✨ Instant data analysis with AI (save hours a week)

Data analysis is a great use for AI. It can quickly go through big amounts of data, and providing a helpful and clear analysis with generated charts—pretty much instantly.

However, it’s often done wrong. It requires a different kind of approach compared to how you would normally do it. 

In this mini guide I’m going to cover how you can do data analysis with AI by turning any kind of data into a complete analysis—saving tons of times in the process. 

Tools: We'll be doing this analysis with ChatGPT Plus. If you don't have ChatGPT Plus, then Julius.ai is a great free alternative. It uses the same technology and is designed for data analysis.

Follow Along: You can use your own data or our example survey for practice that you can download here. In this file we have sample data of a customer satisfaction survey.

Objective: Our goal is to figure out how satisfied our customers are with the different levels of service - Basic, Plus, and Enterprise. We want to find out what they like and don't like, and figure out how to make them even happier.

Step 1: Data cleaning

To make data analysis work with AI you need clean data. If you upload a random file and expect a complete analysis, then it won’t work.

What clean data is to AI:

• Clear and consistent headers: The first row should contain headers, which are the names of each column. These names should be clear and consistent. 

• One row per record: Each row in the data file should represent a single record or data point (for example one customer). 

• Raw data: It should avoid any modifications like including subtotals, summaries, or cells with formulas in your file.

• Avoid Empty Rows and Columns: Empty rows or columns can cause issues during analysis. If a row or column is not needed, it’s better to remove it.

❌ This won’t work

✅ This will work

Most of the times when you export data from a tool (e.g. typeform) you already have the raw-data that already meets the criteria above.

If this is not the case then this is the only manual type of work you need to complete first.

Step 2: Data review

This is an important step in data analysis as it will help you spot trends, unusual bits, and interesting connections in your dataset. It will also tell if the AI can properly read and understand your data.

This is a common step for data scientists also known as Exploratory Data Analysis (EDA). Luckily we don’t really have to know how this works, we can simply instruct AI to do it for us.

Upload the file and prompt the AI with:

“Start with an exploratory data analysis of the file I’ve attached.”

You'll get a detailed summary of your survey data. This summary will show that the AI really understands your data, and it will highlight important findings, patterns, and features in your data.

If everything looks fine then we can proceed to the next part.

Step 3: Deep dive

With our data now clean and checked for quality, we're ready to start with the most exciting part: the deep dive analysis. This is where we start asking targeted questions to get to practical insights. 

To complete our objective to figure out how happy customers are with our different levels of service, we can start asking this question:

“How do satisfaction scores vary among Basic, Plus, and Enterprise tiers? Also visualize your findings.”

The satisfaction score for large enterprises is a lot lower than the other service tiers.

As pointed out by ChatGPT and seen in the chart it’s clear that customer service is the main problem that scores significantly lower than the rest.

Step 4: Analysis into action

Now that we have identified a problem, we can work on a solution. The great thing about ChatGPT is that it can go from data analyst to business consultant instantly.

To find a solution we can ask ChatGPT (within the same chat) on a solution:

“What ideas do you have to bring our satisfaction-score up within the enterprise tier?”

And as expected the main focus in on the customer service, the culprit that we identified with our analysis.

From this point onwards you can dive deeper into the solutions with ChatGPT, but that falls outside the scope of this short guide on data analysis.

Wrapping it up

This was one specific and short scenario, but AI can help you with all kinds of analysis. Just make sure you provide clean data that the AI can work with.

Also, be wary of hallucinations (incorrect answers) and ensure that it has a good understanding of your data. You can double check some of the findings before you share it with the rest of the world. But if done correctly, huge time saver—and it will only get better.

What would you like to see next?

While I try to feature one main piece in the newsletter, I’m also working hard on the new website filled with practical AI solutions. Tell me what you are missing and I’ll share it in the newsletter and on the website.

Leave your suggestion and I’ll work hard on to make it reality. I’ll even personally email you to let you know that your AI solution is live.

Which AI tools were released last week? Here are our favorite picks.

• AI Lawyer - Co-pilot for lawyers, instant legal help for consumers and companies.

• Guidde - The easiest way to create how-to guides and SOPs.

• Sketch Logo AI - Sketch it, let generative A.I. perfect it.

• ubique - Send personalised sales videos at scale.

• Podsqueeze - Automatically turn your podcasts into short clips, blog posts, social posts, and much more.

So… what did you think of this email?

You can add more feedback after choosing an option, or you can reply to this email directly.

Enjoyed this newsletter? Forward it to a friend and have them sign up here.

See you next Thursday!

— Jerry