Big Data, AI, Machine Learning, Hadoop, Predictive Analytics — we hear these terms every day from companies such as Cloudera, Trifacta and Dato (formerly GraphLab) that are securing many millions in financing. I believe that 2015 will be the year when the conversation moves from Big Data to the Dataware stack. Over the past twelve months we have seen a lot of companies across the big data spectrum emerge and while the language can be the same, there are clear product categories that have emerged which describe the market opportunity and future growth.
This is the Dataware stack. Dataware is the combination of infrastructure, data intelligence systems that apply algorithms and machine learning to the data, and the applications enabled by data intelligence that are changing how we do business and how we live our lives every day. And startups dominate the Dataware landscape.
We are at the very start of the data revolution. The consumer world got there first. Apps that know who we are and where we are and some other data points about us help us do a myriad of things every single day. On the business side, there has always been a lot of data but now there is not only an incredible growth in that data, there is an active appreciation beyond business analysts for using data in near real time to improve products and services.
What we are seeing now is a huge shift that is infusing data into every piece of our lives and making every app and service smarter. We have started to see this blending of data with what were rather static services with recent announcements from companies such as Salesforce.com and Workday.
The Dataware Framework is how I look at this new world of software in the age of big data. Dataware includes an Agile Data Stack of components for modern data applications and services and a Continuous Data Loop that brings usable data in and out of the stack. The Agile Data Stack has three layers including the underlying enabling infrastructure, the data intelligence layer, and the data-infused applications and services that benefit from those underlying components. The Continuous Data Loop is the representation of how data is continually being ingested, cleaned, visualized, recycled and refined and put back into the mix for future predictions so that modern applications can deliver intelligence in increasingly dynamic and personalized ways.
While most traditional IT customers and vendors are making moves to deal with the growth of big data — Dataware is largely the territory of startups and early adopters across every layer of the data stack.
At first, new sources of data and data systems will enhance and extend existing technologies such as databases, data warehouses and business intelligence tools. The new technologies will help unlock value in legacy systems and structured data silos along with new types and structures of data sources. But, as the Dataware infrastructure, intelligence and methodologies mature, data-infused applications and services will be built from scratch to disrupt industries and business processes.
Dataware will introduce net new processes and intelligence into the world’s oil exploration, research for cancer cures, advertising optimization, and yes, choosing movies and friends.
Here are four key principles fundamental to understanding the impact that Dataware will have on the technology industry.
- Big Data and traditional structured data work together as a “hybrid” of inputs to feed data-infused applications and services and will complement each other as these applications get built.
- Enabling infrastructure, including new types of databases (Cassandra, MongoDB, Hbase,) and data execution “engines” (Hadoop/Map Reduce, Spark), are primarily enablers and less likely to be where value is captured (when compared to the relational database era) in the Agile Data Stack.
- The Data Intelligence layer is where data, algorithms, data models and “pipelines” intersect to turn data into insights. More value will be delivered and captured in this layer than historic data “middleware” and BI tools have captured in the past. Companies focused on the enabling infrastructure today are likely to try and move up the stack into the data intelligence layer.
- Data driven applications and services are distinguished across two dimensions. First, are the data insights being delivered to a machine or a human? Second, are data insights/predictions being delivered in real-time or a batch/offline mode? Real-time insights delivered directly to a human end-user are the most challenging ones to run at scale. And, they are the most challenging systems to create a continuous feedback loop that delivers both instant gratification to the customer and compelling insights to the service provider.
It’s clear that not every company in the big data arena will succeed. There will be a lot of failures and there is already a lot of confusion around language and markets. The companies that will succeed will make themselves a core component of the Agile Data Stack or the Continuous Data Pipeline and will build their footprint from there.
Dataware is one way to frame the major areas of opportunity in what Mark Benioff recently called the “AI Spring” or what Microsoft is promoting with services like AzureML. But many questions remain including where the most promising markets exist, when those markets will be ready for rapid adoption and who amongst startup-ups and incumbents will emerge as winners and losers. What we do know is that Dataware will dramatically impact the technology industry over the next decade.
Matt McIlwain is an investor in Seattle with Madrona Venture Group who invests in enterprise, cloud and Dataware companies. Dato is one of his investments.