The Umbrex Guide to

Getting Started With

Advanced Data Analysis in ChatGPT
A video, self-guided workbook, sample data set to
help you master the basics in under an hour

Source:_______
Copyright © 2023 by Adam Braff and Will Bachman

All rights reserved. No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording,
or other electronic or mechanical methods, without the prior written permission of the publisher, except in the case of brief quotations embodied in critical
reviews and certain other noncommercial uses permitted by copyright law.

First Edition

Published by Umbrex, Astoria, NY

ISBN: 978-1-961779-13-6

Visit our website at www.umbrex.com

Book design by Umbrex Consulting LLC

Printed in the United States of America

This book is sold subject to the condition that it shall not, by way of trade or otherwise, be lent, re-sold, hired out, or otherwise circulated without the
publisher's prior consent in any form of binding or cover other than that in which it is published and without a similar condition including this condition being
imposed on the subsequent purchaser.
The Umbrex Guide to Getting Started with Advanced Data Analysis in ChatGPT

In this guide you will learn the following skills

§ Enable Advance Data Analysis (ADA) in ChatGPT

§ Upload files to ChatGPT

§ Use ADA to characterize data

§ Use ADA to perform trend analyses and predictions

§ Merge files using ADA and download new files

§ Detect and correct mistakes made by ADA

§ Create graphs and custom visualizations

§ Get Python code from ADA

§ Reproduce the work in Excel

§ Source other data sets to use

Four components to this guide
This workbook Video

File #1 File #2

UNRATE.csv
Patagonia and TNF jacket search.csv
DOWNLOAD
DOWNLOAD
Directions for using this guide

Throughout this workbook you will see action items in red. The key for these actions are as follows:

A red arrow will show you where to click or download.

A red circle indicates an area to select or click.

A red box indicates areas where you should type your text into ChatGPT.

This workbook contains alternating slides:

§ ACTION: The action you should take to obtain a certain result from ADA.

§ OUTPUT: The results you should receive from ADA.

At the bottom of each ACTION page you will see a box that gives a link to the specific moment of the video when Adam
demonstrates that action, if you’d like to skip to certain tutorials.
1. Activate Beta features in ChatGPT-4

Upgrade to a Plus
Sign up for or log Account ($20/month) to
into ChatGPT access Beta features

To see Adam Braff demonstrate this in the workshop, click here to go to minute 16:12 in the video.
Click on your username in On the pop-up screen click “Beta
the bottom left corner and Features” and toggle on
select “Settings & Beta.” “Advanced data analysis.”
2. Log into ChatGPT-4 Advanced Data Analytics

When you log into ChatGPT,

be sure that GPT-4 is
selected at the top, and
Advanced Data Analytics is
checked.
3. Upload the example .csv file
Click the plus sign in
the “Send a message”
box at the bottom of
the ChatGPT screen.

Select the “Patagonia

and TNF jacket
search.csv” file from
your computer, then
click the arrow to
upload it.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 16:43 in the video.
4. ACTION: Ask ADA to describe the data

To see Adam Braff demonstrate this in the workshop, click here to go to minute 17:24 in the video.
5. OUTPUT: Dataset details

You should get something

that resembles this output
although it might differ
slightly.
6. ACTION: Ask ADA to show the trend over time

This is a fairly open-ended

request. ADA might respond
in a variety of ways.

You could choose to be more

specific in your request, or
you could let ADA take the
initiative as we’ve done here.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 19:32 in the video.
7. OUTPUT: Trend over time graph

You should get something

that resembles this output
although it might differ
slightly.
8. ACTION: Ask for a customized trend graph
You can request graphs
showing trends over
time, with specific
parameters.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 22:05 in the video.
9. OUTPUT: Customized trend over time graph

You should get

something that
resembles this
output although it
might differ slightly.
10. ACTION: Download a clean interim data file
It’s helpful to periodically ask
to download the interim data
created by ADA, especially if
it follows complex cleaning
steps and looks correct.

CLICK TO DOWNLOAD

To see Adam Braff demonstrate this in the workshop, click here to go to minute 23:09 in the video.
11. OUTPUT: The new .csv file

You should get

something that
resembles this output
although it might differ
slightly.
12. ACTION: Watch for mistakes

Type your
request.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 23:45 in the video.
13. OUTPUT: Incorrect results

ADA can make mistakes

(e.g., by failing to sort
before graphing), as it
did here.
14. ACTION: Fix mistakes

You can ask ADA to correct

its mistake by pointing out
what was done wrong and
asking for new data.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 24:45 in the video.
15. OUTPUT: Corrected data

You should get

something that
resembles this output
although it might differ
slightly.
16. ACTION: Ask for an area graph

You can request different

types of graphs, such as an
area graph of market share.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 29:38 in the video.
17. OUTPUT: Market share area graph
18. ACTION: Customize data visualization options

You can ask for different

visualizations, including
specific color schemes or
output types, by asking in
plain English.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 30:54 in the video.
19. OUTPUT: Updated visualization

