3 Ways "Advanced Data Analysis" Will Slash Your Data Works To Seconds

Hey there,

When ChatGPT's Code Interpreter was launched back in July '23 (feels so long ago), it promised to be a game-changer.

Meanwhile, it's been rebranded as "Advanced Data Analysis", and most people don't even have it on their radar.

That’s a huge opportunity missed. Because I believe it’s still one of the most underrated features.

Why? Let’s find out!

Want to learn how to 10x your data analysis productivity with ChatGPT?

Sign up for my event with O’Reilly below! Use this promo code to get free access to the platform and webinar for 30 days. Enjoy!

What ChatGPT's Advanced Data Analysis can do

Advanced Data Analysis (ADA - formerly called Code Interpreter) allows ChatGPT not only to write code, but also to execute it (and debug its own errors).

Some people called it "an eager junior programmer at your fingertips", which I think gives the right picture.

In many cases, however, people quickly found that it ran with a bunch of limitations:

• No internet connectivity

• Limited to Python

• Limited memory

It’s not yet the supercomputer we all hoped for.

But make no mistake - tools like this are paving the way for the next generation of AI.

As Andrej Karpathy recently mentioned, we're just seeing the beginning of a new era of LLM-powered operating systems:

Despite all the current limitations, one exciting use case opportunity stuck for now:

Analyzing data! (Which probably led to the new name.)

But we can’t just throw in our data mess and call it a day.

In fact, I found that Advanced Data Analysis works especially well for three common data scenarios:

Example 1: Streamline Exploratory Data Analysis (EDA)

Every analysis of a new data set starts with a so-called Exploratory Data Analysis (EDA). The goal is to find things in the data that surprise you: interesting relationships, outliers, distributions, etc. A thorough EDA helps us understand the characteristics of variables. Unfortunately, it can really be a tedious task to look at all the variables in your dataset. That's where ADA can assist you with EDA.

It works like this: Upload a file like XLSX or CSV to ADA and run a prompt that outlines the detailed steps for conducting an EDA (see this chat here for an example).

ChatGPT will write Python code to load the data, process it, and share its findings along the way, including visuals:

This process works pretty well out of the box in many scenarios. The good thing is that it doesn't require high quality data (like many other analytics use cases), because the goal here is to find flaws and errors in your data.

So it’s really a helpful tool.

Example 2: Get Actionable Insights from Transactional Data

If you want to go beyond descriptive statistics to insights, ADA can help as well. But the data has to be clean and in the right shape.

Unfortunately, in many cases getting this kind of data is the hardest part of all.

But there are scenarios where you can get (relatively) high quality data right out of the box.

The key is: transactional data.

Transactional data like:

• customer purchases

• web traffic data

• sensor readings

• etc.

Data like this can often be extracted from the source system itself, that is responsible for the data collection process - for example Google Analytics, POS systems, sensors, etc.

With the help of ADA we can quickly get high-level insights from this data without intensive modeling.

Just upload the data and prompt some key questions. No need for SQL, Python, etc.

For instance, an e-commerce manager could quickly understand which customers drive the most revenue for their store:

If you want to reproduce this example, check out this chat history and sample dataset.

Example 3: Enhancing Reporting with Dynamic Annotations

Business teams often rely on static reports or dashboards that lack nuanced context.

With ADA, you can quickly create annotations and footnotes to enhance your understanding of this data.

For example, a financial analyst could upload an Excel file with a balance sheet and ask ChatGPT to add commentary for the major asset or liability classes:

ADA can even write this commentary back into the Excel file:

You can find the full chat history here. I used publicly available balance sheet data from Apple, Inc. for this exercise.

This workflow also works well if you want to write explanatory notes tied to other types of reporting data, such as sales or marketing performance.

Limitations

While the applications for Advanced Data Analysis are really exciting, there are still some major caveats that you should be aware of:

• Uploading personal/sensitive data is prohibited: In most situations, you’re not allowed to upload company data to ChatGPT. This can be a problem for many uses, unless the data is already public. Currently, the only options are to build a system like ADA yourself, or use enterprise-ready solutions like ChatGPT for Enterprise, which provides a secure framework for data processing.

• Results are only as good as the quality of the data: Any use case where you want to get high-quality insights, requires high-quality, structured data. At a minimum, you should perform an EDA with data integrity checks before any further data analysis.

• Outputs should be reviewed for errors: Remember that this is still an augmented scenario. You're the one in the driver's seat who's ultimately responsible for the results. So make sure to double-check your work before jumping to any wrong conclusions.

Conclusions

In each of these examples, Advanced Data Analysis unlocks tremendous time savings while empowering more users to leverage data.

The key is blending it in a given business process, while being aware of its limitations.

Finding the right balance may take time, but it's definitely worth the effort.

What we're seeing now is a glimpse of the future of data analysis. Keep in mind that it's still an early beta release.

I hope this gives you some ideas on how to maximize the value of ChatGPT's capabilities!

Let me know if you have any other questions.

See you next Friday,

Tobias

Want to learn more?

1. Book a meeting: Let's find out how I can help you over a coffee chat (use code FREEFLOW to book the call for free).

2. Read my book: Improve your AI/ML skills and apply them to real-world use cases with AI-Powered Business Intelligence (O'Reilly).

3. Follow me: I regularly share free content on LinkedIn and Twitter.