163: How to Analyze Data in the New Era of AI

[00:00:00] Avery Smith-1: The AI revolution is here. We're witnessing one of the biggest shifts in modern history. AI is transforming how we work, how we make decisions, and how we understand the world through data. What used to take me hours in Excel, sequel and Python now can just take me 60 seconds with the power of ai.

[00:00:17] Avery Smith-1: Welcome to the new era of ai. In today's episode, I'm gonna show you what it's like to analyze data in the AI era and how you can analyze data the lazy way. By letting AI do all of the heavy lifting. So if you're listening to this, you either fall into one of two camps. One, you're already a data professional or working towards becoming one or two.

[00:00:36] Avery Smith-1: You're not a data professional, but you deal with data every single day. You work in marketing, finance, operations, HR product, or something similar like that, and you're drowning in spreadsheets, but not exactly sure how to extract meaningful insights from all the data either way. AI is your best friend.

[00:00:53] Avery Smith-1: If you're in camp one and you're already a data professional, AI isn't about replacing you. It's about 10 x-ing you. [00:01:00] It's a tool that'll allow you to automate repetitive tasks and free up time to focus on higher level work. If you're in camp two, as a knowledge worker who deals with a lot of spreadsheets, but struggles to pull meaningful insights.

[00:01:11] Avery Smith-1: AI is going to transform you. It's going to give you the ability to explore your data on your own in an easy and intuitive way. This is where data analysis, AI tools like Julius AI come into play, who happens to be today's episode sponsor. And to give you a real sense of how AI works in data analysis, I wanna walk you through an actual example using real data.

[00:01:32] Avery Smith-1: Using my data actually, so we'll be analyzing customer relationship management data or CRM data for short from my own educational platform, data Career Jumpstart, using Julius ai. In this example, I'll step into the role of a business development rep or some sort of a marketing professional, someone who needs to make sense of numbers and answer key business questions quickly.

[00:01:53] Avery Smith-1: Let's go ahead and get started. All right. This is my actual CRM and I use a platform called Course Creator [00:02:00] 360. It helps me send newsletters, host my course, and manage a bunch of other key parts of my business. It basically works the same way as more popular CRM tools that you've probably heard of, like Salesforce, HubSpot, or Pipedrive.

[00:02:12] Avery Smith-1: It stores all my contact data names, emails, tags, activity, history, everything. In total, I've collected over 20,000 contacts, so this data set's going to have about 20,000 rows. To be totally honest with you, it's kind of a hot mess. There are tons of inconsistent tags, unstandardized columns, random notes.

[00:02:31] Avery Smith-1: It's the kind of messy data set that makes you avoid even looking at it altogether because of how stressful it makes you. So let's go ahead and select all 21,000 contacts. Go to the export and head over to Julius ai. Now that we exported our data, all we need to do is upload our data by clicking on this file icon right now, selecting file upload.

[00:02:51] Avery Smith-1: Selecting the data that we have just exported. Boom. That simple. You can see that Julius is giving us a little preview of what our data set [00:03:00] actually looks like. It's given us a little summary over here on the right hand side of what our data includes. It's generated four preset analyses for this particular data set on its own.

[00:03:10] Avery Smith-1: And then of course, we can come down here and ask Julius to do all sorts of different data manipulation and data analysis. Using just plain English. So let's go ahead and get started. The first thing I'm going to do is move the name and email columns to the far right of the data table. And I'm doing this so I don't accidentally expose anyone's names or emails here on the episode, and I don't really wanna make my editor constantly blur everything like he has done previously.

[00:03:36] Avery Smith-1: So let's go ahead and do that. So it's actually really simple. I can just come down here to Julius and I can just ask in very simple terms here. Please move the first name, the last name, and the email columns to the far right of the data table, and then press enter and Julius will start thinking and working on my task.

