Data is one of those things that’s crucial for any company, but also something that no one likes to deal with.

Well, that era is over.
Why?

Not only is data getting more and more plentiful, but analyzing and managing it well is getting increasingly crucial for any business. As we’ve discussed here) As a result, the role of business analysts is h5 significance.

At the same time, tools like ChatGPT (and especially the new code interpreter) greatly enhance data analysts’ capabilities.

So today, we’re going to dive into why analysts ought to embrace, and not fear AI, given how it augments, rather than supplants them.

Instead of discussing hypotheticals, we’re going to discuss tools that exist today, and how they can help your day to day workflow as an analyst, using a play by play framework.

We’ll have a look at an analyst’s daily workflow, see what inefficiencies these tasks within it may conventionally come with, and explain how these inefficiencies can be solved, using a practical lens. Furthermore, we’ll also have a look at new capabilities that analysts of different skill sets can leverage using AI tools h5 popularity today.

‍

A DAY IN THE LIFE OF A BUSINESS ANALYST

‍

‍Data Collection

A business analyst's day often commences with parsing through a multitude of emails alerting them to fresh batches of data to be analyzed. This could be transaction data, usage reports from APIs, or spreadsheets on customer feedback. ‍

Data Management

Manual data collection, though integral, tends to be quite inefficient. It involves individually connecting to each source, which ranges from executing custom scripts for databases to making API calls and manually extracting data from spreadsheets. These tasks can become particularly cumbersome when considering technical challenges such as occasional server downtime, access restrictions, or inconsistent data formatting across sources.‍

Data Validation

Upon successful data collection, analysts must validate the data's integrity, a step necessary for ensuring the accuracy of any subsequent analysis. This process involves checking for data anomalies, identifying missing values, and verifying data consistency.

Evidently, there’s a lot of tasks within any analysts workflow that, if made more efficient, would be a boon for the analysts and their teams alike.

Fortunately, though, we’re living in an era when there’s game-changing tools coming out seemingly every week-provided you actually stay plugged in, of course (but that’s what we’re for!)

So let’s have a look at some of the best tools you can use to make the aforementioned aspects of your workflow as a data analyst easier:

TOOLING CATEGORIES

1. Opening Tasks/Data Collection:

‍Levity.AI can significantly streamline and enhance the process of parsing through emails. Given how their data pipelines are particularly crucial for analysts and how crucial daily emails are for that, this tool can prove to be invaluable:

‍

• Data from emails can be automatically parsed and extracted, including the sender, subject, date, body, attachments, and links.

• Emails can be automatically categorized into different types (e.g., transaction data, usage reports, customer feedback) using Levity's email classification feature.
• Email summarization provides a concise overview of each email's content, enabling the analyst to get a quick understanding of the data without having to open and read through each email in detail.
• Levity’s email response generation can automate responses based on the email content and the analyst's preferences, thereby saving time spent on crafting responses to each data alert.

By integrating Levity into an analyst’s workflow, the initial phase of their day - sifting through their data pipeline- becomes drastically easier to handle. They can begin their day with an organized and prioritized list of data tasks, enabling them to focus more on actually deriving actionable insights from the data.

‍

2. Data Cleaning & Analysis With The ChatGPT Code Interpreter

If you’re reading this, I have to assume you know about ChatGPT. You may not know about the new Code Interpreter, or at least not know about how it can tangibly make your life as an analyst easier.

Let’s have a look!

Data Collection

Suppose you’re working with a relational database like MySQL, PostgreSQL, or SQLite. You’d typically need to manually connect to the database, write a SQL query to select the relevant data, execute the query, and then export the results for further analysis.

With the new ChatGPT Code Interpreter, you can write a Python script that automates this process.

Showing tends to be better than telling, so I figured i’d just go to the horse’s mouth and asked GPT4+Code Interpreter how it’d go about this. Here’s its answer:

Now a hypothetical framework is one thing.

How would using the tool look like in your everyday workflow as an analyst?

Churn is obviously an incredibly important metric for any product. So, being able to accurately predict which users are more likely to is as well.

Let’s see how the new Code Interpreter can help with that:

Churn Analysis & Prediction

Traditionally, predicting user churn at a place like Netflix or Spotify involves a series of fragmented steps.

Analysts might start by using SQL to extract data from a database, then switch to Python or R for cleaning the data and engineering features, and finally use a machine learning library or statistical software like SAS or SPSS to build and tune the predictive model. This disjointed process can be time-consuming and prone to errors due to the need to switch between different tools and languages.

Comparatively, the Code Interpreter can guide analysts through the entire process within a single, integrated Python-based workflow.

Using data from Kaggle’s ‘Telco Customer Churn’ dataset (hypothetical data about a telecommunications company), let’s see how the Code Interpreter can create a model to predict the likelihood of individual customer churn rates.

1: Code Interpreter identifies the various attributes within the data that are relevant at predicting churn:

‍2: Code Interpreter autonomously spots a discrepancy between how data is presented, and tweaks it to make it workable:

For the sake of brevity, I haven’t here, but if you click on ‘show work’, you can literally see the code the model has employed to accomplish this.

3: Code Interpreter demonstrates its ‘thought process’ behind which method of removal it opts for:

4: Code Interpreter performs Exploratory Data Analysis

‍The following is the ‘work’ the model used to accomplish the same:

5: Code Interpreter employs feature engineering

6: Code Interpreter builds its predictive model

7: Code Interpreter evaluates the viability of its own model:

Obviously, that’s quite a bit to go through. The thing is, it’s so intuitive that honestly I figured I might as well just show and not tell.

Here’s the TLDR:

• 1: Data Loading and Cleaning: Code Interpreter loaded the data into a Pandas DataFrame and cleaned it by handling missing values and ensuring correct data types.
• 2: Exploratory Data Analysis (EDA): Code Interpreter examined the distribution of the 'Churn' variable and some basic statistics of numerical variables.
• 3: Feature Engineering: Code Interpreter transformed categorical variables into a format suitable for machine learning models using one-hot encoding.
• 4: Model Building: Split the data into a training set and a test set. Then created a Logistic Regression model and trained it on the training data. ​
• 5: Model Evaluation: We used the trained model to predict the churn status of customers in the test set. We then compared these predictions to the actual churn statuses to calculate performance metrics such as accuracy, precision, recall, and the F1 score.
  

This entire process essentially resulted in a Logistic Regression model that can predict customer churn based on various features.

In addition to obviously automating several elements of an analysts' workflow, this sort of functionality (that is the worst it'll ever be) also drastically lowers the technical barrier for being an effective analyst as well.

‍

CONCLUSION

‍

As the Code Interpreter shows, AI can be massively helpful with analysts' workflows. The trick is with regards to how it's implemented; in a nutshell, the UX.
Crunch leverages AI and a nifty UX made possible by it to create what is essentially a co-pilot for analytics.

Want to see what it's like?

Signup for our waitlist here!

‍