For all the silliness surrounding Big Data and Data Science, all the hype and all the controversy, there are actually very innovative and disruptive technologies coming from this area, this new approach to data management and analytics [DMA]. How do we categorize the vendors or the technologies that have never existed before?
Predictives
One new area is Predictive Analytics, also called Predictive Intelligence. Since predictions are not analytics, as the term is used in BI, and certainly not the Intelligence used in BI, I don't like either, but prefer the simpler "Predictives". Four companies with which I've had briefings, fall into the Predictives category, but each of these companies have very different approaches and technologies for performing predictives. These companies are Opera Solutions, Alpine Data Labs, INRIX and Zementis. There are other companies that I'll include in a full report after receiving briefings, such as KXEN, Soft10 and Numenta. By the way, Numenta's product is named "Grok". Given their differences, do they really all belong in the same category?
Opera Solutions: Acting on petabytes of data, Opera Solutions provides a signal hub stack starting with data management, going through pattern matching in the signal layer, and, enhanced by their own Data Science teams, resulting in predictions and inferences for better decisions for enterprise advantage, understanding the "signal" is more important than the underlying technology, to actually create front line productivity through signals manifesting and adjusting "gut feel" where machines don't direct humans but do the heavy lifting.
Alpine Data Labs: Alpine Data Labs brings mathematical, statistical and machine learning predictive methods to the data in situ, no matter how small nor how big the data sets, within a variety of RDBMS technologies and Hadoop distributions. Alpine Data Labs helps data science teams address the data where it lays, across data types and functional areas, working with all the data to bring insight to bear on better decisions.
INRIX: INRIX data science teams and technology provides unique predictives using connected cars, connected devices and connected people.
Zementis: Zementis brings predictive modeling into decision management through their data science teams, Adapa product and strong commitment to the predictive markup modeling language [PMML]. Through partners and customers Zementis works with traditional and innovative data sources to provide decision management from predictives, data mining and machine learning for marketing solutions, financial services, predictive maintenance and energy/water sustainability.
DataGrok
One of the more interesting things to come out of data science is how do you really understand the data that is being gathered and presented. Two of the companies with which I've recently have had briefings, challenge the categories of Data Discovery or Data Exploration. However, each of these companies have different technologies, and different approaches to fully, deeply understanding your data, and to being able to draw conclusions from the data before doing other, more formal analytics. Over the past month, I've had the good fortune of having very in-depth, in-person briefings by both of these companies. Both of these companies are helping those who need it most to truly, fully, deeply, easily understand their data. These approaches, while very, very different, both constitute an entirely new category. Beyond data discovery, beyond data exploration, I call this new category Data Grokking.
"Grok" as I wrote in 2007, means to
"to fully and deeply understand"; [but to you need some background on the word's origins]. It's Martian and not from any Terran language at all. It comes from the fertile mind of Robert A. Heinlein, and was brought to Earth by Valentine Michael Smith in Heinlein's wonderful 1961 novel Stranger in a Strange Land.
One of these companies is still in stealth mode, and I won't mention their name here. The other is Ayasdi, and Ayasdi takes a very, very interesting approach to grokking your data.
These two very different technologies, based upon very different science and mathematics, do indeed allow us to fully and deeply understand our data. Much like the Martian ceremony, the DataGrok allows us to mentally ingest our data, to realize creative insights from our data sets, and to recognize the fundamental interweaving among the data, that, prior to these two innovative firms, could only come about through a long, arduous struggle with the data sets.
As I mentioned, the one company is still in stealth mode, so I'll write about Ayasdi here.
Ayasdi
Ayasdi comes out of the intersection of Topology and Computer Science, as brought together by two Stanford Professors, Gurjeet Singh and Gunnar Carlsson. The project started as a DARPA contract that has spanned more than four years, comptop. The CompTop project included Duke, Rutgers & Stanford nodes. Topological methods discover the structure of the data - this is somewhat analogous to, but not the same as the probabilistic or cummulative distribution or density functions [pdf, PDF, cdf or CDF].
Ayasdi is focused on four markets:
- Pharmaceuticals, Healthcare and Biotech
- Oil & gas
- Government
- Financial Services
From this, you can see that Ayasdi customers go after expensive data, i.e. expensive to collect, expensive to use. Iris is the front end to the Ayasdi Platform, and while available as a private cloud, their offering is primarily SaaS.
The analyst community is trying to figure out where to put Ayasdi, thus my category of DataGrok. Another area of confusion is "What is the right tool of each step of the process from DataGrok to inferences and predictions?" Some of this stems from mistrust of machines, but we need machines that do more than count and sort, we need machines that help us to find insight and improve performance.
