When learning data science and machine learning techniques, you need to work on a data set. Matt Harrison had a great idea: Why not use your own Twitter analytics data? So, he did that with his own data, and shares what he learned in this episode, including some of his secrets to gaining followers.

Transcript for episode 159 of the Test & Code Podcast

This transcript starts as an auto generated transcript.
PRs welcome if you want to help fix any errors.

00:00:00 When learning data science and machine learning techniques, you need to work with a data set. Matt Harrison had a great idea. Why not use your own Twitter analytics data? So he did that with his own data. And he shares what he learned in this episode, including some of his secrets to gaining followers.

00:00:31 Welcome to Test and Codecode.

00:00:41 Today on Test and Code, I’m thrilled again to have Matt Harrison. So welcome, Matt. Thanks.

00:00:48 Happy to be here.

00:00:49 Matt Harrison has been on the show before. I’m just looking it up in January of 2019. Wow. That was a long time ago. Episode 63, we talked to Matt about corporate training for Python. And then near the end of 2020, we talked to them in episode 137. Oh, you interviewed me for your book because you have a book authoring course, right?

00:01:17 Yup. Yeah. Thanks for doing that.

00:01:19 Yeah, that’s cool. So you do a lot. So you blog, you tweet, you train people in person and remote and write books, and you’re kind of doing a whole bunch of stuff. Yeah.

00:01:32 Trying to stay busy.

00:01:34 And one of the courses that you had applied, Pandas Twitter Analytics. Tell me about this course. Yeah.

00:01:40 So while I was doing the authoring course, I interviewed like more than a dozen authors and just talked to them about their experience offering. Some went the published route. Some of the self published route, some went self published on Amazon. Some went self published with, like, Gum Road. And so it was really cool to understand how different people are offering and how that works out for them.

00:02:07 And a couple of the people that I interviewed claimed that they made most of their sales through Twitter, which I found to be a little bit interesting in that the typical advice that you’ll hear from self published authors or people hustling doing their own thing is that everything is in the mailing list. Right. Because if you own the emails, then you have access to the customers and you can’t get shut down. However, Twitter and we know that people can get shut down from Twitter for various reasons. However, Twitter is a place where you can reach a bunch of people, and it’s pretty easy.

00:02:50 And these authors demonstrated that they can make a good bit of money from just their Twitter audience. So as someone who like you said, I do a lot of things. I spend most of my time doing corporate training, but I do have books that I’ve written and I have courses that I sell as well. I like to say that I sell snake oil and tell people how to tell lies with data.

00:03:20 I literally teach people how to use Panda Python, which I call snake oil. But I thought, okay, maybe I should look into Twitter a little bit more. And I’ve been on Twitter for many years.

00:03:37 The first time I joined Twitter was I think back. I think I joined around 2006 2007 because I was at PyCon, and one of the talks at Python had like a live Twitter stream and you had to submit questions through Twitter. So I’m like, okay, I guess I’ll join the social media platform Twitter. That was the impetus for me joining Twitter. And I’ve just sort of tweeted randomly. And it’s been great to interact. I think, for people who are listening to this, just in general, if you’re not on Twitter, I think it’s a great place to get side channel information about micro niches or even macro niches. So the Python community tends to be pretty good. I think you have an example of this. Just the other day, you tweeted a question and got great feedback, trying to remember what your question was. But I remember you asked a question. A bunch of people are giving you feedback.

00:04:36 Yeah. And that’s one of the neat things about the network effect. So for instance, people, this is one of my favorites.

00:04:46 I don’t remember the question right off the top of my head either, but somebody asked me a question. I didn’t know the answer, but I have more followers than them. And I thought that the question was interesting. So I rephrased phrase the question a little bit and posed it to the people that follow me. And then a lot of people gave answers, and I thought that was great to just hear get a community to answer.

00:05:12 There is a strong Python community on Twitter and you can kind of find what you want on Twitter.

00:05:20 So I think that’s great.

00:05:24 The take of this course was, okay.

00:05:27 I’ve been on Twitter for a while. What does it mean to take Twitter seriously? And what I did was I actually took a couple of courses on Twitter. Actually, one of the people who I interviewed for a book had a course, and I took that course and I took another course on Twitter, just talking about how people use Twitter and people who claim that they had whatever success with Twitter. And success is one metric for success is your follow account. So why is follow account important? I think that’s probably a question a lot of people might ask.