[00:03:59] Avery Smith-1: You can see that it's [00:04:00] saying that it's going to do what I asked it to do. It is writing the Python code. To actually do that right now and then running the code down below. Okay. Julius just finished running all of its code and it says, perfect. I have successfully moved the first name, last name, and email columns to the far right, and it's given us preview of what that table looks like.

[00:04:16] Avery Smith-1: So far, so good. Also, just a quick note, we don't have to be worried about data privacy inside of Julius because Julius is SOC two type two certified, which basically means they've passed rigorous security audits and strict standards. Each user inside of Julius operates in a private sandbox environment.

[00:04:34] Avery Smith-1: They don't use any of the user data to train any models, and the underlying LLM models that they use like open AI and anthropic are contractually bound by the same rules as well, so we don't have to be worried about privacy issues. Okay, next I'm going to have Julius do some basic data exploration and data cleaning on the columns in this data set.

[00:04:52] Avery Smith-1: This data set has a bunch of columns. You can see that previously it told us that there are 53 columns right here. And to be honest, I [00:05:00] don't even know what all of them are or how useful they are or how much data they have. A lot of people have built columns on this data set and added different things, and I don't really know how many of them actually have valuable data.

[00:05:12] Avery Smith-1: So I'll simply go to the thread chat down here at the bottom, and I will ask a simple question. I'll say, how many of these. Columns are just empty and have nulls. Please identify all columns that are empty and I'll press enter and Julius will go to work for me and we'll fast forward through the thinking here, but basically it'll create a Python script and then tell me the results based off of the script it runs.

[00:05:37] Avery Smith-1: Okay, Julius just finished running and you can see that it found 18 completely empty columns listed right here. It looks like for whatever reason, we aren't using any of these, and there is a bunch that actually only have 5% filled out as well. So it looks like there's 22 columns with substantial data in them, which we can see kind of the null percentages right here.

[00:05:59] Avery Smith-1: On [00:06:00] the right hand side, so summary 18 columns are completely empty, 31 are nearly empty, and 22 columns have substantial data. That's great to know that really, instead of working with 51 columns or whatever it was, we only really need to be working on 22 of the columns. Okay. Next, and this is a great question for the marketing team, and.

[00:06:18] Avery Smith-1: Since I am a one man solopreneur business, that's me. I send out a lot of emails. I send out a weekly newsletter to these 21,000 contacts, and I send a bunch of other automated sequences to people who are inside my accelerator program and people who might be interested in my different courses. One of the things that's really key to sending emails is email deliverability, and this is basically how many times do you actually get in the inbox?

[00:06:41] Avery Smith-1: How many times do you get reported as spam or promotional, even if you're not spamming or promoting anything. And it's highly dependent on what email platform you're sending to. So for example, Gmail has different protocols and rules than maybe like Yahoo emails do. And so it'd be really cool to go through all of my contacts and figure out how many of them have Gmail accounts, how many of them have [00:07:00] Yahoo accounts, so on and so forth.

[00:07:01] Avery Smith-1: So once again, I'm gonna do it the lazy way and have Julius do it for me. So I'm going to say, go ahead and. Look at the email column. I want to know what email platform people are using. Please look at the second half of the email after the at sign. I. And create a bar chart showing the top 10 email platforms.

[00:07:29] Avery Smith-1: Please make the bar chart horizontal and include, uh, percentages on the chart. So I'm giving a little bit more of a complex task here, so it might take a little bit longer. We're not only doing the analysis, but I actually want the spit out visualization so I don't have to make the visualization myself.

[00:07:47] Avery Smith-1: So we're doing the analysis and the data viz in one prompt. Let's see how Julius does. And Julius just finished all of its Python code and it said to analyze which email platforms people are using, I extracted the domain from each email address, the part after the at [00:08:00] sign, and counted how many times each domain appears that I calculated the percentages of users for each platform and visualize the top 10 as a horizontal bar chart.

[00:08:06] Avery Smith-1: So as you can see, this is really important to me as a newsletter sender because really the majority of my contacts are using Gmail. So when I'm sending emails, I really need to optimize for Gmail's protocols and regulations because that's. The bulk share of my contacts in my CRM, so this is really useful to me that like, I don't really need to pay attention to Hotmail or Yahoo nearly as much.

