Think about some of the most popular websites today. One of the main elements Facebook, YouTube, Amazon and Wikipedia have in common is humanity. We may be in the digital age, but digital is just the medium to connect humans to one another via massive platforms fueled by user-generated content.
When it comes to product reviews, such as those on Amazon and other ecommerce sites, the star rating is the precursor to the human element: the review text. It is known that 94% of people read the text of reviews (Fan and Fuel). Shoppers want to know what other humans, like themselves, think about the product’s performance, quality, safety and more. What a powerful human engine reviews are!
But for brands or anyone else trying to make sense of this human element, things can get tricky.
Let’s take a look at Monica Rogati’s Data Science Hierarchy of Needs and relate it to product reviews.
Here’s how this hierarchy is utilized at Channel Signal to bring structure to unstructured product review text.
- Collect. First, the reviews must be collected. At Channel Signal, we collect reviews from 60+ sources.
- Move/Store. Channel Signal populates these reviews into a single, secure platform.
- Explore/Transform. Channel Signal turns all of this messy data into clean data by assigning them to clean product titles.
- Aggregate/Label. Categories vary on different ecommerce sites, so Channel Signal categorizes, labels, aggregates consumer-generated star averages and text, and begins high-level analysis at this stage.
- Learn/Optimize. This actually starts with what the brand already knows. A “gut feel” if you will. Brands indicate common known issues and key brand metrics. This is then weighed against consumer opinion.
- Artificial Intelligence. This is where new concepts are uncovered in the product review text.
And that’s why Channel Signal has invested in textual analysis by partnering with one of the best Artificial Intelligence labs in the world. We believe that many of the underlying issues driving great or poor performances of a product, category or brand, lies in the words customers use.
So, we are developing our AI service to define the issues that are buried in the text, all the way down to the product level.
One of our brands, a large footwear company, wanted to know what their customers meant when they mentioned “heel”. Was the customer addressing their heel, or the heel of the shoe? Was the sentiment of their statements positive or negative? Is there a design or construction flaw? Textual analysis can be used to determine if there is a problem, and if so, put it into context to map out solutions.
The best way to uncover the real market conversation about your brand, categories and products is to balance star ratings with textual analysis.