00:06:03 And here’s my answer to that. My answer to that is, if you look at graph theory, as you add more nodes to the network, the graph grows a lot faster. And so if you have more people who are connected to you, you can get a lot more signal, like you said with this person who asked you a question, and I’ve done this as well. Right. And I think that’s sort of the idea behind Retweeting or whatnot if someone asks a question, they’ve got ten followers, they might not get any responses. Right. But if someone has 1000 followers or 10,000 followers, they’re going to get probably better responses. And when people respond to them, other people are going to see them. And so this was really brought home to me when I started looking at my Twitter data.

00:06:52 And my biggest tweet at that point was a tweet that was a response to someone who had half a million followers.

00:07:06 I think the question it was by James Clear, who wrote I Believe.

00:07:11 What did he write? The habits book. Atomic Habits, I Believe.

00:07:15 Oh, yeah, great book.

00:07:17 Yeah. And so his tweet was like, what’s the best expense, like less than $100 or something that you made? Or what’s the best expense that you’ve done during Covet? And so I just posted a picture of my desk because when COVID hit, a lot of my live training stopped. And so I’m like, okay, I got to double down on this virtual stuff. And so I dropped a chunk of change into getting an SLR camera to use as a webcam, a teleprompter and lighting and all that.

00:07:50 I just posted a picture of my desktop, and I think I got 50,000 views on that picture.

00:08:00 To me, that was like, well, this is crazy, right? I think the next highest tweet I had after that was like maybe eight or 10,000 views, right? So was that the most valuable tweet ever?

00:08:17 I don’t think so. Right. But it just brought to me, it just really made the point that like, hey, if you want to boost your signal, one way to do that, I would say, is to write good responses that get attention to people who have a lot of followers. The other way is to have a lot of followers.

00:08:39 So one thing led to another, and I started analyzing my data.

00:08:47 Then I made this course. It was like, okay, so if I was going to do a consulting gig for you and you asked me to do some analysis of your data, something similar to what I do in this Twitter course would be something that I would do for most of my clients. But I thought a lot of people have Twitter data. This is a great excuse to look at your Twitter data, analyze how you tweet, but also learn some pandas in the process. So that’s the impetus behind that course there.

00:09:18 Thank you, Python for sponsoring this episode. I’ve been noticing lately how much I use PyCharm run configurations. Run configurations are a way to run something from within PyCharm and see the output. The obvious use cases are to run a Python file or run pytest against a file, test or directory. Of course, I use the run configurations for pytest all the time, but I also use it for a whole bunch of other things, from running Pytest and talks to running a Dev server to rebuilding Docker images. Pycharm run commands save tons of time and save me from having one more shell window open on my desktop. So thanks, PyCharm. Try PyCharm Pro for four months for free by going to testandcode.com PyCharm.

00:10:09 Okay, so where do you get the data? Is that something that anybody can access their own data.

00:10:14 Yeah. If you go to Twitter analytics. So I’ll describe this on here.

00:10:22 Obviously most people are listening to this, but if you go to like your Twitter account on the left hand side, there’s a dot, dot, dot more symbol there. And if you click on that, there is an analytics menu option that pops up.

00:10:39 How have I not ever known this was here?

00:10:42 Well, yeah, I mean, if you don’t geek out on that, you don’t find it. Right. And so if you go there, there’s just like a summary that has over the last 28 days. This is your summary over the last 28 days. But what you can do is you can click on where is it the tweets?

00:11:06 So at the top of analytics there’s home tweets and more. Currently, as we’re speaking in summer 2021, that’s what the menu looks like. And in the upper right of that, there is this tweet activity drop down that says last 28 days. And there’s an export data that you can export data by tweet and by day.

00:11:27 And so right there you can basically get a CSV file with your tweet information. Now, just for those who are trying this at home, I found that this is a little bit flaky in that you sometimes click to download and you think it’s downloading and sometimes it doesn’t download or haven’t found it to download.

00:11:51 My advice would be to, if you’re downloading the data there, break it into one month chunks. Don’t try and say I want to download the past ten years of data because I don’t know that it will respond. If you do that.