[00:08:26] Avery Smith-1: And of course you can hit the download button and send this to your boss. Or you know, maybe I'll include this in one of my upcoming newsletters as well. Another important thing for me to check out as a newsletter sender is looking at how many names I have, because sometimes I just get people's emails with.

[00:08:40] Avery Smith-1: Out capturing their names. And this comes into play because sometimes I like to add a name variable in my newsletter, and that is the end receiver's name. And I try to do this to make it feel more personable, like I'm actually talking to someone one-on-one. So if you ever see me use your name in the newsletter, that's how I do it through this variable automation thing.

[00:08:59] Avery Smith-1: But even [00:09:00] though it is automated, I still hope it makes it feel more personable, and I hope it makes you feel loved and seen because you are. I love you guys, but let's go ahead and see how many names I have and how many names I have missing. Once again, this is extremely easy. All I need to do is go to the bottom of Julius here.

[00:09:15] Avery Smith-1: Say it in plain English. I'll say, let's do an analysis on names. I have a lot of missing names. What's the percentage? How many, first name only names do I have versus full versus no names at all? Press enter and once again, sit back, relax, grab some water. Actually, my water bottle's empty, so that was just a fake right there and let Julius do everything for me.

[00:09:45] Avery Smith-1: Boom. Julius is just finishing up right here and basically gave a breakdown for me. It looks like I have no name at 29%, and it looks like only about 48% of my contacts have [00:10:00] a full name and the remaining are missing. So it looks about 60% of my contacts I have names for, which is actually pretty good. That makes me feel more confident and like I should be using these personalizations more in the email than I am right now.

[00:10:12] Avery Smith-1: So this is exactly how this data from the AI is going to change what I do as a business user. Like it's going to inform what decisions I make when I send these newsletters. Pretty dang cool. Now obviously I don't personally know all 21,000 people on my email list. That would be impossible, but I wish I knew each and every one of you.

[00:10:31] Avery Smith-1: And let's go ahead and spend just like two minutes here to learn a little bit more about. Who is on my email list? Let's start by just getting to know your names. So Julius, I'm going to ask you a quick question here. Please create a top 10 chart of the most popular first names. I just wanna know what your name is and see if we have any common first names on this email list.

[00:10:55] Avery Smith-1: Let's see what happens. Alright, look at that, Julius spit out a count [00:11:00] of the top 10. With Michael having the most with 69. Daniel having the next most is 68. John, David, ham, Sai, Andrew, Alex, James, and Sarah. I'm glad we got at least one more female name in the top 10 because this was a pretty male dominant list here.

[00:11:17] Avery Smith-1: But super interesting. If you're listening to this and one of your names popped up, Michael, Daniel, John, David, ham, Andrew, Alex, James, or Sarah, please comment below, especially if you're on YouTube or Spotify. And I wanna say hello to you. Thank you so much for subscribing to the newsletter. It's great to meet you.

[00:11:32] Avery Smith-1: I also love that Julius mentioned down here that a lot of these names are more common Western names, but there's also names like Mohammad and Sai that are maybe more common around the world. I think that's just a fun little tidbit now that we know some of the most common names on my email list. What about engagement?

[00:11:46] Avery Smith-1: Because that's what we care as. People who send newsletters we're like, you know, are you engaging with our emails? And as a data analyst, I try to run my business and my newsletter as data-driven as I possibly can. So inside of my CRM, I've set up some cool automations [00:12:00] to try to keep track of how engaged you are with my newsletter.

[00:12:03] Avery Smith-1: So basically I give you points based on how many emails you open up for me and how many clicks you do inside of that email. And we call that metric engagement score or something similar. So let's go ahead and ask Julius down here. I wanna find who is the number one engaged person. So who is the most engaged person on my email list?

