In her blog post last week, Erin Rodat-Savla highlighted the “serious limitations” of sentiment analysis today, supported by commentary from leading text and sentiment analytics consultant Seth Grimes on the implications for marketers. From the inability of algorithms to accurately decode complex human interactions and adapt to the evolving language of consumers, organizations are continuing to struggle to truly understand what their customers are saying.
The state of voice of customer analytics
More than half of all analytics projects are unsuccessful because technology providers fail to deliver on their promises, according to Gartner Research. Concurrently, for the first time in eight years, marketing technology budgets are increasing at the same rate as marketing revenues, and are projected to increase 3.5% this year. The demand exists for a solution that truly understands voice of customer content, but tools that claim to have the “all in one” solution have consistently disappointed end users.
True insight has never come solely from a pie chart or word cloud
Dashboards decorated with pie charts and word clouds are fun data visualizations (we even use them) when placed appropriately, but they do not deliver actionable insights alone in a dashboard. Analysts are commonly forced to dig deeper into the data to extract common themes within the conversation and provide examples to decision makers to drive change within their organizations.
Humans are at the end of every analysis process
“The market, unfortunately, is polluted with tools that claim to have sentiment abilities, but are too crude to be usable,” said Grimes, when interviewed by Rodat-Savla. Machines are not yet able to understand the context and nuances that a human can. At best, the leading sentiment analytics solutions are 60-70% accurate.
Brands cannot afford to ignore or misunderstand the intent and emotions behind conversations surrounding them. For that reason, it is understandable the results of our recent industry survey revealed that 89% of all analysts are still manually reading everything.
Rodat-Savla identifies the ability for machines that teach themselves as the next frontier for sentiment analysis, but until this technology is developed the solution is a harmonious balance between the humans and technology.
It’s not about replacing the analyst, it’s about enabling them to do their job faster by finding a way to automate the processes that that do not require contextual and nuanced knowledge. The future will embody greater automation with the advancement of technology, but humans still play an essential role in engaging with and teaching these machines what it truly means to capture the voice of the customer.