00:12:03 Then if you look at. Yeah, so there’s like the last 28 days or whatever, and then you can’t do months.

00:12:10 Yeah. So if you click on that, there’s a little dropdown, you can get the calendar days.

00:12:15 My advice, my experience has been it’s been most robust when I just do one month at a time and then I export the by tweet data.

00:12:26 Okay. Yeah. So there’s by day and by tweet. Okay.

00:12:31 And so if you do that, you’ll get a CSV come separate value file. And in that it will have like, here’s your Twitter link, here’s the text of the Twitter, here’s how many views it had. And then it has these other interaction stats as well. Like, here’s how many people clicked on your profile because of that. Here’s how many people clicked on the link in there. Here’s how many people clicked on the image.

00:12:55 For someone who’s like a data scientist, this is a pretty good data set for starting to understand how your tweets performed because it has interaction with the individual tweets. And one thing that I learned from taking these other courses is that like, if you want people to whatever follow you, then a lot of people will follow you if they click on it. If the tweet forces them to click on your profile. Right. So if they’re interested, if your tweet causes them to say, oh, maybe I should look at the profile of this person. Right. And then from the profile, you get a click on the follow up.

00:13:35 That data is there.

00:13:38 Yeah. It’s basically like how many people let me pull it up really quick here.

00:13:49 It’s got like how many people went to your I think they clicked on your follow on your follower. But I don’t know if it has people followed you directly from the I don’t remember.

00:14:00 Let me I’m pulling up some of mine.

00:14:03 Okay. See what it’s got pulled up here as well.

00:14:06 Oh, user profile click. Yeah.

00:14:08 So you have those user profile clicks, and presumably from that you can somewhat infer that some percentage of people who click on your profile will click on follow you.

00:14:19 Right. And so the higher the number, the more people you get to follow you.

00:14:26 Okay.

00:14:28 Yeah. Because that’s pretty important.

00:14:32 Yeah. So pretty important. And so maybe I’ll just describe the course a little bit more. So what the course does, it goes through some cleanup of the data, some exploratory data analysis of that all using Pandas and Jupiter and doing some visualizations of that. And then we wrap up with some machine learning. So one thing I like to do when I get structured data like this is to do what’s called principal component analysis, which is a nice little well, it’s called machine learning, but it sort of predates what the range of machine learning is. But the basic notion of principle component analysis is that you have a bunch of data in tabular form, and it tells you which features or columns are the most important for those. And then you can do a scatter plot with this.

00:15:22 I had probably 20 columns or something. And then you can scatter plot this in two dimensions. And it basically throws all this data in two dimensions. But you can tell that along one axis is like how much interaction you get with your tweets. And another one, it might be like what type of tweets they are. And so you can do kind of like a Magic Quadrant, like a Gartner Magic Quadrant doing something like principle component analysis where you say if I’m in the lower left hand corner, then I have tweets that maybe a lot of people didn’t look at and maybe they were pretty simple tweets. And then if I go to the right hand side, I have more people looking at them. And then I have the more people who are looking at simple tweets and then a pie I have, the more people who are looking at complex tweets, something like that. And so it allows you to pretty not always, but a lot of times come up with some insights into what kind of tweets are similar to each other because they bought similar tweets next to each other in this principal component 2D plot. And then another thing I like to do is clustering, which again, is not really. I mean, it’s classified as machine learning, but the clustering stuff is like from the 70s or whatnot.

00:16:39 But basically you’re taking the tweets, they’re similar and then putting them all together. And so you can apply some techniques to determine how many clusters there are. And I think I had like seven or so different clusters of tweet types that I had. And then you can dive into those clusters. It’s like, oh, these are clusters that are high performing. These are clusters that are responses. These are clusters that are responses and known as bad performing. And so the idea with the clustering there is say, I want to find out what my best performing tweets look like, look at that cluster, and then maybe I want to focus my tweeting attention on making tweets similar to that. Right. And so if you tried different styles of tweeting, I mean, one thing that’s pretty popular these days is like a thread where you’re like, oh, I’m going to deep dive into this as much as you can deep dive into something at Twitter, rather than having a single tweet, you’ll say this is a thread and then you’ll have like ten tweets responding to that. And so if you start experimenting with these, you can see whether these thread type tweets are just your simple, like, snarky response tweets or your tweets that have polls in them or tweets that are questions, which types of tweets your audience resonates more with, and which types of tweets you should probably be using more if you want to have better interaction with your audience.

