In skiing, the “black diamond” run or ski slope is often referred to as “high risk/high reward.” You receive lots of “reward” skiing the black diamond slopes, but you have a significant amount of “risk” associated with variable terrain, such as the presence of trees and the possibility of injury. However, the black diamond slopes are very fun to experience, and you can mitigate the risks with preparation, practice, and a really good skiing helmet.
Data science is one of those “high risk/high reward” concepts in technology. An organization can gain huge financial rewards by being able to create competitive advantage from internal and external information. However, there is a downside. Technology costs and time to implementation timeframes can be significant. In particular, these costs can overwhelm organizations that do not have existing data center facilities or the specialized staff to install, administer, operate, and create an effective data science program. In addition to these costs representing a significant risk, they can negatively impact the return on investment (ROI) of data science initiatives.
What organizations need is the ability to limit their risks associated with the costs of a data science initiative, and thus increase the opportunity for a speedy breakeven point and positive ROI. Data analytics-as-a-service offers the rough equivalent of that “ski helmet” for infrastructure, and a “ski instructor” for operating the environment and getting the best results.
Data science platforms such as Apache Hadoop are often referred to as “free” since you do not have to pay for the technology, you can just download it and install. However, Apache Hadoop is not like a “free” game that you install and play. There is a significant amount of time and staff cost to install a Hadoop environment and manage the continuing operations of the environment. At EMA, we like to refer to this as Hadoop being “free” like a free puppy as opposed to “free” like free beer. Technically, both the puppy and the beer are “free.” However, unlike the beer, the puppy requires ongoing upkeep –feeding, training, walking, etc – once you accept it. Data analytics-as-a-service allows organizations to avoid many of the costs associated with Hadoop, and to effectively have a technology “ski helmet” to avoid the risks and costs of implementation. Data analytics-as-a-service takes on the responsibility for building, maintaining, and operating a data science platform environment.
Next, organizations need the practice and preparation to make the best of a data science initiative – or, the “ski instructor.” Getting the most of advanced analytics such as descriptive and predictive analytical models requires skills and certain amount of experience. Companies usually spend considerable resources on either hiring or contracting with data scientists. For a permanent hire, this can be anywhere between $150k-$170k per year in salary alone, if not more. For a contractor, the hourly rate for a data science resource can be upwards of $300 per hour, depending on their skill level. Many organizations can have difficulty allotting the salary to a specialist, and the hourly costs can be more than most groups are willing/prepared to pay. Data analytics-as-a-service providers will often provide guidance within their subscription price to provide data scientist expertise without the high price, and thus avoid a team wasting time with the wrong analytical models for their business problems. This reduces the risk of failure and lowers the costs associated with the data science initiative. Again, providing a direct impact to the ROI of the initiative.
Several software vendors offer this type of data analytics-as-a-service. Several of the established “mega-vendors” provide solutions such as those listed above based on their traditional approaches. There are also smaller organizations that offer new methodologies to the business intelligence concept. A new entrant to the field is Perceivant. Perceivant focuses on assisting organizations with:
Perceivant does this with a cloud-based architecture that provides analytical results many times faster than environments such as Hadoop, and without the risks associated with attempting to manage the environment and analytics internally.
What say the readers?
Do you think that your data science initiative could use some risk reduction? Do you think Hadoop is easier to install and operate than depicted above? Do you wish that someone had provided you with a “helmet” before heading down the slope of your first data science project?
Provide your comments below and/or ping me via twitter at @JohnLMyers44 with the hashtag #cloudDataAnalytics