Earlier this year at SAP's Business Suite 7 launch I saw a demo of some
text analytics capability developed by Business Objects for sentiment
analysis used as a kind of an online sentiment monitor feeding into SAP
CRM. Interesting concept and one that is increasingly compelling for
businesses as more and more real time social content is created. There
are a couple of key bits of technology at play here: real time content
creation, real time search (text mining) and some linguistic analysis
algorithm. Beyond these basics it comes down to what do you do with the
analysis once it is located, captured and run through the language
sausage grinder.
The information available about your brand on the real time web
falls into two basic buckets. On one side there's the immediate
customer issue / problem / question. I've talked about this customer
service opportunity, the delivering great service "when, where and how"
the customer wants it and plan to make the topic a part of a larger
post on social customer service in the near future.
Today I'm talking
about a broader category that is your real time social web brand image
and opinion (or sentiment). The customer service information is
specific and individual, social brand sentiment is an aggregation of
social data from across the real time web. From this aggregated data it
is possible to gauge overall public opinion about your brand at any
given time (assuming you can collect that data of course). The data
source is any or all online forum where people gather and discuss their
opinions from public sources like Twitter and Facebook to your own
private label communities. Once collected the data can be analyzed as
text and using linguistics sentiment trends can be established from the
tone of the language and the type of words used to describe feelings
and opinions. The more data and data sources the more likely that the
sentiment represented accurately reflects online opinion trends.
OK, cool technology, but why should you care? Think about brand and
how, in the past, in an offline world you would have tried to capture
public opinion about your brand. First of all it was very expensive
using surveys or focus groups. Secondly it was very time consuming and
the data could easily be out of date by the time you had completed the
analysis. This method is really historical brand sentiment at best.
Today, using real time web, you could know with reasonable accuracy
what the online world (a world this is rapidly growing all the time)
thinks about your brand with reasonable accuracy. As a marketer that's
a powerful tool. The uses span from crisis monitoring to marketing
messaging (and method) testing.
Tools to enable sentiment monitoring and analysis are starting to
emerge from both existing CRM vendors and also start ups with new
product offerings. I mentioned the SAP / Business Objects
capability; while not a full product the text analytics engine, it does
have the capability to do sentiment analysis. Web analytics vendor Omniture
offers Twitter integration to its SyteCatalyst analytics platform that
allow sentiment monitoring. It does not, however have a linguistic
analysis capability but instead lets you search keywords and
combinations of keywords and slice and dice those searches. Sysomos
offers broader social web analytics (2 tools, Map and Heartbeat) by
searching blogs, wiki's, and social sites like Twitter.
Once the data
is collected it can be analyzed in a variety of ways (but no text
analysis / linguistics capability). Evri,
a Web 3.0 Semantic search engine, has a sentiment web API that pulls
search data into it's semantic engine to provide sentiment measurement.
The tool is basically binary, measuring positive or negative sentiment.
SPSS has rich text mining capabilities in its PASW Modeler 13 product that allows deep analysis of text mining results.
Autonomy, Nielsen and Techrigy all have sentiment analysis capabilities as well.