00:18:05 Yeah. And I’m thinking also, just not both of us are people that like to teach other people. So I’m using Twitter not just to reach people or try to sell them or anything.

00:18:17 I just also want to help the community. So these sorts of analytics can also just help see why waste your time on stuff that people are ignoring.

00:18:30 Yeah, exactly. I mean, sometimes you might think, oh, this is an awesome tweet. You spend sometimes, like writing a tweet as much as like 280 characters is nothing. Writing a tweet can take a long time.

00:18:43 So I tend to be, I guess, someone who paints my own bike shed or whatever. I actually wrote a program so that I can write tweets in this markdown esque type markup language. And then if I want to embed code snippets in there, it will take the code and then we’ll actually use pillow and make a screenshot of the code in there and I can embed images in there. And then I can do threading in there and respond to other tweets. So I wrote a program to do this so I can have different tweets and do, like, grammar checking on those.

00:19:15 That’s incredible.

00:19:18 I love that.

00:19:20 But the point is, you can spend a lot of time going through and making a tweet. Right. And thinking, oh, this is the awesomest tweet ever. And then you look at the stats of the tweet. It’s like, you know what? People liked your tweet. That said, how do you use pytest? What are your best recommendations for pytest? Right. Which takes 10 seconds to write rather than a half hour to think of and make sure the threads are all working and all that. And you get better interaction with the question, what are your best recommendations for using Pipe test?

00:19:53 Yeah.

00:19:58 So you did all this analytics, you did the course around it, but then you applied it to your own Twitter account. Right.

00:20:09 You tweeted in, let’s see, on May 10, I just looked it up this morning. You posted, hey, at the beginning of the year, I had 6000 followers. I just passed 10,000. And when I look today, you’re over 24. Yeah. So that’s pretty impressive. So Congrats, I guess.

00:20:31 Yeah. It turns out that with a couple of $100, you can buy some robots and you can get a lot of followers.

00:20:38 I wasn’t going to actually ask that. I was kind of thinking, so how much did it cost in robot followers?

00:20:46 Yeah, it’s all Bitcoin.

00:20:51 So to be serious, I believe most of these people are real people who are following me. And I have made an explicit attempt to get more followers. Right. Because just that idea of network effects. Right.

00:21:11 I think I share some pretty good content and I have resources that I think a lot of people want access to. And so why not spread that? And there are a lot of people who are interested in the things I’m interested in. I mean, mainly I use my Twitter for Python and data science stuff. I don’t tweet necessarily a lot of like personal stuff per se. But yeah. So in the past two weeks, I believe I’m looking at my stats right now and I’ve gained almost 12,000 followers in the last two weeks, which is a lot of followers because I’ve got 24,000 right now. So basically, I’ve doubled in the last two weeks here.

00:21:55 And so it was not with nefarious purposes. It was not like paying people to follow me.

00:22:04 This is another thing that I came away with. Well, let me give some Twitter advice and then I’ll talk about my specific attempt here. So I think there’s the Pablo Picasso quote, good artist copy, great artists steal or something like that.

00:22:23 Something wrong along those lines. But I mean, you’ll see similar stuff in Twitter. Right. Tell me how old you are without telling me how old you are. Tell me how long you’ve been using computers without telling me what year it is.

00:22:38 Sort of Twitter means. Right. And so one thing that you can do is just look at those things that are popular and say, okay, I’m going to apply that. Right. And maybe that mean, maybe a lot of people are seeing that particular tweet, and maybe they’re in your network that day. So maybe you want to just sort of jot that down as a note and say, okay, maybe in a week I’ll do my own version of that. Right. So I think that is one thing that you can do is look at tweet styles that are successful and then build on those. And like I said, some of the courses that I took sort of walk through some of these tweet styles, and some of the courses that I went through were like, okay, if your tweet takes up more vertical space, people are going to see it. And so people will like it more because they’re seeing it more. And so part of my analytics was that I went through that process. I actually counted, like, number of characters in my tweet, number of new lines in my tweet, basically as a proxy for vertical space there and looked at to see if there with my tweets, if there was better interaction, if my tweets were longer, if I spaced them out more people like breathing room in their tweets. So if you have breathing room, people will interact with them better. I actually didn’t see that with my tweets. So maybe people don’t care about that. And it might be somewhat Twitter independent. But, yeah, I guess the point there is that it’s hard to predict something that will go viral. Right. And so just because someone, you see someone’s tweet that went viral and got a lot of hits and it had white space in it and lots of lines does not necessarily mean that you’re just adding white space into your tweets or emojis. I also looked at emojis, right? That’s a feature that I use when I did my machine learning and analysis to see whether emojis make my tweets better. And you’ll see that I do use a bunch of emojis in my tweets. But my analysis is also that the emojis don’t really have an impact as to whether interactions better with my leads.

