Detecting a reputation crisis through machine learning

Is it possible to evaluate the online reputation level of a brand and manage the crisis, through artificial intelligence and machine learning?

That’s what we did by analyzing Tyson Foods and classifying online conversations through dynamic market research techniques.

On the web, every day there is a huge amount of talk, of writing and a lot of content is. Each user talks about many brands, both directly and indirectly, posting photos showing brands, making reviews, giving judgments, advice, commenting on current events and why not, criticizing.

For each brand, all this information is very interesting. Each company must be careful to monitor and receive all the signals coming from social networks and the web, in order to understand the perception of its brand reputation, brand awareness and to be able to define some strategies and foresee crises.

So how can a brand anticipate and carry out efficient crisis management by monitoring the web? KPI6 has an answer for all of its customers’ needs, and the answer is called BrandX.

Brand X is a new product within the KPI6 suite that will serve as a global benchmark for analyzing a brand based on different variables, which may be the level of reputation, the trust perceived by consumers and citizens too, advocacy level about their best products.

In particular, the Reputation index is that part of KPI6’s suite that measures the perception that users have of a particular brand. It is a variable index over time, as it allows us to measure the variation in perception day by day by crossing different data sources. It is a multidimensional index, that measures the perceived about certain dimensions such as product quality, company policies, sustainability and CSR policies and the financial situation. 

Thanks to machine learning techniques, we have built a classifier for each industry and each brand. Within our analyzes, we use the principal Artificial Intelligence and Machine Learning technologies, typical of digital consumer intelligence, from sentiment and emotion analysis to the use of automatic text classifiers, capable of reading, understanding the semantic level of sentences, assign meanings and classify by label.

A very important example can be applied to the Food & Beverage sector, as it is one of the most talked-about industries on the web. By applying our analyzes, our CML, and the work of our data quality trainers, we were able to analyze the online brand reputation of Tyson Food, managing to capture some crisis situations related to Corporate Social Responsibility.

Tyson Foods, Inc. is an American multinational corporation based in  Arkansas that operates in the food industry.

The company is the world’s second-largest processor and marketer of chicken, beef, and pork and annually exports the largest percentage of beef out of the United States. The company has a long history, full of brilliant innovation… but they’ve had their fair share of controversy, too.

They’ve been in the news recently due to dire warnings about the state of the country’s meat supply chain. The criticism of Tyson has extended to its treatment of its employees as well.

Through our dynamic market research tools, we have analyzed the whole web conversation on the Tyson Food world, in the last months. The goal was to understand how, by applying Custom Machine Learning techniques, it was possible to detect a corporate crisis, above all in which sector.

In almost 5 months, our monitoring has collected more than 1600 conversations and user-generated content on Tyson. We recorded a spike in conversations on July 8, the day when #Justice4TysonWorkers went viral. The web audience joined in asking for clarification on the death for coronavirus infection of many Tyson foods employees.

It was in fact easy through the tagging work of our data quality managers, to divide and label posts into different categories:

• Product: posts that talk about the product and all intangible services such as customer care and support, product/service quality or product/service judgment, after-sales, price perception, advice and chatting on the product, product/service comparisons.

• Governance: posts commenting on external and internal company policies, comments to CEOs, boards and management, partnerships, collaborations, events.

• CSR: posts concerning sustainability choices, scandals, geopolitics, environment, cultural, sporting, art etc.

• Innovation: Posts that propose new ideas about a product or that want to review a certain product.

Once the posts in those categories were separated, it will be applied to a sentiment analysis algorithm that will assign a value to each post. 

As we can see from the graph, there was a majority of posts on CSR compared to other categories. But…what happened? We can understand it from the emotions of the contents and from the cloud of the top words.

The conversations are split over two areas of the CSR: the situation of workers and the exploitation of animals.

There was a highly negative sentiment, with emotions of disapproval, regarding the attitudes towards workers and the exploitation of animals for the production.

Through our suite of dynamic market research and digital consumer intelligence it is possible to extract and analyze many useful insights, such as all the information on the audience that have talked about this topic. Which one?

• Socio-demographic: gender, age, location, language, consumers indexes, job.

• Behavioral: interest, entity, top mentions, hashtags used.

• Psychographers: personality traits / big 5.

Are you interested in learning more about the audience that chatted online on Tyson Foods? Download the report to have all the data available and learn more about the potential of our suite.