The first real look at Big Data analytics of the year comes in the shape of a new Forrester Wave report. The space may still be dominated by IBM and SAS but it appears that 2013 is going to be a year with some serious competition with a substantial number of strong performers jostling for a place in the Leaders category.
In all, this Wave, the Forresters Wave Big Data Predictive Analytics Solutions, Q1 2013 -- which was researched and written up by Mike Gualtieri -- assessed ten different vendors using 51-criteria. The vendors were ultimately broken down into Leaders, Strong Performers and Contenders, depending on how they performed across those criteria.
Again, like other Forrester Waves and Gartner reports, Forrester insists that these findings are only guidelines and that potential clients should only use the Wave report as a starting point for more comprehensive research before they start investing in anything.
Big Data Predictive Analytics Solutions, Q1 2013
That out of the way, the vendors that made it into the Wave this time around are: Angoss Software, IBM, KXEN, Oracle, Revolution Analytics, Salford Systems, SAP, SAS, StatSoft and Tibco Software. In an interesting side note buried deep in the report, Gualtieri points out that Oracle was not very forth-coming with information on its solution.
While Oracle chose not to provide full information for its big data predictive analytics solution, we included it in the Forrester Wave based on our analysis of publicly available information,” Gualtieri writes.
Oracle, as anyone that has followed it knows, is never reticent about publicizing its achievements, so the question to be asked is what is Oracle doing? Why won’t it discuss its products? Are we to take it that there is a problem or -- more than likely -- is it preparing something major this year?
Predictive Analytics Rise
All that said, there are three takeaways in this research that clients and vendors should keep in mind:
- Enterprises have woken up to the advantages of Big Data predictive analytics and vendors need to respond to that growing market need.
- The result is that the market is growing because more business and technology professionals see these solutions as a way to address opportunities.
- For vendors, Big Data handling, modeling tools and algorithms are key differentiators.
The result, Forrester says, is that it expects the market to be “vibrant, highly competitive and flush with new entrants” over the coming year.
Big Data, Information Explosion
The Big Data market is a result of the massive growth in the availability of information and data since we all went electronic way back when. The result for enterprises is the development of huge data warehouses with buckets of information that, until relatively recently, no one was really able to manage.
Recent advances in business intelligence tools have opened up this information to enterprises to use as an asset in building their business. But predictive analytics goes a step further with advanced statistical, data mining and machine learning algorithms digging deeper to find patterns that can be projected into the future behaviors of customers.
The starting point then is here. Forrester describes Big Data analytics solutions as:
Software and/or hardware solutions that allow firms to discover, evaluate, optimize and deploy predictive models by analyzing big data sources to improve business performance or mitigate risk."
Data analysis can be done daily, weekly or even continuously. In order for enterprises to do it successfully they must:
- Set the business goals: Outline what business goals they are hoping to achieve with predictive analytics.
- Analyze data from a number of sources: Enterprises need to assess and identify all the possible sources of information. The more information, the better the analysis. With the current level of information, enterprises are advised to use a data visualization tool.
- Prepare the data: Data must often be re-processed before running analysis algorithms. For example, data analysts may need to enrich the data with calculated aggregate fields.
- Create the predictive model: There are hundreds of different statistical and machine learning algorithms and combinations that data analysts can run against the data to find predictive models.
- Evaluate the model: Predictive models are about all the probabilities contained in the information. Analytics professionals need to constantly work with different models that work with the data they are using for a specific problem.
- Monitor: Enterprises need to constantly assess the effectiveness of their results. Keep in mind the business idiom “Past results do not guarantee future performance”, Forrester says.
Big Data Predictive Analytics Evaluation Overview
To see how vendors compare to each other, Forrester evaluated the strengths and weaknesses of the top players in the predictive space. While the market is too young to really tease out trends the way it would in more established IT areas, Forrester says it expects the market to be highly competitive with new entrants over the next three years. Evaluations were made on the basis of:
- Current offering: Each solution architecture, data handling capabilities, discovery and modeling tool algorithms and model deployment options, among others, were all examined.
- Strategy: Each vendor was evaluated on its ability to meet customer needs and anticipate emerging customer demands
- Market presence: Each vendor’s financials, global presence, installed base and partnerships were all assessed.
Forrester also conducted a three hour interactive lab series with all of the vendors being evaluated to get a look at what each solution offered.
In the case of the ten assessed vendors, all were able to provide the following capabilities:
- A core set of predictive analytics functions and vendors that can offer one or more solutions. The features that Forrester considers core are: data preparation, predictive model building, predictive modeling and predictive modeling management.
- An analytics solution that crosses domains. The products covered in this paper are general purpose predictive analytics solutions that cross all verticals and sets of functionality.
- A market presence and reference-able customer base and reported at least US$ 5 million in revenues from predictive big data analytics solutions. They must have 50 in-production customers for big data analytics, spanning more than one geographical region
- A technology that has sparked interest from the market, or has attracted Forrester's interest because of a technology.
The Leaders offer a rich set of algorithms that enable users to analyze data, and architecture that can handle big data and tools that cover the entire analytics life cycle. As might be expected from news over the past year, the biggest players are IBM and SAS. SAP however, which Forrester describes as a relative newcomer is also performing well.
The Wave is broken down as follows:
- Leaders: IBM, SAS
- Strong Performers: Tibco, Oracle, StatSoft and KXEN
- Contenders: Angoss, Revolution Analytics and Salford Systems
The SAS Enterprise Miner tool, Forrester says, is easy to learn and can run in database analytics or in distributed clusters. On the other hand, IBM’s Smarter Planet campaign and the acquisition of vendors like SPSS, Netezza and Vivissimo has put IBM up where it is.
IBM also comes with complementary solutions like InfoSphere Streams and Decision Management that adds to its attractiveness across organizations that are looking to integrate predictive analytics across the entire enterprise.
SAP is a newcomer to the big data predictive analysis market.Its architecture and strategy has been built on years in the information space, pushing it straight into the Leaders category.
It has also differentiated itself by placing its SAP HANA in-memory offering at the head of its portfolio, along with an in-database predictive analytics library and a modeling tool that looks like SAS Enterprise Miner and IBM SPSS Modeler.
Tibco, Oracle, StatSoft and KXEN all made it into the Strong Performers space, but missed the Leaders category because of lower architecture scores compared to the top players.
However, they all had substantial offerings. For Tibco, its Spitfire advanced data visualization tool was considered one of the main attractions for Forrester, who says it will be attractive to data scientists.
Oracle's solution centers on offering in-database R as well as a strongly built and robust analytics technology. Oracle’s recent strategy is to incorporate R into its database and datawarehouse offerings, but R is a language that is difficult to learn and Oracle needs to hide that.
StatSoft comes with a comprehensive set of analytics algorithms and is particularly useful in manufacturing use cases, while KXEN collapses the normal predictive analytics life cycle by automating the predictive model discovery process.
Angoss, Revolution Analytics, and Salford Systems all fall into the Contenders category, having a narrower focus than other vendors that were evaluated.
The best set of tools for decision trees comes from Angoss, which also offers cloud solutions that can be introduced quickly for fast results. Meanwhile, Revolution Analytics is trying to set itself up as the provider of solutions for the commercial sector, using the R programming language as well.
For its part, Salford Systems says that its implementation of analytics algorithms -- including CART, MARS and TreeNet -- are better than any of the other vendors evaluated. In this respect, it has managed to build up a reputation among the data scientists that it works with. However, unless it broadens its focus in terms of architecture (support for big data and life cycle tools), it will remain a niche player.