We all wish we could pick people’s brains for honest opinions and emotions…And we Brand owners are no different. We also seek the honest subjective opinion and feedback of our fans towards a new feature, product, or even an ad campaign we just launched. And since today’s technological advancements all combine to aid marketers sell their products and services, new tools were developed to extract this type of information from people’s minds. These tools are called Sentiment analysis tools or also known as opinion mining. Before you start thinking about voodoo dolls and black magic, allow me to clarify how these tools work. Sentiment analysis is using natural language processing, artificial intelligence, and statistics to extract and compute the sentiment content of a piece of text.
What is Sentiment Analysis, and how can brands benefit from it?
Brands are often talked about in multiple online social and nonsocial platforms and it’s nearly impossible for brand owners to keep track of all that information unless they are willing to hire a team of 50 digital analysts just to gather this data and it would probably take them around 10 years to do that. Anyway, to explain what sentiment analysis tools do in simple words is basically these tools get alerted whenever the brand name or company name is mentioned in any online source. The system then tracks this conversation in which the brand name is mentioned and starts scanning the text for familiar keywords. These keywords are entered beforehand and categorized into one of three groups; positive, negative, or neutral. So when the system scans the text and comes across one of the previously entered keywords it automatically analyses this comment as negative, positive, or neutral.
After the gathering and categorizing comes the analysis and reporting part. The previously collected reviews are interpreted and summarized to finally show a conclusion and statistics of whether this item in concern is mostly positively, negatively, or neutrally perceived.
The findings of these reports give brand owners an insight of how well a product is doing and if necessary optimization is needed.
In every performance analysis benchmarking must play a role, as there’s no point in knowing people aren’t buying your product without knowing what they are buying instead. Sentiment analysis could be used for competitor brands and products to evaluate their performance as well and your performance against theirs.
Regarding my personal experience with sentiment analysis tools and during my hunt for finding the one that fits the particular needs of the company I’m part of, being an Egyptian digital advertising agency with most of our clients based in Egypt and their fans speak Arabic or Franco Arab or a mixture of both. As you have guessed my first and main criteria for narrowing down the tools was the ability to gather and comprehend data in Arabic with its slangs and multiple dialects. I ended up with four tools for further consideration, which are; Brandwatch, Syenthesio, CrowdAnalyzer, and Radian6. And I have to mention this, most tools never tell you the full story upfront so you have to endure long (sometimes tortures) hours of tool demos with chatty sales people. I will try to save your precious time and offer a summary of my findings, however different business and social objectives and goals may require different criteria.
Criteria of Analysis
When analyzing and comparing the different tools the following criteria were considered for the analysis. 1) Sources of Data: Some tools gather data from social platforms like Facebook and Twitter only while others scan other online sources such as blogs, forums, and website reviews. 2) Audience Analysis: Analyzing the audiences’ demographics like their language, country, age, and gender. 3) Retrieving historical data: The ability to retrieve past data, which is essential for comparisons. 4) Support: The availability of a support team.
Note: The price was not included in the criteria since prices vary depending on service packages and requirements, also these tools use different currencies and are subject to exchange rates fluctuations.
|Data Sources||Multiple online sources||Multiple online sources||Multiple online sources||Multiple online sources||Multiple online sources|
|Audience Analysis||Analyzes audience’s language and country. But not their age and gender||Analyzes audience’s language and country. But not their age and gender||Analyzes audience’s language, country, age, and gender||Doesn’t provide language, country, age, or gender audience analysis.|
|Retrieving historical data||1 Month historical data||4 years historical data||2 weeks of Historical data||No historical data|
Challenges in Sentiment analysis
Seems like a marketer’s dream come true, doesn’t it? But just as every other tool there are some challenges.
Challenges of Sentiment analysis are;
- The complexity of human emotions towards things, people do not necessarily love or hate something (we all wish it was that simple). The same person can have mixed feelings about a certain topic or product which makes his reviews hard to categorize.
- Lexical content, sarcasm, and irony when extracted out of its sentence, can totally change the meaning.
- Negatives and word additives can also be misleading.
- Words with changeable meaning among different industries, so when entering these keywords one must be very careful as keywords don’t always reflect the same for different products. For instance the use of “thin” or “ultra-thin” when describing a smart phone is considered a positive review, while being thin when describing a mattress reflects a negative review.
- These tools face real challenges when it comes to analyzing foreign languages, let’s take the Arabic language as an example since we have personal experiences with that, analyze the heavily used Egyptian slang, the multiple Arab dialects or even the Franco Arab writing style all make it nearly impossible to find a tool that’s smart enough to analyze that with a decent accuracy at reasonable cost.
When using the incredible sentiment analysis there are some points that you need to look out for:
- The Maximum accuracy percentage I received after surveying different tools was around 88% accuracy.
- If you require sentiment analysis for a foreign complex language like my case, I would suggest the manual humanized revision of the analysis offered by some tools even if it’s at an additional cost, but it’s worth it.
- To get a tool that fits your needs it will probably cost a lot ranging somewhere between 500- 700 USD per month, so I recommend you make sure your clients/ top management will appreciate that service, know its worth, and are also open to the fact that some people will have a negative feedback to their product. Otherwise you will be wasting your money.
- Additional features are usually available at additional costs, so if you wish to add some features like humanized revision, or further historical data retrieving that’s mostly possible.
- And finally regarding keyword choices, I believe it all comes down to a trial and error strategy. Experiment with keywords and short phrases until you reach what best describes your product from customers’ perspective. You might be surprised what customers are calling your product out there in the digital world.