00:24:57 Any idea if threads are read more, if there’s the thread emoji?

00:25:05 Yes.

00:25:06 Threads are sort of hit and miss as well. Right.

00:25:12 I’ve found that sometimes threads can be good, but a lot of times, going back to that, I spent a lot of time composing these shreds, and it’s like, didn’t really resonate with some people.

00:25:27 But here’s another thing. Another thing is you have to realize that anytime you tweet due to how Twitter works, it’s not guaranteed that everyone who’s following you is going to see your tweets.

00:25:41 Right?

00:25:42 Right. And so another thing that I’ve been doing with some of these, like, remember I talked about that program that I made some of these tweets that take me a long time.

00:25:53 They’re like more masterpieces. Right. Rather than just stroke of genius, here where it’s like, yeah, I spent like half hour an hour writing out this thread and putting all these examples in it.

00:26:05 Well, I have a program that I can repost that. Right. And so that’s one thing that you can do is repost your content.

00:26:16 And a lot of times people won’t see it. And I don’t think it’s necessarily bad content bad to repost it, especially if you think it’s relevant to your audience. And a lot of times it will hit different people or might have a different impact.

00:26:34 Yeah. As an example.

00:26:37 So for this podcast, I often announce stuff, but I’m also often working late at night and I’ll post it like at midnight or something that it goes live. And I know there’s people in Europe that are already up that will see it. So I go ahead and I’ll tweet it out. But a bunch of people in America are asleep. So at eight in the morning, the next day, I’ll repost that notification.

00:27:05 Yeah. And I think that works. One of the courses I took literally said if you post it three in the morning, I think my time. So like two in the morning your time. I’m in Utah. So three in the morning my time. That’s the optimal time to post on Twitter. So actually, one of the features that I added to my Twitter data set was time of day when I’m posting. Right. And I don’t think I actually had any tweets from three in the morning unless I was traveling and happened to be like on the other side of the globe, typically asleep at that time. So maybe I’m missing out on that. But I, with my Twitter data set, did not see a spike on time of day. Right.

00:27:49 I don’t understand how does that make sense? Because of Europe.

00:27:54 Yeah. The idea what they claimed was that basically you’re going to catch Europe as it wakes up and then Europe is going to kick it off, and then it’s going to be popular by the time that us wakes up, and then it will sort of flow through to Asia after that. And so we’ll sort of be sort of like doing the wave at some sporting event. Right.

00:28:20 Okay.

00:28:23 My analysis did not prove that to be correct with my data set. But again, when I ran my data set, I also had a lot less followers than I have now. So maybe if I reran that, I might see something a little bit different.

00:28:37 Well, and I like the idea of continually checking this. So every couple of months you can do a recheck to make sure that it’s going to be different for everybody. So what’s working for you today might not work for you two months from now, and it might change because you have a different set of followers.

00:29:00 Whatever.

00:29:02 Yeah. And that’s it in general. That’s a problem with machine learning in general. This notion of drift where if you have some sort of model, presumably your model works as well as it can or as well as you want it to, given your business considerations when you release it. But after that, it gets out of date. Right. One of the nice things about the course is that there’s a bunch of notebooks there. And basically, if you just reload your data, get your new data and run it through there, you can run the analysis again and see whether things are changing or not.

00:29:35 That’s cool.

00:29:38 We got to talk about the secret sauce, right?

00:29:40 Yeah.

00:29:42 So what worked?

00:29:43 Yeah.

00:29:45 Like I said, at that point, a couple of months ago when I created my course, my most popular tweet was this tweet that was showing my desk. Right.

