Coca-Cola taps into the power of product reviews and online ratings with Harmoni

Background

Large consumer brands like Coca-Cola understand the power of social media data and product reviews when it comes to company decision-making. In fact, data shows that around 9 out of 10 consumers rely on digital information while shopping, and two-thirds rely on reviews as their number one purchase influencer.

The Coca-Cola Company North America’s Technical Consumer Research group (TCR) is continuously scraping reviews and online ratings for an in-depth understanding of consumer opinions. These sources can prove to be extremely messy, fragmented sources of data. From the same reviews left on multiple domains, to proper coding and sentiment analysis of numerous topics, or tagging multiple themes to the same review, making sense of these is more than just downloading them from the internet.

Coca-Cola’s TCR team gathers product reviews from major online retail sources (e.g., Amazon.com, Walmart.com, etc.) and sorts these into meaningful topics.

The Solution

Coca-Cola’s TCR team turned to Infotools to help them bring this rich data to life in an easy and cost-effective way. Infotools has been working with Coca-Cola North America for over 20 years as its data analysis platform partner. After an initial proof-of-concept using social listening data, it was confirmed that Infotools’ cloud-based Harmoni platform was an ideal solution for providing easy user access, analysis and reporting of product reviews. The platform is made for handling large, diverse data sets.

The text-analytics solution not only allowed Infotools to ingest the metadata and pre-coded themes, it also provided TCR with multi-level sentiment analysis and AI-powered theme development for the open text product reviews themselves.

How we tackled the project

Each review is broken into theme sentences.

• Just under 137,000 reviews revealed over 410,000 theme sentences
• An average of three theme sentences per review
• Themes can be established across an entire set of data or a range of categories/products
• Data-specific themes (applying to a subset of the data) can co-exist within the Harmoni platform

Each theme sentence is associated with a base theme and sub-theme.

• 17 base themes were identified (incl. Other and None)
• 89 sub-themes were identified

A sentiment score indicates whether the theme sentence has a positive or negative sentiment within the sub-theme.

• Values lower than -0.1 indicate a negative association, e.g. “Don’t like the taste.”
• Values greater than 0.1 indicate a positive association, e.g. “Great taste.”
• Values between -0.1 and 0.1 indicate a neutral association, e.g. “I don’t like the taste on its own, but I like it in coffee.”
• Harmoni’s DISCOVER feature, which uses advanced statistical testing, automatically brings the most interesting findings to the surface to find stories in data faster by profiling the groups that matter

Economies of scale; insights of importance.

After using Harmoni, TCR’s product review database has quickly become an integral part of its “Voice of the Consumer Exploration” protocol. Having this data easily accessible and usable allows Coca-Cola’s TCR team to make decisions based on real customer feedback and avoid time-consuming manual steps for data analysis and reporting.

In Harmoni, TCR found a solution that could:

• Quickly ingest and organize large, diverse data sets, like the kind coming from product reviews.
• Handle multi-level sentiment analysis and AI-powered theme development.
• Provide a single environment for not only handling the gargantuan data processing tasks, but also take the data all the way through to analysis, visualization, dashboards and reporting.
• Allow team members to investigate and interact with the data, by doing things like sorting results by concept or key phrase to examine a specific data point.
• Give internal researchers easy ways to share the insights and democratize the data.

By employing Harmoni, Coca-Cola Company’s North Americas Technical Consumer Research group has greater customer understanding and can add business value through product reviews data.