The result should be

something that matches
the visualization you
requested.
20. ACTION: Ask for a seasonality search graph

To see Adam Braff demonstrate this in the workshop, click here to go to minute 32:36 in the video.
21. OUTPUT: Seasonality search graph
22. ACTION: Ask ADA to explain the Python code

If you plan to use Python to

analyze or operationalize the
data later, you can ask for
explanation of the code as a
kind of high-context tutorial.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 34:49 in the video.
23. OUTPUT: Detailed Python code instructions
24. ACTION: Ask ADA to decompose the trend for prediction

You can ask ADA to plot a

decomposition of trends
and predict trends.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 36:11 in the video.
25. OUTPUT: Trend decomposition results
26. ACTION: Teach ADA to reproduce the work in Excel

You can ask for the same

kind of high-context
tutorial for Excel.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 39:07 in the video.
27. OUTPUT: Detailed Excel instructions
28. ACTION: Ask ADA to predict and plot the trend

To see Adam Braff demonstrate this in the workshop, click here to go to minute 42:08 in the video.
29. OUTPUT: Forecast graph
30. ACTION: Ask ADA for simple explanations

Explain like I’m [insert

age or sophistication level
desired].

To see Adam Braff demonstrate this in the workshop, click here to go to minute 45:42 in the video.
31. OUTPUT: Explanation
32. ACTION: Ask ADA for more technical explanations

The explanation can be

simple, as in the prior
example, or more
technical, as in this
example.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 46:37 in the video.
33. OUTPUT: Explanation
34. ACTION: Request ideas for other data sets to join

You can ask ADA to

suggest other data sets to
predict trends.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 47:59 in the video.
35. OUTPUT: Suggested data sets
36. ACTION: Request sources for other data sets to join

You can also ask for a

source for these data
sets, including links.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 49:00 in the video.
37. OUTPUT: Data set sources
38. ACTION: Use sources to find another data set for comparison

Using the “U.S. Federal

Reserve Economic Data
(FRED)” from the
previous sources
provided, click the link
to open FRED in a new
tab, and search for
monthly unemployment.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 16:12 in the video.
 We’ll use the
Unemployment
Rate provided as
a search result
from FRED.
 Clicking the
“Unemployment Rate”
result gives us this
graph, and we can
download this data.

CLICK TO DOWNLOAD
39. OUTPUT: Downloaded UNRATE.csv file

We have now
downloaded
and saved the
file to our
computer as
UNRATE.csv.
40. ACTION: Upload the new data set (UNRATE.csv) for comparison

You can use multiple data sets

in ADA at once, although it can
get confused if there are many
transformations of both.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 49:52 in the video.
41. OUTPUT: Merged comparison data sets

ADA might detect an

issue and fix it, as it
did here.
ADA might provide a
structure for the
merged data. If not,
you can request it.
42. ACTION: Ask for a comparison graph

Using multiple data sets, you

can ask ADA to create
comparison graphs with
specific instructions.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 50:43 in the video.
43. OUTPUT: Comparison graph
44. ACTION: Request a summary

ChatGPT can summarize the

analysis done within the chat
session—but won’t necessarily
remember “all the way back”
(limited by number of tokens it
can hold in memory).

To see Adam Braff demonstrate this in the workshop, click here to go to minute 52:00 in the video.
45. OUTPUT: Executive summary

This is a weak attempt,

lacking substantive
answers. You can ask
ADA to elaborate.
46. ACTION: Ask for a more detailed summary

Ask ADA to provide a

more detailed summary.
47. OUTPUT: Detailed executive summary
48. ACTION: Ask ADA to auto-write a Python program

If you want to create an

interactive tool that
works with the data
(including the data as
cleaned and joined by
ADA), just ask.

To see Adam Braff demonstrate this in the workshop, click here to go to minute 53:03 in the video.
49. OUTPUT: Basic interactive Python program

You can go through

many rounds of
iterating on the code
to obtain the desired
results.
More resources available from Umbrex
How to find clients for your consulting practice The top 500+ LinkedIn content creators

How to build a consulting website How to name your consulting firm

PowerPoint templates & examples How to improve the client experience

Guide to serving on a board of directors How to start a podcast

Guide to CRM software for consultants How to create a newsletter

Business & correspondence cards Executive handover playbook

Best practices to optimize your LinkedIn profile Library of industry-specific KPIs

How to set consulting fees Library of functional KPIs

How to solicit client feedback Email deliverability playbook

How to get business insurance RFP playbook

Invoice tips & template Mental math reference guide

Consulting contract tips & templates Voice of the customer interview guide

Best practices for working with staffing firms B2B survey question bank

Business coaches for consultants PESTEL analysis playbook

Go to the Umbrex Resource Library

 THE
ANALYTICS
INTENSIVE
AI and data literacy training
for business enterprises and investors
Why the Analytics Intensive?