[00:12:25] Avery Smith-1: Show me their first and last name, please. And I'm really curious. Once again, I'm just gonna let Julius do its work. Sit back, relax, and uh, watch this robot work for me. Oh crap, you guys, this is embarrassing. Look who the most engaged person on the email list is. It's me. And that's true. I do open up a lot of my emails and check the links and everything to make sure it is working.

[00:12:49] Avery Smith-1: So I have a score of 2100. Okay. Uh, give me the top five. Now, the person who reads my own emails, that's okay. I'm exposed, but yeah, that's who I am. But let's see the top [00:13:00] five. Okay, here we go. Drum roll. Nice. So we got me number one. We got Isaac or as Sonya two. Um, that still is, uh, he is one of my employees, so that is a good employee Isaac way to keep reading my emails.

[00:13:12] Avery Smith-1: I appreciate it. We have Jen Hawkins, number three. I did an interview with her previously. She landed a data job at a FANG company. You can watch that episode. If you want to hear her full story, we'll have a link to it in the show notes down below, we have David Mills and MTA Pandy, who are two of my most engaged students inside my bootcamp, the accelerator program.

[00:13:32] Avery Smith-1: So shout out David and Muk. They're both looking for jobs right now. So if you're watching this in your hiring manager. Please consider them. They are awesome data analysts. Another thing that sales and marketing teams would be interested in is the sales cycle length, and if you've never heard of this term before, the sales cycle length is basically the number of days between the first interaction with a potential customer and the moment they complete a purchase or sign a contract.

[00:13:54] Avery Smith-1: For us, we'll consider that the day that they got on my email list and compare it to when they actually made a [00:14:00] purchase. So I went ahead and told Julius to compare the created date to the last purchase date. And yes, there are some spelling issues there. And what you'll see is that Julius went through and basically walked me through, uh, what was going on and.

[00:14:13] Avery Smith-1: After some back and forth. Basically, this is a pretty complicated task because this dataset has multiple columns of created and purchased dates because it's basically the blend of two different CRMs. So after some going back and forth with Julius, we were able to figure it all out and be able to create a metric to show, you know, when people purchased versus when they actually joined my email list and we were able to create this really beautiful histogram right here.

[00:14:40] Avery Smith-1: That basically lets me see that. Wow. You know, a lot of people are actually getting on my email list and purchasing almost immediately, and this is really interesting because it lets me know that like we're doing a good job when people are really interested of selling to them and or it could be because we are actually nurturing people for a long time.

[00:14:57] Avery Smith-1: But they use a different email when they sign up. I [00:15:00] see that happen, uh, quite often. So for example, some people maybe have an email where they get, you know, really important things and then maybe an email with only semi-important things. And when they end up signing up, they switch their email to the most important email.

[00:15:12] Avery Smith-1: So. This is a pretty interesting graph to look at my sales cycle and, you know, it gives me an idea that, wow, there are some people that take it takes over almost a year for them to end up purchasing, um, if they haven't purchased already. So this is pretty interesting data that I think would be valuable to anyone that's in sales or marketing.

[00:15:27] Avery Smith-1: Now imagine trying to do all of that manually, digging through all 21,000 rows, cleaning all those columns, coming up with all the formulas, making the charts. Heck, it would take hours if not days, to do all of that. But with ai, we got this done in about 20 minutes. I just typed what I wanted in plain English into Julius, and the insights came straight to me.

[00:15:47] Avery Smith-1: That is the power of working with ai, not against it. So now it's your turn. Go to julius.ai right now and upload a data set and try it out. We'll have a link in the show notes down [00:16:00] below for you guys to test out as well. Whether you're already a data professional or just someone who works with messy spreadsheets, AI tools like this are becoming a must, and this is just the beginning.

[00:16:10] Avery Smith-1: If you wanna be part of the next data analysis revolution, do not wait. Start exploring these tools now. Start practicing, and I would love to help. In fact, what would you like to see me do next with AI commented down below? And you might just see it in an upcoming episode. Good luck everyone, and welcome to the new era of data Analysis.