Using ChatGPT for Customer Sentiment Analysis


Caroline Gong

In prehistoric times when things like AI, ChatGPT, and models weren’t quite as hot, data collection and analysis was – and often still is – very manual. We’ve all probably seen massive datasets filled with hundreds of thousands of rows and thought, “Geez, how am I gonna get through all of this?” Sometimes, maybe a few Excel functions and tools could help. But especially for qualitative feedback, such as customer reviews, there is no way to know all the takeaways without reading them through yourself.

Now, though, the data analysis landscape has massively improved with the rise of AI! But don’t worry, you don’t need a computer science degree in order to keep up. If you’re interested in learning more about basic AI tools and how they can be used to unlock your customer insights, you’ve come to the right place. If you’re not as convinced, let’s see if I can change your mind.

Why would I even want to use AI in the first place?

Great question, and answers you shall receive. Here are three main reasons:

  • It’s fast as heck. Like, you will literally get information within seconds that would’ve taken you hours had you done everything by hand.
  • It can identify insights that will help you effectively prioritize and resolve customer concerns.
  • It’s easy to use. The tools we’ll look at will only require you to provide your customer data and ask the computer your most pressing questions.

Before I dive into those basic AI tools, let’s set up some basic vocabulary.


Customer Sentiment Analysis - the automated process of understanding how your customers feel about your brand and services/products (we’re going to teach you the beginnings of this!)

AI model - a program that detects patterns in a dataset. You won’t need to know how to build a model or how one even works, but just know that whenever you ask an AI tool for data analytics, the results you get will be from the observations of an AI model.

Prompt - a fancy schmancy term for “question.” Whenever you ask an AI tool a question or request, that is a prompt.

The AI Tool of All Time

You’ve probably heard its name. You’ve probably seen ads and headlines about it. You’ve probably even used it yourself…

Yes, we’re going to look at how ChatGPT can be applied to understanding your customers’ needs and wants! We’ll look at the capabilities of the free version and the paid 4.0 version as well.

Demo Set-Up

Pretend you're a new clothing shop owner. You want to create a minimalist athleisure brand that prioritizes comfort and quality above all else, and you’ve already launched three products – sweatpants, a sweatshirt, and leggings. You’ve gotten a few sales already, so you want to see the feedback.

I created a very rough and simple dataset that we’ll feed ChatGPT:

(For the purposes and intents of this demo, there are only 10 reviews and a few categories for each item, but in reality, your dataset will almost certainly be larger and more extensive.)

Let’s see what the free version of ChatGPT can do.

Free ChatGPT Demo

Step 1: Enter your data into ChatGPT. The free version only takes in “raw data,” or simply put, the copy-paste of your dataset. 

Immediately, this is the response ChatGPT gave me:

This list outlines ChatGPT’s capabilities. We’re going to ignore #4 because you won’t need this, especially if your brand doesn’t have a data or programming team.

Step 2: Let’s start easy and test the tool’s quantitative abilities. You can ask questions/prompts regarding basic metrics, like the mean/average, median, customer lifetime value, churn rate, and more. 

Since our sample dataset is quite simple, we’ll ask ChatGPT to give us the average rating of each product.

Hm, interesting. Seems like our sweatpants are doing pretty horrible.

Step 3: Ask follow-up questions! These can be open-ended – the whole point is that you can glean even more information regarding the performance and customer satisfaction of your services/products.

Focusing on the sweatpants, we need to figure out why the ratings for it are so low.

A few things to note. First of all, we know ChatGPT is able to look at the relevant rows, so that’s a relief. Second, it concludes with quite a few sentences that summarize its findings from the qualitative feedback. But especially if there are various different complaints, where do you start?

Great! This aligns strongly with your brand vision, and you know where to start. For instance, if product durability and quality is indeed the most pressing concern, you can start by looking at the materials and stitch of your products or the manufacturers that you’re working with.

Let’s try it again with another product.

Fantastic – looks like the sweatshirts are doing well, and there are a lot of things to love about it. Don’t forget that it’s equally important to understand the strengths of your brand!

Here’s a tough question for ChatGPT, though. Is there any way to apply what your customers love about the sweatshirts to your sweatpants?

If you haven’t noticed already, ChatGPT has a habit of going on for quite a bit, so make sure you ask for a clear-cut explanation if you can’t be bothered to read an essay! 

(By the way, here’s the link to this demo if you want to read through it yourself!)

Anyway, it seems in this example, it is possible for ChatGPT to draw conclusions and answer more open-ended questions. It interpreted and answered every prompt, and was able to give some actionable advice. All of these answers also only took a few seconds to generate at most.

ChatGPT 4.0 Demo

Using the same dataset, let’s see if ChatGPT 4.0 can do any better.

Step 1: Enter your data into ChatGPT. Unlike the free version, ChatGPT 4.0 allows users to directly upload their datasets, so just upload the right file and you can request information immediately!

Immediately, there is way more information, specifically quantitative and statistical observations. It also was able to tell that sweatshirts were the most well-received among customers and sweatpants not so much.

Step 2: Ask away! I wanted to ask more challenging questions, so I made up a hypothetical.

I made up prices for each product:

  • Sweatpants - $45
  • Sweatshirts - $45
  • Leggings - $35

The reason for this is because a lot of customers specifically said that the sweatpants didn’t seem to be worth the price. And one of the hardest challenges for businesses is to determine the right price point for its customers. Screw this up, and you might lose a good chunk of your customer base.

ChatGPT 4.0, in fact, gives me three options! Right now, I’m the most confused by #3, so I need a follow-up explanation.


(Link to demo.)

So not only is ChatGPT 4.0 able to answer more nuanced questions with multiple solutions, it can also reasonably and effectively justify its answers. That’s… pretty awesome! At the very least, it was able to provide a few ideas that we can play around with!

Small PSA: AI is Still a Work-In-Progress

ChatGPT in both its free and paid versions both worked. But obviously, the sample dataset we gave it was small and easy to work with.

Before we set you loose to test out the tools yourself, there are a few caveats about ChatGPT that you should be aware of too:

  • ChatGPT cannot process large amounts of data. If your dataset is humongous, it’ll either take the program forever to process or it simply might not even allow it.
  • Sentiment analysis in AI is still being developed. The free version specifically may not be able to dive deep enough to answer very layered questions and prompts.
  • You can’t create other automations based on your results. That’s a fancy way of saying that ChatGPT can’t watch your data change over time – you’ll have to re-upload the file every time. It also means that ChatGPT can’t really do much besides produce insights from your data. It won’t be able to predict or test or do anything really that goes beyond the scope of your dataset.

But again, these warnings are only applicable to ChatGPT. There are more advanced AI tools out there that do have the ability to dive deep and look at every nook and cranny of your dataset. For example, here at Cotera, we’ve built some pretty cool models that can help you segment your customer base based on their purchasing history and behaviors or when to best remind your customer to re-purchase from your brand.

If you’re interested in AI models and how they can help you deeply understand your customers, book a demo here to see ours in action!

Stay in touch

Not ready to reach out yet? Sign up for news on our latest product and content updates.