I finished writing my book on sentiment analysis in fintech marketing last week. It’s a 13 page long eBook. It’s about using sentiment analysis to predict the performance of fintech posts on platforms like Substack, Medium, and LinkedIn. The eBook contains several experiments where I used sentiment polarity, subjectivity, and word count to predict engagement.
The results were less impressive than early experiments suggested. I was setting the y-intercept for the linear regression formula to zero in earlier experiments so I was getting R squared scores that were much too high. After I stopped doing that, many of the experiments gave me R squared scores in the 0.10 to 0.20 range instead.
So the linear regression formula wasn’t as effective at using sentiment analysis to predict the performance of posts as I had hoped. But the original results were unrealistic. One of the reasons why I went back and rechecked the formula was that I read an article explaining that major ad agencies normally don’t get an R squared score above 0.10 when they analyze their campaigns.
I still think these results show that sentiment analysis is useful for identifying the posts that will get the most engagement. It could also be used in a linear regression model along with other factors and I plan to test that later. Word count improved the accuracy of the model in several experiments. I’m also considering other factors such as the grade level of the text. Simplifying your writing makes it easier to read and lots of marketers and journalists recommend doing that.
Benefits of Doing This Project
I also improved my Python programming skills by doing this project. Originally I was cutting and pasting posts into my sentiment analyzer. But that was taking a lot of time and it reduced the size of the experiments I could carry out. So I eventually started using Beautiful Soup to scrape the posts and feed them directly into the sentiment analysis software.
Similar SaaS tools and apps often use a pay-per-use model. You have to pay more money if you want to carry out larger-scale projects. And some of these apps are designed for business users and not for individuals, so they can have high fees. By using free Python libraries like Beautiful Soup and TextBlob, you can analyze more information without incurring extra costs.
The tutorials that explain how to do this are considered beginner level. The biggest challenges were solved already because other programmers built the Python libraries that perform the analysis and handle the scraping. The tools I’m using are widely used for research and if you look around you’ll see other papers where the researchers are using TextBlob and Beautiful Soup instead of developing their own libraries.
Another reason why this is important is that the technical SEO experts who I follow on LinkedIn and Twitter have been performing experiments with similar technology. They also use Python and spreadsheet macros to analyze large collections of web pages and documents. Ahrefs is even set up to export data in CSV, a format that’s similar to the data I was analyzing in OpenOffice for this project.
I did the analysis on my laptop and not my smartphone. But it is possible to install Python libraries on a smartphone and run this type of analysis using apps like Pythonista. I’ve done similar types of text analysis projects with my smartphone before. That may be useful because certain websites will show you different results when you visit them with a smartphone instead of a laptop.
The Purpose of the Book
As for the book itself, over the past few days I’ve shared it with several writers and marketers so they can review it. I’ve also been posting about it on LinkedIn and those posts attracted a different audience that was mostly programmers. In general, though, it’s meant to show fintech founders a way to attract larger audiences for their writing using a method that’s not widely covered by other marketing books and courses.
The book explains a way to make fintech articles more exciting and interesting to read. It’s not about making fluff posts or discussing unrelated topics to bring more visitors to your website. The idea is that you can still write about things like payment processing and money transfer apps while optimizing factors like sentiment and word count.
That’s important if you want to target fintech keywords near the bottom of the marketing funnel. People who search for those terms are ready to sign up for a fintech app. But when you’re targeting those keywords, you’ll get readers who aren’t interested in the broader topics that attract the most visitors from Google. They want articles that answer specific questions about the app. So it’s important to find other ways to make those articles more entertaining to read.
Where You Can Get the Book
I plan to make the book available on my website as a lead generator soon. The idea is that you give me your email address and I’ll send you the book. And I’ll use the email addresses I collect to send subscribers other information, such as newsletters about other research projects, later. I think it’s possible to create a more accurate linear regression model by adding more metrics. The model will use natural language processing to predict engagement for fintech articles, posts, landing pages, newsletters, and other content.