Blake Reichenbach on Marketing, Curiosity, and the Love of Books

In this 350 2nd episode of data driven, Frank and Andy speak with Blake Reichenbach. Blake is a product manager at HubSpot, focusing on the Content AI platform, and is the owner of Howdy Curiosity, an online nonfiction bookstore and learning community. Stay tuned for a delightful conversation on data, AI, and the love of books. Hello, and welcome to Data Driven. The podcast where we explore the emergent fields of artificial intelligence, data science, and, of course, data engineering,

Which is really the underpinning that makes it all possible. And to that end, I have my very favoritest, data engineer in the world, Andy Leonard. How are you doing, Andy? I am unlike both of you, I have not yet had COVID, but I and I'm doing well. But it's in it's It's in our house. We have a a home member here who has tested positive. So we're all walking on eggshells, over here. And, but I am doing well. I love the, you know, the data engineering part. I really love Frank's

article that it was written way back last year. It's So 2023 about roadies about roadies versus the rock stars, and he calls weed data engineers the, roadies. And, yeah, doing well. Frank, I got to present last night in person In person. Richmond Richmond, Virginia, Not not Kentucky. Which is a good segue to our guest It is. Who is in Richmond. Andy and I met in Richmond. We organized the Richmond code camp,

although that was Richmond, Virginia. We are here today with Blake Reichenbach, who is a project manager product manager, sorry, at HubSpot Focusing on the Content AI platform, and I love to know more about HubSpot in general. One of the podcasters that I follow and and admire is John Lee Dumas, and I know he has with HubSpot. But welcome to the show, Blake. As we were talking in the virtual green room, you're recovering from COVID. I had COVID flu strain a, Sinus infection then

followed up with this week. You're such an overachiever, Frank. I have to do it all day. I just have to do it all day. I know. He's putting he's putting my COVID to shame. I Should've got out and gotten back into something else before joining so that I could keep up. I have 3 kids, also known as bioweapon incubators. So Fair. Fair. My my only kid is, you know, won't

be visible on the podcast, Yes, of course. But he's the, large bulldog sitting behind me, and thankfully, he doesn't tend to break too many germs into the house. That's awesome. That is That's awesome. So so tell us about HubSpot and and what it is you do there. What is HubSpot exactly? It's one of those things that's on I have a board of things I'm supposed to look at. And HubSpot is is is on the list, but with everything going on, I haven't had a chance to.

Yeah. Well, if that's like your Kanban board of Software to dig into. I would definitely recommend moving that up in your backlog to dig in HubSpot. I'm clearly biased as a HubSpot employee, but It is a really cool company and a really cool product. So we're a leading customer relationship management platform or CRM platform for scaling companies. Our platform includes a bit of everything that a business

needs for their front office. So we have marketing, we have sales, we have Service, we have data operations, and we have the part of the platform that I live in, which is our CMS. And All of these different hubs as we call them or or product lines,

exist around that central CRM. We try and make everything as, you know, fittingly for this this podcast, as data driven for our customers as possible so that they have these Integrated systems across their different business pillars, so that things can stay in sync and aligned, and, you know, staying true to The data that they have about their customers to make the most informed decisions possible.

As far as what I do at HubSpot, I've been with the company for, I'm going into my 7th year now, which I guess by, like, SaaS industry standards makes me a grandfather. But I for the last about, two and a half years or so, I've been a product manager. I started out in our product security organization. My focus is really on, you know, maturing our content abuse and fraud detection systems. And then I moved over more

recently into our content AI platform. So thinking about this really exciting emerging world of AI and generative AI and how that's reshaping or informing the way that content marketers work and figuring out what solutions are going to have a meaningful impact for content marketers. Interesting. I would imagine AI and generative AI is probably very much on your radar. Oh, yeah. Has that been I don't think I've Sorry. Go

ahead. I was gonna say I don't think I've had a conversation in the last 6 months that has not included something about generative AI Or, you know, machine learning or chat GPT or Sam Altman. It's very, very central to, you know, what I'm working on and where my focus is. Interesting. So so how disruptive has it been for your your business? I mean, so is it I guess it's fair to say that that that HubSpot is a CRM, right? And how do people use AI? Like, how is

AI integrated into your platform? Yeah.