00:29:56 Which had 50,000 views.

00:29:59 Now you just tweet pictures of your desk now, right.

00:30:03 There have been quite a few pictures of my desk. None of them have been that viral. That one has. But as of today, I’m looking at my stats right now. My most popular tweet was tweeted in this month, and it has over half a million views. And the tweet is a tweet that I’ve actually done a lot. And I’ve done it for the past four months. And it is basically a tweet that says, I’ll just read it. It says Hi. Then it has an emoji waving, if you’re interested in colon, Python with a Python emoji, data science, I think, with a wizard hat emoji and career advice with a hammer. Follow me with a check Mark after it. I’m planning on using Twitter to share a lot of content that you won’t want to miss with a little party buzzer or something after that emoji. So that’s my Twitter tweet. And that one, I believe. I tweeted a little over a week ago, and it basically went around and has over half a million views. Like I said, this tweet is not original.

00:31:11 I have this program, and I originally tweeted this, I believe, in February.

00:31:18 And when I tweeted it in February, I think I got actually I can look at my stats here. I have my stats pulled up here. Okay. So in January, I had 98 followers that were new followers. That was sort of when I started saying, okay, I’m going to start taking Twitter seriously. My course had just come out, and so I was going to start taking seriously. The first time I tweeted this was in February. And I got around 500 followers from that. And I’m like, wow, this is crazy, right? I mean, I had 600 followers at that. .60 followers at that point. And so I’m like 10% increase for doing that tweet the next time in March. I just figured, oh, let’s tweet it again in March. I’m like, probably won’t get as much, but I’ll try it again and I got like 900 followers.

00:32:07 Okay, that’s pretty crazy.

00:32:11 Go to April. I’m like, well, can it hurt? I got 1500 followers in April.

00:32:19 Okay, May.

00:32:22 I got 3000 followers in May.

00:32:25 And then last month.

00:32:31 So I’ve basically gotten almost 12,000 followers this month from tweeting that. So the growth of that has not been linear. It’s sort of been crazy doing that. And again, I think that goes back to this network effects, right, where I think the people are liking what I’m tweeting.

00:32:52 It’s not just empty content. I try and share super relevant Python advice and data science advice.

00:33:00 And I think the people who appreciate that are liking that and retweeting that. And so I think I got 400 retweets on that and 4000 likes, which for me is pretty good.

00:33:12 Okay, on the flip side, you’re not looking at this and going, hey, I’m getting a whole bunch of followers. So I’ll tweet it every day because that would be annoying.

00:33:26 So thinking about like the I don’t know who did it, the left right job sort of thing.

00:33:36 You’re not selling all the time or in even this. It’s not really selling, but it’s like, hey, please follow me sort of thing.

00:33:46 You’re not spamming that all the time once a month seems completely reasonable. But if I was following somebody, an example is threads. Like, some people have had success with threads and I’ve actually followed people because I’ve found an interesting thread that taught me a lot. But then I noticed that every day they’re posting like these huge threads and sometimes that kind of gets annoying. And what’s annoying depends on the person, of course. But I like that you’re not just bombarding people with this everyday.

00:34:16 Yeah. What I try and do is I try and before this tweet and after this tweet, I try and actually provide some tweet that has pretty good advice in it. Right. And so what’s going to happen, presumably, is that the person who sees this is going to click on my profile, they’re going to look at my recent tweet and they’re say, oh yeah, he’s not just lying here. I mean, like I said, if you look at my tweets, most of them that aren’t replies are pretty focused on Python and data science. Right.

00:34:52 And so they’re going to look at those and they’re going to say, oh yeah, there’s tweets here that are relevant that are exactly what he’s talking about here. And so I think that’s relevant. I think another thing too, Brian, is again, one of the courses that I took here, which was talking about reusing your content was like the advice went something like this.

00:35:16 If you can’t remember the last time you tweeted it, feel free to tweet it again. Right.

00:35:23 If you tweeted it yesterday, yeah, don’t tweet it again. Right. But if you can’t remember the last time you tweet it, tweeted again. Because like I said, it’s probably pretty likely that most of your followers didn’t see it anyway, just due to how Twitter works. So it can’t really hurt.

00:35:41 But yeah, if I was constantly spamming this and this was my only tweet. Yeah, that would be pretty annoying. I would probably unfollow myself too.

