How does an enterprise leverage Big Data Technologies? What is the right strategy?
Despite the current buzz around the benefits of Big Data, the CXO community is still learning how to strategically use it. Based on my discussions with more than 25 leaders across large and medium enterprises, there is an inhibition not only due to initial investment and ROI but also a misconception that Big Data requires a large amount of investment, transformation and change. The essence of any good cloud strategy is that it highlights key initiatives and prioritizes them based on value, cost and trade-offs. The dilemma starts from relevance of Big Data usage in the organization: does it fit into the marketing, product or brand team analytics needs or technology team’s operational data processing requirements? Would the starting point be structured internal data or un-structured data from social media?
To add to these woes, leaders must evaluate the current technology to see if the current set of technologies can support business needs and the initial investment required for transformation initiatives. There is a misconception in some sectors that Big Data is relevant only if the enterprise is already leveraging analytics in a traditional manner or using data from social media. This could be attributed to examples from companies like Amazon, Facebook and Google, however there are numerous scenarios where Big Data can do things faster and cheaper when compared to the current technologies that allow enterprises to service more customers with lesser staff.
Contrary to general belief, I have seen enterprises or lines of businesses within large enterprises taking plunge into Big Data by starting small. While it is not surprising that the business impact of this would not be huge, this approach has given these enterprises the confidence to embrace Big Data.
For example, a telecom giant had all the necessary data in disparate systems but was not using the data efficiently because of a long processing time (3-4 days). With relatively small investments in Big Data technologies, the data is now readily available within a few minutes. The data is not only more useful for planned use cases but has also created huge business opportunities from the near real-time data insights.
How Small Can I Start?
The return on investment can be calculated by considering the time, people and material required to execute the project. The general rule for IT systems investment is to get return of investment within three to four years. Since the technology transformations have been more frequent in recent times, I have taken two years into consideration to demonstrate return on investment.
As an example, we can look at a theoretical existing IT system that is storing and processing large volume of data from disparate sources with data having potential of giving business insights. To look at the ROI I look at an existing IT project with standard investments around software, hardware and operations. Since it difficult to measure business benefits coming out of data processing using Big Data technologies, I will focus on the tangible benefits, from investments in cloud, required to execute medium IT system using open source technologies like Hadoop(1), Mahout, etc. over two years.
- HW and Software costs: IT systems play a significant part to keep things in context, I will be considering the popular Big Data Technology Hadoop ecosystem which runs on commodity hardware and low cost software. I have taken the cost of existing hardware and software in table 1 and replaced the same with new technologies in table 2.
- The other significant contributor would be the implementation cost and developing employee’s new skills over a period of time. I have added implementation costs of the new project in table 2.
- Operations cost include ongoing operation and support for the system along with any annual software support contracts.
Table 1: Existing Software and Hardware. All numbers in thousands USD
Assuming the cost of capital is 3% per year, the net present value of the investment is around 2.2 million USD.
Now consider the same system to be replaced by Big Data technologies in table 2. Taking implementation cost of the new project to be 1.1M USD with minimal software license investment (only open source support considered) and commodity hardware, the first Big Data implementation would be coming to around 2.5 M USD.
Table 2: New Software and Hardware replacing existing
technologies. All numbers in thousands USD
The operational efficiency would not only have an impact on productivity and labor savings but also allow enterprises to avoid expensive software and hardware costs. The direct benefits include improvement in operational efficiency and reduction in hardware costs due to usage of commodity class hardware. The intangible benefits, apart from enterprise getting competitive advantage due to adoption of Big Data technologies, could be in the form of enterprises getting an opportunity to do near real-time data analytics using historical data for business insights, launch new products, accelerate time to market and agile customer responsiveness.
The article was originally published on Express Computer on June 11, 2014 and is re-posted here by permission.