So we have, quite a bit of of new AI features that we've rolled out Over the last year or so, you know, I I again, my my focus is really within the CMS So what I'm most familiar with is within our our tools for managing content, building websites. And we've rolled out quite a few, different AI and ML assistance, within our feat our feature set. A lot of those are still, you know, in beta and have to be opted into, But, you know, introducing different, generative AI models to help marketers

just streamline their efficiencies. So doing things like Generating meta descriptions for their pages or rewriting content, you know, pretty what I think in the market is becoming standard for generative AI tasks, has really been our our starting point there. You know, I think we as a company have been, Looking at AI, you know, longer than it's been this, like,

flashpoint in public conversation. Coming from the security background, one of my first Big projects and product security was in, you know, maturing our abuse detection systems and figuring out, you know, how can we leverage LLMs? How can we leverage machine learning models To improve our precision and make sure that fewer, you know, fraudulent pieces of content slipped through the cracks.

And then once, you know, ChatGPT became, like, the tech topic of the day, you know, that's where, HubSpot along with a lot of other folks in, in the same space started saying, okay, cool. How can we pull these features in app to, You know, give our customers new new tools to use, new things to play around with, and better ways to improve their own efficiencies. Interesting. And and you're in the since you're in that marketing space, like, the and and and I would imagine

It's a very data heavy world anyway. Right? Like, it's a very it's you already start off with a bias towards being data driven. Yes. I said the name of my own show. But but, I mean, like and I think that, you know, I'm just fascinated by marketing. Right. Like, marketing is my new fascination for 2024, because I realized in some ways, I'm good at it. In some ways, I'm horrible at it. Actually, Really got awfully horrible at it.

So, but it's funny because as I look into it more, I've reached out to people to kinda help, and they're like, oh, no. You gotta separate the data. I'm like, this is a lot of data analysis. This is this is my jam. Like, I I should be better at this. Yeah. You know, good marketing is data driven. I think that, you know, in marketing, especially content marketing, It's often seen like as much an alchemy as it is a science, where on the one hand, you have some marketers who

are, like, Data obsessed. You know, they will only write a blog post if they have estimated search volumes and, like, you know, customer persona data. And Then you have other marketers who are kind of like, let's crank things out and see what sticks. And I think that's a really kind of interesting intersection with generative AI Because a lot of, you know, early generative AI tools for marketers, and I'm not going to name

specific companies. I don't wanna start any kind of, you know, your flame or there, but A lot of of GenAI tools have kind of just been like a a churn and burn factory where they're cranking out a lot of mediocrity really fast. And, you know, you kind of see if you if you look at performance graphs of companies that have gone this route of just, like, Cranking out generative articles without, you know, human in the loop processes.

Like, you'll see their web traffic and their conversions kinda going up, up, up, up, up, up, hit a cliff, Boom. Drop. Right. And, you know, there's not that long term ROI. There's not that, meaningful customer connection that lets The brand really build upon itself. And so I think that where we're at as an industry now is this really cool place where The marketers, but also the Gen AI tools that are winning or going to win long term are the ones that are able to incorporate data in a

meaningful way. And the products that are able to, present generative AI functionality In a way that is intuitive and that prioritizes UX, which frankly has not been a big emphasis in the Gen AI industry, for the last Dear, I think a lot of companies are rushing to get to market and really focusing on, like, what can the Gen AI do

and not how do customers use it? So that's where, you know, things are super exciting for me right now is we're at this place of combining generative AI with Customer data with user data and also figuring out what's the right balance of having humans in the loop To make sure that brands are able to have that content that's really unique and that is special for their brand and that lets them build relationships with Customers who are probably pretty skeptical, frankly, of of

