, from assembling the system, to operating and maintaining it. 4 To put it simply, implementing a Big Data solution is a complex procedure, due to: The sheer amount and diversity of data The number of points of internal and external integration The wide variety and large quantity of users The expertise to ensure the analytic models and results can be trusted [3] (Rouda, 2014) 4 (CoolaData, 2014) 3
All of these suggest that organizations should view the implementation of Big Data analytics infrastructure as a continuous process, rather than a one-time project, that should be designed and maintained by dedicated experts in the field. 5 With that in mind, let's proceed to compare the two options, and identify the best way to ensure the most suited solution for your business. The DIY Temptation: Why hire someone else when I have my own in-house IT department? Well, that's a good question. There's no doubt building a solution yourself fits the organization's very particular, customized needs. Since no one knows your business better than you, it seems absurd handing the job over to an outside source; you select tools, you identify the problems, you chose the team of experts and you manage the whole operation from beginning to end. It also seems like the most logical option economically, since you already have the work force, and all you need is to have your IT team dedicated on this project. This is a common approach within the business culture and it often has to do with NIH (not invented here) syndrome. In other words, often, the preference to use an in-house solution stems from the organization's cultural attitude towards using third-party solutions to an internal problem. So, is this the preferable option for your business? Let's take a look at some other factors that might be of value when trying to solve our dilemma- 5 (Rouda, 2014) 4
When Temptation Meets Reality The nature of Big Data complexity dictates that we perceive it as a serious operation, not to be handled as another one-time project with one-time resources dedicated to its completion. From that perspective, building your own solution raises a number of crucial disadvantages that can largely be avoided when choosing to buy an end-to-end analytics solution. This is even more so in Pharma, due to the importance of domain expertise. Let's review some of them: The cost you pay for your Trial and Error process. Investing time and money on research, choosing the right tools, etc.; these are all inevitable stages in designing and implementing the system, and they're mostly avoided when turning to data analytics specialists. The difficulty to implement changes once the system is up and running. It is a vital characteristic for the system to have enough flexibility to allow adding changes over time, according to changing dynamics, and when you don't have that ability, your system lacks basic requirements. "Soft costs": although it seems a DIY solution may save money, in the long-run it's clear that this is not the case. The costs involved in developing, implementing and maintaining the system are much greater than those invested in buying an end-to-end solution; according to ESG, a "buy" infrastructure will yield approximately 21% lower costs than a "build" equivalent. 6 This is mainly due to the shortages of skilled personnel, money, time, resources and expertise of about 20% of the companies. 7 Not only that, but according to Gartner, more than half of all analytics projects fail because of insufficient budget; the system not being completed on schedule; or because they fail to deliver the features and benefits that are optimistically agreed on at their outset. 8 DIY no doubt increases this risk drastically. 6 (Rouda, 2014) 7 Ibid 8 (CoolaData, 2014) 5
Time to market: Keeping in mind our main goal in choosing the best analytics solution, which is using our data to benefit our business, the faster we're able to deliver the Big Data solution, the faster the business will realize any associated benefit. While a "buy" solution can be up and running within 9 weeks, it's estimated that a DIY solution takes 12 to 18 months longer! This gap in delivery time is critical for the business, and might stall other projects. Defocus of the organization from its core business. In the end, buying an end-to-end solution enables the organization to focus its resources (time, money, man power, etc.) less on infrastructure development and deployment, and more on the value the infrastructure is providing. Conclusion In our hyper-digitalized world, rather than a "one-size-fits-all" BI tool, companies are now moving towards vendors that provide industry domain expertise that can integrate their knowledge into their solution, thus taking the development burden off the IT department, and providing top notch analytics dedicated to the field. In Pharma, an outsourced Big Data end-to-end analytics solution can really build the business case, and overall lead to a more effective use of sales and marketing spend. The business value presented in a "buy" analytics solution is tremendous- identifying correlations, understanding causation, and allowing better informed decision making. Ultimately, accurate, precise metrics analytics will allow a business to optimize processes, collaborate and improve revenue growth and profit margins. Last, but definitely not least, in Pharma, an end-to-end solution provides what is perhaps the one most important benefit in the business world - a short time to market. 6
Commercial operations are able to use data sets to optimize sales performance, monitor new drug launches, optimize sales force activity, formulate closed loops marketing campaigns to physicians, track messaging effectiveness or social media patient outreach campaigns; and so much more. 7
Build vs. Buy Pros and Cons BUILD Pros Cons 1. Tailored to your needs 2. You are the only one in charge of data 3. You feel you have greater control over analytics 4. Everyone gets input 1. High infrastructure costs 2. Engineering time takes from core business 3. Challenging data and cleansing integration 4. Not flexible: harder to add features over time 5. Long development time 6. Siloed data/island of information 7. Greater potential for errors because developed bottom up 8. Difficult to utilize (ad hoc development) 9. Consuming top executives and expensive resources BUY Pros Cons 1. Faster time to solution - Up and running within 9 weeks 2. Easy to use and maintain 3. Feature rich 4. Unified reporting and data repository 5. Proven expertize: Pharma by design, knowhow and best practices 6. Replicable 7. Easy to integrate and add features 8. Doesn t require as much of your development team's time 1. Less organizational control 2. Outsourcing not always preferred in some companies' culture 3. Seemingly high cost 4. Consuming top executives and expensive resources 8
References 1. CoolaData. (2014). Building Analytics vs. Buying an End-to-End Solution. CoolaData. 2. Gartner. (2013, December 16). Gartner Predicts Business Intelligence and Analytics Will Remain Top Focus for CIOs Through 2017. Gartner. Retrieved from http://www.gartner.com/newsroom/id/2637615 3. IDC. (2014). The Digital Universe of Opportunities: Rich Data and the Increasing Value of the Internet of Things. IDC Analyze the Future. Retrieved from http://www.emc.com/leadership/digitaluniverse/2014iview/index.htm 4. Rouda, N. (2014, July). Getting Real About Big Data: Build Versus Buy. ESG. Retrieved from http://www.oracle.com/us/corporate/analystreports/esg-getting-real-bigdata-2228170.pdf 9