00:35:52 I try to use the same advice for jokes and not tell a joke again to my family if I remember the last time, but I barely don’t have that good of a memory. And they will remind me that I told it yesterday.

00:36:05 Yeah, that works too.

00:36:08 Well. Speaking of that, though, you were thinking about in the programming space of Twitter, you’re using the Twitter to communicate around Python and Pandas, things like that.

00:36:21 But people also put personal stuff on there, even if you don’t have analytics around it. Do you have advice around how much personal stuff to share or if that’s just a bad idea if you want to increase your followers.

00:36:37 That’S a great question.

00:36:43 I mean, my take on it is maybe here’s how I’ll summarize it. I think people want to know that you’re a real person. Right. And I think people can connect with you if you have particular interests.

00:37:02 But my interests are like ultimate Frisbee. I do a lot of mountain biking, that sort of thing, which I could talk about. And I know that there are like tech people who like ultimate because I’ve met a lot of people who like ultimate. Right. But that’s a pretty small niche and would probably bore a lot of people similarly with mountain biking.

00:37:24 It’s interesting. I like going rides, but geeking out about that on Twitter, probably not too much. Right. Maybe applying it to teaching and learning. I think it is more interesting. So I think it’s important to come off as a real human and not a robot. I think people like that. But I don’t know. My personal take is that at least for me how I’m using Twitter, a lot of people who I’m following, I like to see the human side of them, but I’m not following them for the human side directly, if that makes sense.

00:38:01 Okay. One of the things you mentioned was like having people look at your profile because they’ll probably look at your profile. They have to follow you. I think so. Looking at your profile once in awhile and looking at what do people see when they look at that? And does that represent really who you are or what you want to represent? And if not, maybe take it out.

00:38:27 That’s a good point. Yeah.

00:38:29 Both the courses that I took are like, you should probably look at your profile. Right. Most people have pretty bad profiles or poor profiles. Right. And this is going to be the first thing people see. And you should also look at your pinned tweet. Right. Because the Pin tweet is something that you can control at the top here. So my Pin tweet is currently like it’s a thread. And the first one is announcing when my machine learning Pocket reference came out, which is now two years old, but it’s an O’Reilly book. It’s got an animal on the cover so I think for a lot of people who are interested in what I tweet about, I think that’s kind of like, oh, there’s kind of instant credibility there. This is an author, they have a book on machine learning with Python.

00:39:13 Maybe I should follow them. Right. One more thing that I think is important, in addition to your profile is sort of a call to action. And I guess this is the thing that my secret tweet here is the thing that maybe impressed me the most is that a lot of times we just assume that I’m going to make great content and people will follow me.

00:39:35 But it’s amazing that just saying, hey, follow me. Right. And having that call to action kind of makes all the difference. Right.

00:39:47 That’s just something that I’m going to be a little bit more aware of with my intentional tweeting is like, what is my call to action? Right. And if you just assume that people are going to follow you just because of whatever, I mean, they might.

00:40:08 And I’ve noticed this now that I’ve started paying attention more to Twitter, but a lot of the threads that you sort of talked about at the end of the thread will be like, hey, if you like this, share it and follow me. Because I’m going to share more content like this. Right.

00:40:24 And most of these people have lots of followers because people were told to follow them.

00:40:31 Yeah. Interesting. Well, watching YouTube, I’m watching maybe a five minute YouTube video. A few seconds at the end, somebody’s saying, hey, smash the button or whatever.

00:40:44 I assume they’re doing that because it works and it’s not too annoying to have that at the end. I actually kind of like the idea of doing that as part of a thread because it’s not totally in your face. If I’m not interested that I got to the end, why not?

00:40:59 Yeah.

00:41:01 Okay, so we only talked about one successful tweet so far.

00:41:06 Any other winners or secret sauce that you want to share.

00:41:11 To be honest, for getting followers, saying that you’re going to post about something and asking for me, that’s what’s worked. Right.

00:41:21 Okay.

00:41:21 But let me talk about maybe some other patterns that I’ve seen that have been good with my community. And that’s been I mean, I’ve been sort of the mentality, oh, I’m going to write these threads and they’re going to have all this information. I just taught this class on Python. We had this great discussion in my class. So I’m going to take one of these discussions and distill it into a thread. So I’ve done that a couple of times, and the results have been hit and missed. I’m like, I spend an hour, like, composing the thread, making the source code for it, making the examples, and then I’ll go off and do another tweet that’s like, okay, what is your advice on using Pi test? Right.