generative AI on the whole. Well, I love that phrase, humans in the loop. And, usually, when I like a phrase that a guest says, I'll say I'm stealing that, but we're recording this on the 12th January 2024. So I will put it in quotes and I will credit you, for that. Well, I've I've stolen that from another A number of other articles, so I can't take total credit for it. But if it's the first time you've heard it yeah. If it's the first time you've heard it, you can attribute

that to Blake Reichenbach of Richmond, Kentucky? There we go. I'll I'll

do that, and I'll throw in a just, you know, a footnote that says Blake says he heard this elsewhere So that you're covered as well. We wanna be above board here on data driven. I I just wanna do that. But I one of the reasons that phrase strikes me Is that the successes that I've seen, you were mentioning the the successes going up and up and up. I have seen humans in the loop, You know, for those those types of

of, solutions. And what it this is just my simple Bonneville, Virginia, You know, mind the way that I think about things, but it it appeals to me as, a little bit like the old mechanical Turk type thing Where in that, you've got a person doing what people do best, and you got LLMs doing what it really does best. And I mean, on both counts. They outshine the other. Saw an interesting tweet not long ago that said, all

LLM hallucinate. And it's just the answers that you get that you like, you know, that help you or accelerate you are, You know, are are the ones that are just finding the next phrase or nailing the topic closest to them, whatever, though the next word. And and they just are they're doing all all the time that's happening. It's just some of the times the closest word is, you know, half an inch away. Other times, it's half a mile. And

so their hallucination is what they do. And I found that was an interesting take, on that, but that's where the human In the loop, the person, you know, do they they the person in the box in the mechanical turk, that's when they shine because they can look at this and go, well, that no. That's not right. You know, we can't we can't send that forward. So don't really have a question. I just was, very intrigued

by That phrase, that turn of phrase. And again, if I use that, I'll make sure, Blake Reichen, Reichenbach from, yeah, from Richmond, Kentucky. I'm making sure I've got it written down. I was making sure I was gonna say Richmond, Virginia. It's such a I almost said Richmond, Virginia, and I almost didn't. I was like, was it Reichenbach? I didn't wanna I didn't wanna call you the, name of the guitar the really cool guitar manufacturer, Which is sounds close. The old

Rickenbockers. Oh. But which would be a compliment. Now I don't know if it'd be a compliment or not. I like Rickenbockers. To me, it would be, but to you, I'd be saying your name wrong. So Then break it at this part out. Just just just pull thing out. I have a lot of experience with people saying my name wrong. Yeah. That I say your name wrong. Oh, everybody says my name wrong. Even technically, even I say it wrong. I got 2 first names. So, you know There you

go. I I spent this past summer in Switzerland, which is where my family's originally from. And So technically, I've learned that I am also saying my name wrong. But, you know, I I have lived in Central Kentucky, almost my entire life. You know, foothills of the Appalachians. And so, I I have heard pretty much Any variation of the combination of letters in my name, I've I've heard it. Even even my own father often says our last name is Rickenback. Which that's

A little bit further off base than Reichenbach. No. Not at all. That that is in earnest. Yes. Gotcha. Well, it's a little I'm I'm sorry. I wandered off. This is my job on the podcast. That's what we were doing. We were doing. True. But, What do you see as kind of the next step in and it's kind of 2 things in it, but we'll focus on the, the hallucination part of it. And I think that's maybe part of the driver for when you, you know, you were describing it goes up and up and up and falls.

I think that may be part of what's defining the fall. So what do you think the next step is to maybe manage that or mitigate it? Yeah. So, you know, I think

the Sort of first element in that equation is something that I I think guests on this podcast have talked about before, which is like having, smaller, more precise models. They're trained on more nuanced datasets. Right? One of the really powerful things About, an LLM like chat g p t or or one of OpenAI's models is that they are, pretty

solid generalist. Right? And they have this really wide swath of training data, but the sort of double edged sword there is Oftentimes, they're looking at such a huge dataset that the lines start to blur between entities, between topics. And so that sort of predictive language capacity to understand what should come next gets a little bit diffused.

