Democratizing Analytics: A ChatGPT Plus Case Study

For years, there have been no code offerings in the realm of artificial intelligence (AI) and machine learning (ML) like RapidMiner, KNIME, and Dataiku. However, with the launch of OpenAI's ChatGPT Plus in February 20231, the landscape changed dramatically. Professionals from various industries can now utilize the features of ChatGPT Plus for just $20 per month, and the best part is there's minimal learning curve involved. This unprecedented ease-of-use introduced by ChatGPT Plus truly marks the onset of "Democratized Analytics," opening doors for a fresh wave of minds to participate and innovate.

Advanced Data Analysis Tool: A Playground for Data Lovers

If you work with data in any way and you haven't tried this feature yet, you're missing out big time. ChatGPT Plus comes with an Advanced Data Analysis Tool that doesn't just chat—it computes! With the power of Python at your fingertips, you can upload files, download results, and do so much more.

ChatGPT Plus Advanced Data Analysis Tool in Action–A Business Use Case

I wanted to show my colleagues a "real" business use case that highlights the raw power of ChatGPT Plus. But there was a hurdle—what data to use?

Data Sensitivity: A Common Concern

One significant issue when using ChatGPT Plus for business purposes is the need to avoid sharing any sensitive or personally identifiable information. This is a concern of those introducing any cloud-based tools into a business setting.

It occurred to me that I could also use ChatGPT Plus to generate a synthetic data set (artificial data). This allows the demonstration of its capabilities without compromising data privacy. There was an option to use public government data, which comes with its own drawbacks, but a synthetic data set seemed like a lot more fun.

Creating Synthetic Data

I wanted a data set that would speak to all stakeholders. Another criterion for the data set is its belonging to a vertical different from that of my clients’ in the financial services industry, so it would not distract from what the tool can really do. This led me to data sets within the travel industry—hotel bookings and cancellations.

My first step was a simple one: I created a CSV file with just headers. But get this—in hindsight, I realized ChatGPT Plus could have created the headers for me too! I later tested ChatGPT Plus’ header generation and the tool surpassed my expectations. It created not just realistic headers but also fake names and email addresses for each record. The ability to generate table elements earlier in my career would have been invaluable!

Project Details

Once the synthetic data set was prepared, my next steps were to upload it into the tool. The following prompt was also provided to the tool to provide it with context and guidance:

“Attached is a CSV file with column headers for hotel booking and cancellation data. Please generate realistic sample data for all the columns, totaling 12,500 records. Have the data reflect bookings over the past 2 years. As you populate the dataset, ensure that the data reflects trends and patterns for a hotel in Orlando, Florida, and its relevant market dynamics such as seasonality and competition. This particular hotel has an unusual number of late cancellations. Make sure that is reflected in the data.

Within the data, create four trends in the data that illustrate strong relationships between different data values. These patterns should be credible and could likely be uncovered during an analysis of the hotel's actual reservation data. Ensure one of the trends is surprising, and potentially uncommon, but not unbelievable. One of the relationships should be tied to the late cancellations. Some value in the data should be strongly correlated with the late cancellations, this correlation should be both interesting, unique but also believable.”

ChatGPT Plus produced a step-by-step overview of its process to generate the request:

Here is the LINK to the full prompt thread. The tool even shared the Python code and the results generated at each stage. Since the links in the thread don't show the Python code, results, or the graphs, these are included at the end of this article for you to marvel at. Spoiler alert: ChatGPT created a PowerPoint presentation documenting the use case with minimal human intervention.

It's worth mentioning that like any tool, ChatGPT Plus may occasionally overlook a detail or make an error. However, what sets this tool apart is its self-awareness. The tool can often recognize its own mistakes, notify you, and make real-time corrections.

Here’s an example of this self-correction feature from a separate project. In this case, ChatGPT Plus autonomously identified an error it made and promptly reprocessed the task:

While ChatGPT Plus may not be flawless, its adaptability and self-correction features add a layer of reliability that makes it an invaluable tool for various projects. Much like humans, it's continuously learning and evolving.

In my usual workflow when using ChatGPT Plus for data analysis, I generally start by prompting the tool to describe the provided data set. I then guide it through exploratory data analysis (EDA) tasks and delve into modeling. Finally, I present the business problem and ask for potential solutions. This structured approach has often led to a more in-depth and insightful analysis—though, it could be serendipity at play.