■ Business analytics, including big data and AI, is a domain that continually presents
new opportunities as well as new challenges
– Opportunities include the ability to answer new kinds of questions (such as
understanding your competitors’ tactics and success) with new kinds of data
– Challenges include slow pace (having the answers locked up in one silo of the
org), confusion (misinterpreting data), and losing ground in AI
■ You might think the solution is to hire many more data scientists—but in many
industries that’s expensive or impractical, and it turns out it’s not even required
■ The solution is the Analytics Intensive, an AI and data literacy course for businesses

1
The opportunity for big data analytics is
growing as sectors learn from one another

Public equity
• Business investors • All of that, plus
intelligence • Nowcasting and • Target
• Improving forecasting screening
performance • Portfolio • Due diligence
monitoring
Private equity and
Corporates
venture capital

2
Business analysis is enhanced by an
increasingly diverse range of data sources
Questions to be
I E E solved with
What are my revenues What is the overall size What promotional
and costs by region? of our market? actions are my Internal data

competitors taking?
External data

I E E
What operational metrics How is my share changing Are my competitors’
are driving them? over time? actions working?

E E
Do I have any unforeseen What are sales in an
I adjacent product category
competitors coming up
How are they trending strongly? or geo that we might
over time? enter?

3
But implementing an analytics strategy
creates new challenges
What is your strategy for sustaining a competitive advantage in using data and analytics?

Are you able to get the answers to analytical questions quickly enough to take action?

How are you incorporating the latest advances in generative AI into your analytical strategy?

What alternative and external data sets are you using today?

What are the next analytical projects on your wish list, ranked by impact and feasibility?

What could you do if you had many more AI- and data-literate colleagues?

4
You can solve this by hiring more triple-skilled
data scientists…

Domain Most knowledge workers have some combination of

knowledge the three skills, with data scientists in the center

Statistical Coding
acumen ability

5
…Or by cross-training the (many other) people in
your organization to fill in their skill gaps

Show how to link analyses

to business outcomes

Build
Teach what makes
comfort with
an analysis valid
big data
and how to
sources and
visualize it
tools

6
The Analytics Intensive

AI and data literacy for business enterprises and investors

Topics taught in Adam Braff’s Brown and NYU graduate classes

6 modules over 6 weeks, taught in hybrid format

7
 Practical knowledge—in 6 modules

Module What you’ll learn

1 Problem solving with data How to define and solve problems as the top consulting firms do

2 Alternative data How to use the data sets that give big hedge funds an edge
3 Data visualization How to make your analyses instantly understandable and powerful
4 Practical forecasting How to predict key events with a useful range of probabilities
5 Generative AI for analytics How to use ChatGPT with plug-ins to accelerate and deepen data work
6 Persuading with data How to tell your analytical story in a way that will influence others to act

8
Customizable content and format

Default 6-week format Customization options

(not exhaustive)

Sessions 2–5 held Remove any of modules 2–5 to

Sessions 1 & 6: Live
over 4 weeks @ shorten course
in person (on site or
2.5 hours each
offsite), half-day each
(remote)
Hold all sessions live in person
Students work in
Up to 20 students
teams on relevant
per cohort, exclusive Incorporate a case study using your
final project, to
to your enterprise enterprise data (under NDA)
present in session 6

9
Who is the Analytics Intensive for?

Senior executives championing AI & data efforts

• No coding or advanced statistical knowledge needed

‘Analytical but not technical’ employees & managers

• Across finance, marketing, operations, HR, and other functions

‘Technical but not business-y’ data science colleagues

• Who would benefit from understanding analytics-business link

Pace and content are adapted to the specific people in the cohort

10
Practical, interactive exercises—not
boring lectures
■ Individual diagnostic to understand
each participant’s ‘type’
■ Hands-on work with Excel, Tableau,
and ChatGPT
■ Real data about real companies
(including, optionally, your enterprise’s
data)
■ Forecasting contest
■ Work performed both individually and
in teams
■ Final presentations made by teams of
students, with actionable feedback

11
About the instructor
■ More than two decades of data & analytics
Quotes from Braff’s students in course evals
leadership roles (McKinsey, JPMorgan Chase,
DirecTV, Zurich Insurance, Point72)
■ Extensive teaching experience in wide range of His suggestions are super
contexts useful and thought-
The environment provoking; one of the best
– Data & analytics “boot camps” taught to of Prof. Braff’s professors.
many different business functions, from IT to class is great! We
operations to investors have respect for
each other.
– Classroom teaching experience 2019–
present in graduate schools at Brown and
NYU; has created and taught 5 different I liked that the professor
course syllabi related to business analytics Class materials are was enthusiastic and
engaging.
– In consulting: led many McKinsey training designed to be fun
courses, including Basic Consulting to interact with.
Readiness, EM Basics, and Big Data class at
EM College in Cambridge Very practical and
useful skills.
■ Author of Business Analytics: A Digital Textbook I think it’s a perfect Everything is clear
and “bookie” of annual forecasting contest class! and organized.

12
Learn more about the Analytics Intensive

https://braff.co/intensive

13