00:42:04 And it’s just open ended, but somewhat specific and catered to my audience. Right. Like, what’s your best advice for doing that? And I’ve seen really good interaction there and a lot of value. Right. Because I think you’re going to get different points of view. And I found things about how people are using pytest. I taught Pietest and used pytest for a long time, but I saw a lot of that’s. Cool. I didn’t realize that. So I think being willing to have an interaction and show that you want to have an interaction rather than just pushing, pushing, pushing, saying, hey, let’s chat about this or let’s share about this has been successful as well.

00:42:53 Yeah. Like you mentioned at the beginning, one of the things with my network was now I am looking at my analytics. I can totally see which one it is.

00:43:01 Somebody asked, hey, Brian, what’s the best way to distribute private what’s the best way to distribute private internal packages?

00:43:10 So I just shared that and a lot of people responded, yeah.

00:43:15 And when I was looking through those responses, I’m like, oh, there’s some packages in here I didn’t know about or some features of, like, GitLab. I didn’t know that GitLab could serve as a private internal repository.

00:43:27 Yes, that’s totally one of my side projects is figured that out yet now because we use GitLab. And I didn’t know that.

00:43:35 Some other advice is media does work. Or at least I found that it works. And I found that what my audience likes is some of them like to see my desk. My desk is a little bit weird. I mean, I have a weird keyboard and multiple monitors and my teleprompter. But also people like to see books. Right. And so being published with O’Reilly, I occasionally get like, oh, they’ll send me a copy or translation of my book. And so some other popular threads have been, oh, here’s a bunch of my books. Right. In different languages or some other stuff that I’ll do is when I’m teaching a class, one of the things I do for some of my classes is that, like I said, I invested quite a bit in my virtual training setup. And so I’ve got a computer that I can write on. Well, it has a digitizer on it.

00:44:39 So I have my slides, but I’ll annotate my slides. And so I’ll draw little pictures about, you know, here’s the code. But here’s sort of what’s going on behind the scenes. Here’s the Python objects that are being created, here’s the local namespace, the global namespace. Here’s how these Dunder things work and whatnot. And so I’ll take slides that are heavily annotated from my courses and then share those.

00:45:04 And it looks like there’s good interaction there and people like that. So the courses said vertical space is important.

00:45:17 My take is not that vertical text space is important, but if you have an image that’s relevant to your audience, I think that can be very useful for boosting a tweet.

00:45:31 Wow. I just found one of your annotated slides that’s pretty interesting.

00:45:41 Yeah.

00:45:41 I guess that’s another thing too is we said look at what the tweets that people see when they put your profile but one of the buttons people have is media so it’s important to look at that too.

00:45:59 A lot of people will just put in a bunch of animated gifts and I don’t do that. But you look through some people’s medians like it just replies with gifts. Yeah it just replies with gifts like the dog burning and everything’s okay or some get cat or something like that.

00:46:15 I don’t think that should be included in the media section to be honest. If it’s just a gift but a gift but if somebody’s taking a picture of their own cat all the time then yeah, sure.

00:46:27 Yeah.

00:46:29 Anyway.

00:46:30 Well cool. Well thanks a bunch. I’m actually looking forward to learning more about I do care about follower account but mostly I care about reaching people and helping people so I’m going to take a look at some of this stuff with my own account so thanks.

00:46:49 Awesome. Yeah.

00:46:50 And we’ll talk to you next time and good luck with everything.

00:46:54 Yeah. Thanks, Brian. Pleasure talking to you. Take care.

00:47:03 Thanks Matt for that fun interview and for your efforts with teaching and training. Thank you, Patreon supporters join Them at testandcode.com support thank you Pie charm for sponsoring the show. Check them out at testincode.com PyCharm those links are in the show notes@testandcode.com 159. Oh, and one more thing. There’s a link in the show notes for Matt’s course that includes coupon code Test And Code already in the link and gives you half off the course. Thanks Matt. And again those show notes are@Test And Code.com that’s all for now. Now go out and analyze some Twitter data.