A Guide to Sentiment Analysis Tools

Page 1 of 2

BEST PRACTICES SERIES

Article ImageSentiment analysis is like social media therapy. We attempt to learn how our audience feels about us-what are the complaints, issues, suggestions, and praise? We can also learn what they think of our competition, our industry, and our employees. Perhaps the most valuable data in sentiment analysis is what we don't find. It is alarming if your marketing is trying to highlight price, but there's no mention of it online. That indicates the need for a significant reassessment of your efforts. Sentiment analysis also becomes a powerful benchmark of your marketing programs as you can track the change in overall sentiment over time.

At its root, sentiment analysis is getting and analyzing data. The details and uses of sentiment analysis are more involved, so take time to understand your needs and choose an appropriate tool. Here are some points to consider while researching sentiment analysis tools.

REAL-TIME VS. AGGREGATE DATA

There are two ways to get your data. Real-time sentiment analysis tools track all conversations and push alerts about certain keywords, which is helpful for customer service. For example, if a customer is in your store and he's waiting on a long line, can't find what he wants, or balks at prices, you will get a notification as soon as he vents his frustration online.

The alternative is a tool that gathers aggregate data for a specific time period. We find that for most businesses, a quarterly review is sufficient. However, you may supplement this analysis with reports that focus on individual campaigns to assess their success.

DATA PROCESSING

Once you determine what data to analyze, you can decide how to process it. Most tools offer some graphical gauge of your overall sentiment. For many, it is a score, scale, or grade that helps you quickly understand the situation.

Most sentiment tools rely on keyword analysis and assign sentiment based on language algorithms. The issue here is that software is never perfect. The more data points a tool finds, the more accurate it will be, but small and medium businesses (SMBs) may have trouble finding a significant dataset.

Here is an example: You are doing sentiment analysis for a film whose target audience is women 35 and older. A tweet from a woman in the demographic that says, "So much fun! X is my new favorite movie!" would count as two bits of positive sentiment. Her 14-year-old son who saw the film against his will tweets, "So much fun seeing X with mom ... NOT."

Many algorithm-based tools will still mark the son's tweet as positive. A second issue arises here as well. If the son's tweet does get marked as negative, it should not hold as much sway because it is not about the movie itself. We have yet to see a tool with this level of scrutiny.

Page 1 of 2