SAP Overhypes Its Hana Sentiment Analysis
SAP's rapid-deployment offering promises 'big data' analytics results in less than six weeks. I don't buy it.
That's a long name and a bold promise on which SAP might have trouble delivering. For starters, many companies would be hard pressed to get a conventional database appliance into meaningful production use in less than six weeks, yet here is SAP saying you can deploy what will be entirely new types of systems for many of them--an in-memory database requiring specific new hardware and a sentiment analysis application--at break-neck speed.
More Software Insights
- The Critical Importance of High Performance Data Integration for Big Data Analytics
- Mobile DevOps: Achieving continuous delivery with multiple front ends and complex backends in Banking, Financial Services, and Insurance
- Managing Volatility through Smart Inventory Planning
- Measuring the Total Economic Impact- IBM Rational Application Portfolio Management
What really raises doubts, however, is the fact that SAP has virtually no track record in social network sentiment analysis. Yes, BusinessObjects acquired Inxight, and its natural-language processing and text analytics technology way back in 2007 (before SAP acquired BusinessObjects). But it didn't do much with that technology until it was reintroduced last year--just in time for the fervor around all things social. The Inxight technology is now a part of the SAP BusinessObjects 4.0 platform, announced in February 2011. More specifically, it's embedded as part of SAP Data Services. The 4.0 upgrade was actually released in September, and SAP offered no sentiment-analysis customer use cases at the time. There are also no reference or beta customers for the RDS for Sentiment Intelligence.
[ Want more on sentiment analysis? Read Sentiment Analysis: How Companies Now Listen To The Web. ]
SAP's natural-language processing and entity-extraction capabilities (for spotting people, places, events, or, more likely, brands and products within text) are more like toolkit-level components that a developer would use to create an actual sentiment-analysis application with user interfaces and workflows. Byron Banks, VP of Database and Technology at SAP, says that SAP Data Services capabilities have been enriched with templates and "standard capabilities" (whatever that means) that the company came up with working in collaboration with the SAP CRM team.
"You can reach out into the social media stream, pull in Twitter and other social media feeds and get some insight, doing some parsing of the data and categorization to get a view of the discussions and who's doing the talking," Banks told InformationWeek.
These words suggest that SAP can deliver both sentiment-analysis (insight into topics and relative positive or negative feeling) plus network analysis (some sense of who's doing the talking and whether they're influencers). I pressed a bit and Banks granted that knowledge of who is talking is dependent upon the commenter's Twitter or Facebook handle being on record with an SAP CRM system, not on actual analysis of network influencers.
Can these templates help you automatically prioritize brand-relevant comments that come from the most important network influencers, whether they're customers or not? That's unclear. And can the sentiment technologies recognize the intensity of sentiment and automatically classify posts by whether they are questions, complaints, or customer service problems requiring a response? Also not clear.
Social-media analytics companies like Attensity, Radian6, Crimson Hexagon, and NetBase have been working on these problems for years. And in fact, SAP already has a partnership with NetBase through which it's reselling that company's SaaS-based services with added social engagement capabilities through SAP CRM.
Sentiment analysis is an inexact science with many challenging subtleties that confound the most sophisticated tools. If companies want a rough thumbs-up or thumbs-down sense of what people are saying about brands and products, there are plenty of dirt-cheap tools that don't require a major database system install. In my book, a toolkit and a few templates are about as close to a true social-media analysis app as a dashboard tool is to a true executive decision-support system.
Even when implementing a mature and widely deployed sentiment-analysis application, experts will tell you not to expect meaningful results in as little as six weeks.
"It's not a simple matter of flowing data into prebuilt interfaces or applications," says independent analyst Seth Grimes, organizer of the Sentiment Analysis Symposium. "There remains the 'last mile' need to adapt the preconfigured software package to your own business processes, hooked into your own unique mix of enterprise and social information sources and databases."
It should tell you something that SAP's first instinct was to partner with NetBase, a company that is aggregating and analyzing 100 million social network and Internet feeds per day. Nonetheless, Banks says the RDS for Sentiment Intelligence will make a nice complement to that offering.
"If NetBase isn't tracking a feed that is important to your industry or geography, [the RDS] gives the flexibility to choose any stream you want to listen to, whether it's internal or external," Banks says. So one RDS advantage would be that you could point the sentiment-analysis capabilities at, say, customer comments in the CRM system gathered by phone, email, and other non-social sources. But the appeal of this RDS will likely be limited to SAP CRM customers using SAP BusinessObjects who also have other reasons to use the Hana in-memory platform.
SAP certainly isn't the only vendor trying to grab a share of the hot social media analysis category. Larry Ellison, for example, demoed "Oracle" sentiment-analysis capabilities in June that were based on technology from two companies, Vitrue and Collective Intellect, that had been acquired only days before. No doubt the capabilities exist, but integrating newly acquired technologies into an Oracle-branded product would clearly take longer than a week.
The real agenda of the RDS for Sentiment Intelligence is offering another reason to buy SAP Hana. I think it's a mistake for SAP to market Hana as a tonic for all ails. SAP Business Warehouse deployments and breakthrough, real-time applications of Hana are compelling. A sentiment-analysis solution that starts with a database is mission creep.
It's nothing new to see vendors checking every technology box and offering a product for every sales opportunity. But sooner or later, hypey claims and promises unfulfilled will catch up with you.
Every company needs a social networking policy, but don't stifle creativity and productivity with too much formality. Also in the debut, all-digital Social Media For Grownups issue of The BrainYard: The proper tools help in setting social networking policy for your company and ensure that you'll be able to follow through. (Free with registration.)