The online reputation management industry has grown considerably and matured while doing so. In 2008, finding the conversation was all that mattered. Web software companies offering this service staked their existence on tracking down those all-important brand mentions in the Internet ether. However, knowing what to do with them once found was another matter entirely. Automated sentiment analysis didn’t exist. It wasn’t easy finding out whether your brand mentions were positive or negative.

Of course, the volume of online conversation about brands at that time was also quite low. Facebook and Twitter were still in their relative infancies and if you picked up 500 mentions of your brand in a month, it was considered, well… considerable.

Fast-forward to the beginning of 2012 and the online reputation industry has come a long way. Specialist Social CRM businesses have entered the fray and the demand for online reputation management has become serious. Radian 6, one of the leading ORM tools, was purchased last year for a sum in the region of $200 million. This has encouraged many competitors to try their hand at meeting the ever-increasing market need.

Stumbling blocks with automated sentiment analysis

Today, automated sentiment analysis is a reality and it plays an important role in meeting this need. Through learning algorithms, many of the online reputation tools out there will provide brands with an insight into the positive or negative nature of their current conversation. Being able to do so with a relatively high degree of accuracy has become the elusive and ultimate treasure.

There are some very good reasons for this. Computers have limitations when it comes to interpreting and contextualising human language. Detecting irony, sarcasm, humorous nuance and a multitude of other emotions is tricky for machines, to say the least. Needless to say, the accuracy that automatic sentiment analysis tools can achieve has inherent limitations. Crashing through those limitations and delivering greater accuracy in contextualising online brand conversation has become the industry’s Holy Grail.

Finding the pot of gold in online reputation management

Current sentiment analysis tools are built on learning algorithms. If you feed computers a set of training data, they will learn how to predict sentiment in a fresh set of mentions from this data. There are two limitations to this process, however. The first is that the training data needs to be of an exceptionally high quality. It can’t be generated by computers or algorithms as this would result in inaccurate and flawed training; ultimately a vicious and unsuccessful cycle.

The second is that the learning process for any machine algorithm is exponential. To go from 60% to 61% accuracy requires far more training data than the jump from 40% to 41%. Achieving the levels of accuracy everyone seeks would require an alarmingly large volume of training data.

Accuracy is important. Making strategic business decisions based on the insights generated from a brand’s online conversation requires a high level of trust in the quality of the insights and the accuracy of the data being presented. The market knows this and has responded to the current limitations within the online reputation industry by assigning their own resource to manually manage the data that’s being found online. This is a solution that’s less than optimal. Addressing this optimally will lead us to “the treasure” in terms of online reputation management; otherwise known as the pot of gold or the Holy Grail. The company that can offer a scalable and robust solution for delivering accurate online conversation and insights to brands will meet the market demand. There lies fame and fortune.