A Simplified Approach

For the hotel bookings project, I opted for a more streamlined method to keep things straightforward. I began by initiating a new chat session with ChatGPT Plus, uploaded the CSV file, and entered the following prompt:

"Our hotel is grappling with the issue of last-minute guest cancellations. These not only make it hard to rebook rooms without slashing prices but also disrupt our staff and inventory management.

Please scrutinize the attached hotel booking data. Run a full EDA and run analytical models to pinpoint trends affecting bookings, cancellations, and their subsequent impact on operations and profits. Your analysis will guide us in formulating strategies to lessen the financial toll of cancellations and enhance profitability.

?Specifically, highlight any actionable trends or customer segments that could help improve room occupancy, sales, and reduce cancellations. We welcome all insights that can drive bookings and revenues.

Also, incorporate relevant charts or graphs to support your analysis.

Your findings will be compiled into a presentation for executive management. The presentation will articulate the data-driven strategies and policies designed to mitigate cancellations and maximize profits.

The structure of your analysis and summary should be as follows:

1. Business Problem Overview and Solution Approach: Define the problem and your methodology.
2. EDA Results: Present the results of Exploratory Data Analysis.
3. Data Preprocessing: Explain any cleaning or transformation of data.
4. Model Performance Summary: Evaluate the accuracy and effectiveness of your models.
5. Strategies and Policy Recommendations to Reduce Cancellations: Offer actionable strategies to cut down on cancellations.
6. Strategies and Policy Recommendations to Minimize the Impact of Cancellations on Profits: Suggest ways to cushion the financial blow of cancellations.
7. Strategies and Policy Recommendations to Maximize Profits: Advise on overall profit maximization.
8. Summary: Recap the main points.
9. Next Steps: Outline future directions for further analysis or implementation.
10. Appendix: Include a data dictionary and any other pertinent information not covered in the above sections."

Anything I write here cannot illustrate how amazing ChatGPT Plus' response was better than a LINK to the actual chat.

Minor Tweaks, Major Outcomes

The only follow-up prompts needed were a few affirmative responses and minor adjustment request prompts for the PowerPoint presentation that ChatGPT Plus generated.

Upon downloading the generated PowerPoint presentation, I opened the file and used the Designer tool for slide enhancements. A few clicks and three minutes later, I incorporated screenshots of the graphs and descriptions from the original ChatGPT Plus thread into the presentation. While it wasn’t quite presentation-ready, the presentation was a solid start achieved in under 20 minutes. Here’s a copy LINK. You be the judge.

Comparing ChatGPT Plus with Other AI Presentation Tools and Plugins

While there are numerous AI-driven presentation tools and ChatGPT plugins available, none have met my criteria for quickly creating a nearly finished product. Many tools, especially business productivity plugins for ChatGPT, serve more as advertisements for other software rather than practical tools.

Limitations and Cautions

Is ChatGPT Plus impressive? Unquestionably. Could it replace a seasoned data scientist or marketer? Not likely. It's essential to remember that while the tool is reliable, it's not infallible. Any data or conclusions should be vetted thoroughly before making business decisions.

Practical Business Use Cases

If sharing sensitive data with ChatGPT is a concern, companies could consider enterprise-level solutions and policies that secure data. In such an environment, a junior data analyst could:

• Begin data cleansing.
• Conduct basic Exploratory Data Analysis (EDA) and data preprocessing.
• Review and augment the tool's conclusions and recommendations.

Subsequently, a more experienced data scientist or marketer could validate the findings and take the project to the next level.

Additionally, ChatGPT Plus has the following additional applications:

• Training data: Useful for training junior analysts and marketers.
• Idea generation: Use ChatGPT Plus to create synthetic data sets that can serve as a basis for brainstorming.
• Testing and auditing: Employ synthetic data to test new application logic or auditing processes.

Concluding Thoughts

I hope this use case has been insightful. ChatGPT Plus' Advanced Data Analysis Tool is versatile and capable, offering a wide range of applications for businesses and individuals alike. I’d love to hear your own experiences and creative uses for this tool.

Published by John Minze, Director of Client Solutions at Intelisent

John possesses extensive leadership experience from top-tier advertising, financial services, and production agencies. He specializes in crafting comprehensive solutions for clients that encompass integrated data-driven assessments, process enhancements, campaign strategies, creative executions, and print/digital production. John is eager to explore the latest artificial intelligence technologies and offer innovative solutions to his clients.

Sources

1Open AI. (2023, February 1). Introducing ChatGPT Plus. https://openai.com/blog/chatgpt-plus