Yeah. Right. So I think that as we see more industry specific or topical specific, or even like, I think Data training sets that are honed in on a specific brand's voice and their own, like, you know, existing corpus Of, of published data. That's where I think we'll see some pretty big improvements in the

quality of these gen AI outputs. Yeah. But The other part of the equation and what I'm really excited about and really interested in is figuring out what that right balances between giving autonomy to generative AI tools and having humans guide those gen AI tools. Right. Gotcha. Because Ultimately, I

think we're still in a phase of, you know, speaking about the content marketing industry specifically. I think we're still in a phase where People want to connect with people and for, you know, brands to be able to demonstrate their own, expertise, authority, and trustworthiness. You know, that's still really critical for building those relationships as a

business to your customers. And so I think that What's going to be a big improvement when it comes to incorporating GenAI into these processes is figuring out that right balance Of saying, here's what I'm willing to offload to an l l m versus here is what, you know, explicitly requires human intervention or human guidance or human prompting. And what makes that equation really complicated, like, talking through that in theory, it sounds

pretty straightforward. But then as a product manager, what I'm always thinking about is, like, how does the customer experience that? So how does the average user Who's, you know, maybe not coming from a data science background or an AI background or a software background. How are they going to interact with these products? You know, are they going to feel like you're giving them a worksheet and they have homework and they're saying, what the heck is this? Or are they going

to feel like, I'm losing control. This is, you know, a runaway train and I'm overwhelmed. Right? There's a a really fine Balance to be struck there. But I also think it's it's an important balance to work toward, and I think it's really important for Companies building generative AI tools, myself included as a, you know, PM at

HubSpot Sure. To pursue that right balance and to, you know, figure out how users interact with the with these tools in a way that gives them a sense of control And that lets their own expertise shine through while also helping them work more efficiently. I I love that, Juxtaposition, if you will. And I I see it, you know, it there's the human, driven part of this, And then there's this other, vector. See what I did there?

Where you're you're using the data to inform the human And both those lines keep shifting and the intersections also shift along. Well, it's more than that, but that is a great, a way to look at it, kind of a, you know, a a 50,000 foot view, and those lines will continue to shift like you said. The autonomy part, I totally agree. I think that's you that is a hard call.

And I you know, like, from what you just said, I gather that The answer may be, several different spots, you know, kinda like, less interactive, Medium interactivity, more interactive depending on the the users, just acceptance dealing with that. Some people may not have an issue with doing the worksheet or answering the quiz, questions, the survey questions so that you can gauge. And that, in my opinion, will put them higher on that interactive scale.

You know, they may be more tolerant. I don't know if that's the right word, of the, of AI. But then you got old cooch like me, you know, that see see all of these questions that have that reaction like, Come on. I got things to do. Just answer the question. So interesting. Exactly. Interesting stuff. It is interesting stuff, and I'm always fascinated by content marketing, and how how the success, like, what you said was very true. Like, there are people that

The either it tends to be bifurcated. Right? Like, you have people who do just will just spew out stuff and not think about the data, and there's people who will, Like you said, like, unless I'm guaranteed x number of this, I'm not gonna write a post about that. And I think that the sanity there's probably some kind of distribution of effectiveness That probably

skews towards the middle, whether it's towards one side or the other. I think that's up for debate, but, clearly, it's not the outliers.

