Twitter is one of the largest microblogging social network of this decade and CEOs are increasingly flocking to this site. Who can blame them, given that this is where all the action happens - the latest news of the companies, marketing and leadership. Whether you are the CEO of a Fortune 100 company or a small business owner, your Tweets matter and it is an effective way to strengthen ties with the customers, build your brand image, network with influencers and crowdsource to learn and improve on products and services.
In this project, we analyze the tweets of various CEOs and try to ascertain the characteristics and tweeting styles that make them influential. We also look into customer engagement based on these tweets, and what topics help the company build their brand presence. We tie it all together by looking at whether these tweets offer any support for increased success in the company’s economic value.
The project was implemented in Python 3.7.3.
- tweepy
- spacy
- gensim
- nltk
- textblob
- Scraped Twitter accounts of 31 CEOs from different industries using the API. Also scraped daily stock data and financial statistics from Yahoo Finance for the companies whose CEOs we analyzed.
- Sentiment Analysis on the tweets for each CEO
- Topic Modeling using Latent Dirichlet Allocation for each CEO to determine their tweeting styles and a breakdown of their common topics
- Stock Price Analysis (Regression) using a combination of CEO attributes (age, compensation etc.), Tweet attributes (tweeting style/topics, sentiment scores, number of retweets and likes etc.) and Company attributes (company statistics, daily returns, volume of stock traded etc.)
- Engagement Analysis to determine what kind of topics garner more engagement from their customers using Logistic Regression
- The sentiments revealed that most of the CEOs' tweets are positive or neutral. The tweets that were classified as negative were general condolences to global tragedies.
- A closer look revealed that the sentiment analyzer was unable to distinguish between genuine and sarcastic tweets. For example, John Legere (T-Mobile) has a good number of sarcastic tweets directed towards his competitors but they were classified as positive.
- Topic Modeling revealed different tweeting styles of CEOs and how they maintain their online persona in the eyes of their customers and follwers. For example, Elon Musk has a very casual tweeting style and uses Twitter mainly as a marketing tool where he talks about his various products. On the other hand, Tim Cook makes a lot of statements about current events and clarifies what Apple is doing to help out. We also saw that he tries to connects with his followers by retweeting their pictures and captions it with the very popular hashtag #shotoniphone
- Some of the CEOs focus on tweeting about their business, whereas some CEOs don't use Twitter as an advertising platform. Instead, they talk about their personal interests and their take on social and global issues. These CEOs are using social media to establish an emotional connection with their followers and who by association can also connect to their brand.
- From our analysis, we saw that engagement and influence has an economic impact of the company. A single tweet has the potential to affect the stock market prices. The influential CEOs can leverage Twitter to turn the negative sentiments of the crowd to the positive ones and affect the market (Examples are given in the presentation).
- Engagement analysis showed that customers repond well when CEOs engage in conversations, apart from pushing their products or services. When CEOs use the platform to promote philanthropic efforts, impart life advice or share their personal interests, they tend to give the impression of being authentic and relatable, increasing their engagement.
- Using Linear Regression is not ideal in this scenario but it was a good place to start to discern if there was relation between tweets sentiments and the stock price.
- Extend this by building more complex predictive models using Time Series, ARIMA etc and incorporate social network analysis of the twitter users for influence and reputation.