Data Science Platforms: Myth v. Reality

The phrase “data science platform” has been bandied about a lot recently — at conferences, in market research, and in tech publications like this one. Forrester named data science platforms a top emerging technology last year, and companies using data science at an enterprise level are being wooed by offerings in a rapidly expanding marketplace of platform providers. But what is a data science platform, really? And is it more than just a buzzword?

First, a definition: Data science platforms are meant to encompass the whole of a data scientist’s work. That means they typically provide tools that help users integrate and explore data from varied sources, build and deploy models, and make the outputs of those models operational. Essentially, this suite of tools is meant to keep data science work transparent, reproducible, and scalable — and make it easy for a data scientist to push dynamic results (like the predicted outcomes of ad campaigns) to the people who make decisions based on those results, replacing or supplementing static (and quickly outdated) reports.  

These platforms are no flash-in-the-pan-product, either. Data science as a profession has blown up — data scientists have had the best job in the United States for two years running according to recruiting site Glassdoor, and data science teams at Fortune 500 companies like Cisco number in the hundreds — and enterprise-grade technology is just beginning to catch up to demand. How do I know? We asked Forrester Consulting to hold a barometer to the industry to find out if — and why — businesses are using platforms*.

The Rise of the Platforms

The last major wave of big data tech investment was focused on enabling data science for organizations: building data lakes, centralizing data, and scaling support to continually integrate data through technologies like Hadoop. But now that companies have access to big data, Forrester has found that data science platform adoption is poised to more than double in the next two years — rising from 29% to 69% by the end of 2018. The reason, the firm concluded, is that more and more companies will soon realize the potential benefits. Among them, survey respondents suggest, are an improved customer experience, more informed business decisions, better business planning, and increased operational cost efficiency and customer retention.

Those aren’t the only benefits to performing data science work around a central software hub. Forrester’s survey also found that tool sprawl, where the volume of tools exceeds an organization’s ability to effectively utilize them, was the number one challenge data-driven businesses face, with an average of 6.7 tools being used to find value in data, from business intelligence tools and relational databases to predictive analytics, streaming analytics, and NoSQL databases. And almost half (46%) of the 208 companies Forrester spoke with lacked an integrated approach to their data science technology stack.

‘Insights leaders’ are the real MVPs

Companies already using data science platforms, on the other hand, are excelling. Forrester identified a group of businesses that regularly exceed profit and growth expectations, which it dubbed “insights leaders.” These leading companies were most likely to be small and agile (53% report having less than 5,000 total employees) and — most notably — 88% of them use a fully functional platform to do data science work. The majority (62%) also have a data science development plan and roadmap in place, as well as top-down support for data science initiatives starting in the C-suite.

Insights leaders currently make up only 22% of the market, and are far ahead of their less data-driven peers when it comes to investing in data science and retaining analytical talent. But nearly every company surveyed — whether insight leader or laggard — reported that data science is an important discipline to develop, and ranks among their most important corporate initiatives.

Clearly, there are a lot of components involved in running a business that does data science well. But as the buzz surrounding platforms becomes steadily louder, it’s my belief that these tools will become a vital ingredient in the recipe for overall business success. Having the ability to iterate on live data models, share code, and push results to other departments doesn’t just affect the reports that land on your CEO’s desk — it informs product development, helps optimize marketing decisions, and much, much more.

*(Full disclosure: We wanted to take stock of the market because we offer a Data Science platform.)

Like this article? Subscribe to our weekly newsletter to never miss out!

Follow @DataconomyMedia

One of the most important tasks that a Data Architect is often asked to help with is the creation of an Enterprise Data Strategy. But why is Data Strategy so important and what exactly does it consist of, and lastly why is this a task that a Data Architect should be leading or supporting?

So, what is a Data Strategy? Let’s review what it isn’t first…

• A Data Strategy is not a list of generic principles or obvious statements (such as “Data is an Enterprise Asset”)
• A Data Strategy is not merely a laundry list of technology trends that might somehow influence the organization in coming years
• A Data Strategy is not a vague list of objectives without a clear guiding vision or path for actualization.
• A Data Strategy is not merely the top level vision either, it can expand into critical data domains such as Business Intelligence and eventually represent a family of strategies.

Now we will attempt to define what an Enterprise Data Strategy really is:

Enterprise Data Strategy is the comprehensive vision and actionable foundation for an organization’s ability to harness data-related or data-dependent capability. It also represents the umbrella for all derived domain-specific strategies, such as Master Data Management, Business Intelligence, Big Data and so forth.

The Enterprise Data Strategy is:

• Actionable
• Relevant (e.g. contextual to the organization, not generic)
• Evolutionary (e.g. it is expected to change on a regular basis)
• Connected / Integrated – with everything that comes after it or from it

This definition helps to understand what Data Strategy is; so now we need to understand why most organizations need one. Here are a few of the reasons why…

1. Without a centralized vision and foundation, different parts of the enterprise will view data-related capabilities differently. This inevitably leads to duplication of both data and data systems across the organization and thus makes it quite difficult to determine the ‘truth’ of one’s data and will also drive up costs.
2. The Data Strategy provides the basis for all enterprise planning efforts connected to data-related capability.
3. The Data Strategy is the tool that allows for unification of Business and IT expectations for all enterprise data-related capabilities. The more detailed and comprehensive it is, the better the chance that both sides will fully understand each other.
4. There is no better place to define the metrics or service level expectations that should apply across the enterprise.
5. This is the best place to explain thoroughly how management of enterprise data can be leveraged to support organizational mission objectives or processes.

A Typical Enterprise Data Strategy includes the following components:

• A definition of what types of data or information needs to managed from an enterprise perspective (and yes this ought to be fairly specific).
• A determination in regards to roles (organizational) in terms of who owns what data or data systems.
• A mission statement in relation to exploitation of data assets. So, we’ve taken for granted here that these are enterprise assets – what’s important is understanding how they ought to be used.
• Initial or top level expectations for enterprise-wide service level metrics (for data systems and data quality).
• Introductory versions of all domain-level or specific sub-strategies, such as; Information/Data Governance, EDW, MDM, Content Management, Big Data etc.
• Top level planning decisions or expectations for making those designs.
• Identification of key enterprise challenges and anticipated design decisions.

Now we’re ready to address why Data Architects are typically involved in creating and executing Enterprise Data Strategy. Data Architects are specialists within the larger field of IT Architecture, while some have wider architecture experience – others do nothing but work with data and data systems. Data Architects make good candidates for helping to craft Enterprise Data Strategy because they are typically charged with defining all existing and future data related systems capability. Architects often also have a good deal of experience working directly with business stakeholders and thus help to ensure both business and IT perspectives are taken into consideration while crafting the Data Strategy.

There are in fact few other roles qualified to lead this type of an effort. While CTO, CIO or CDO’s (Chief Data Officer) might quality to lead such a task, often times they are stretched too thin to focus on building the comprehensive Strategies necessary to make a real difference for the organization. The Data Architect can typically dedicate their full attention to this task and have the full support of all necessary resources (including the CXO level personnel) to ensure that the necessary analysis, negotiation and planning goes into the Data Strategy so it can be relevant and ultimately successfully.

Follow @DataconomyMedia

Stephen Lahanas is the Vice President & IT Architect of Semantech Inc. His work is primarily focused on SOA, Data, Strategy, Cyber Security and Semantic Technology. He has provided critical technical and leadership support in a number of industries including defense & government, telecommunications and education. His writing has also appeared on Dice.com & Slashdot.com.

(Image credit: Peter Miller)