Turn Data Into Insights With NLP: A Case Study With Kantar on Satisfaction Drivers for Snack Consumption in China

The Rise of Three Squirrels

Alibaba Singles Day 11.11 2019, domestic Chinese online snacks retailer, Three Squirrels, sold CNY ¥ 100 million (US $14.2 million) in just 19 minutes and 23 seconds. Founded in 2012, the company is now the biggest snack retailer in China, registered annual sales around 10 billion CNY ¥ (US $ 1.6 billion) in 2019, and particularly popular with millennial females.

Three Squirrels rise can be considered a flash success, which the company attributes to having a better idea of the dietary preferences and behaviors of local Chinese consumers. Because of this they able to push out rapid innovations that match the demand of consumers at that present moment and outdo local and overseas competitors.

The Potential of Opinion Analysis & Why Most are Not Good At It

In fact most brands are sitting on a wealth of consumer opinion data that could allow them to be as agile and effective and Three Squirrels. Most decision makers in China are aware that the paramountcy of online retail in the country has gifted them with new levels of information regarding customers, through online reviews, but not many are able to really utilize this data.

There are several reasons why only a few are gleaming actionable insights from customer reviews. Many brands and retailers lack either the right partners, internal capabilities to analyze this data for insights that can actually benefit their business.

The other key stumbling block is that while there are already a large number of opinion analysis tools on the market, the majority suffer from the same shortcomings:

  1. Require long manual setup
  2. Only possible to covering 8-10 predefined topics
  3. Limited by number of languages that can be analyzed
  4. Can only analyze at around 65% accuracy
  5. High set-up costs.

In essence, a lot of the AI used in the majority of opinion analysis tools to this data are not intelligent enough and are still constrained by human bias.

Beyond Human Bias: Using NLP & Big Data to Investigate Consumer Insights in China

At Re-Hub, we have worked together with Kantar to put together a 14-page report that effectively uses big-data and NLP to perform opinion analysis on consumer reviews in the snack category.

This was only possible using the NLP tools of our partner Revuze. Unlike other opinion analysis solutions, Revuze has removed the problem of human bias and created an AI that trains itself, providing insights unconstrained by human imagination.

Together with Kantar, we leveraged Revuze NLP capabilities to analyze over 400,000 consumer reviews on Chinese eCommerce platforms and automatically extract what are the aspects that consumers mention and its related sentiment.

Findings are as below


Overall Satisfaction

The first area we looked at was the overall satisfaction that consumers have toward snack brands. This is the cumulative sentiment representing the overall experience of consumers with the products.

From the data we can observe that 66% of both positive and negative experiences can be distilled to 4 factors:

  1. Price
  2. Shipping
  3. Flavor
  4. Packaging

Another key takeaway is that a bad flavor experience is more likely to lead to a negative experience than a good flavor dictates a positive experience. Conversely, price is likely to drive a positive experience but has much less effect on whether an experience was positive.


Aspect/Sentiment Breakdown

Category analysis was performed through the study of two elements:

Aspect sentiment Breakdown Snacks
  1. Frequency of Mentions: Our algorithms calculate how many times an aspect in mentioned in the pool of reviews.
  2. Sentiment: We asses the language used when discussing a specific aspect as being negative or positive. We then give an overall percentage score to the sentiment and index it against the entire category.

This is represented on the chart on the left, where bar-length represented total frequency of mentions and while the color indicates positive/negative (green/ orange/red) sentiment as indexed against overall category sentiment.

Four Drivers

Fut four aspects drive the majority of positive consumer experience within the category. We took the top five brands from our list and performed a deeper dive into sentiment towards different aspects.

  1. Flavor: The key aspect that drives positive sentiment is flavor, with Oreo coming out top for this. It is possible also to go deeper in understanding the specific flavors that engender customer satisfaction.
  2. Shipping: To Chinese consumers who prize convenience, shipping is extremely important.
  3. Price/Value for Money: Pricing is important as a product can be equally regarded as too expensive and too cheap. The aforementioned Three Squirrels is performing weakly in this category as consumers reflect sense that the price has risen over the years.
  4. Packaging : Chinese consumers react positively to packaging, this is one area where Three Squirrels really outperforms local competitors.

Conclusion

Opinion analysis with NLP gifts brands the opportunity to make better decisions and quicker decisions while leveraging data that is already at their disposal. The difficult part is finding the right tools to use to garner actionable insights. Using advanced data and AI solutions like Revuze in this case, brands in China can now take get their hands faster and deeper insights regarding the key aspects that drive consumption for their products and learn where they stand in comparison to their competitors.

The full report can be downloaded here. Reach out to Re-Hub to find out how you can digitize your insights functions to understand customers and improve decision making.