Yeah. You know, speaking speaking as a, former freelance content marketer. So rather than, like, as a a PM in the space, but just as someone who Loves content marketing, and the

the sort of science and orchestration of content marketing. I think that treating your content marketing sort of like a multi bandit test Is the best way to approach it so that, like, you're investing, let's say, like, 70 to 80% of your efforts Into these marketing initiatives where you have really strong data to indicate that it's going to be successful. And, you know, you can say, Like, based on past performance or Google Analytics data or heat map data

that this is likely to resonate with your audience. And then reserving that other, you know, 30 to 20%. I hope I said 70 to 80% earlier or my math is gonna be way off. Okay. Great. No. You nailed it. You know, reserving Perfect. You know, with that other, you know, 20 to 30% of your marketing efforts, doing some experimentation and seeing what sticks. You

know? I I think that, having room within your marketing strategy to say, okay, I'm gonna make a really opinionated post on LinkedIn about this topic And just see what my

audience's reaction is. Or I'm gonna record a, you know, TikTok Even though the majority of our audience is on this other platform just to see, like, how does it perform, what aspects work, And use that as a way to continuously collect new data about, you know, what does actually resonate with your audience, What segments of your audience may you be missing, and are there emerging audiences that you haven't considered yet that may still be a good fit for your product or service?

That's a good point. That's fascinating. Yeah. Try and balance all of those. Goodness. Yeah. You mentioned the, The multi armed bandit program problem, and you basically just described explore versus exploit, like, very well. Right? And it it it applies to more things than just slot machines. So so for those wondering what the heck we're talking about, there's this problem in typically reinforcement learning, Where it's the explore versus exploit.

It's also known as, the multi armed bandit problem, where you basically given a simulated bank of slot machines, How do you maximize your winnings? Is that a good way to describe it, Blake? Yeah. I think so. Cool. I've done a number of presentations on it, and I've had a lot of fun with it. Even in Vegas, I think I actually presented in Vegas. Ironic. Alright. So now, well, let's switch to the pre, done questions. This is a great interview.

Absolutely. Definitely would love to know more about that, but, we wanna be respectful of everybody's time. So here's the first question. How did you find your way into data? Did you find the Data Life or did the Data Life find you? The Data Life certainly found me. I did not go looking for it. So my educational background, my degrees are actually in English and sociology. And when I started working at HubSpot, I So that company seems pretty cool. I'm gonna work there as a gap year before

I go do my PhD in American literature. Clearly, that is not how things played out. Ended up Falling in love with the product and the sort of SaaS ecosystem. And along the way, I realized that to, Meaningfully invest in growing our product and growing my own career, I had to become much more data informed and data conscious. So I am Squeezing every drop out of that single stats

class I took in undergrad that I can. And thankfully, I've I've been able to work with some really, really, really brilliant Data scientists who have been more than willing to say, Blake, what you're proposing is statistically impossible and stupid. Let me educate you on how this actually works, to, you know, flesh out my own skill set and familiarity. Very cool. I love having those people around that'll just say, hey, wait. No.

I have some of those around me as well. Frank's one of them. So, what what would you say, Blake, is the the favorite part of your current job? So my gut reaction was making flowcharts. I love a nice

flowchart. I love, you know, diagramming out customer problems, but, to take that 1 step deeper, I think that for me what I love most is Being able to explore really complex problem spaces where there's not a single right answer And being able to be in a position of influence to say, okay, based on this abundance of choices and abundance of options for how we go, Here's how I think we should approach solving for our customer, and here's how we can

measure whether or not we're successful at doing that. Gotcha. You are such a geek, loving flowcharts. Just say. I I am. I I fully embrace being a geek. I love a nice flowchart, and my coffee cup this morning even said as, oh, this calls for a spreadsheet. So it's he gets very on brand. I love that. I love that. That is awesome. So we have a couple of complete the sentences questions. When I'm not working, I enjoy blank.

Reading and selling books. So, last year, I started up, a bit of a side hustle selling nonfiction books. I've got a website, howdycuriosity.com. I, you know, I I spend so much time reading nonfiction, especially in the, entrepreneurial and marketing and strategy spaces and recommending those books to people. So I decided, you know, maybe If I'm spending so much

time doing this, maybe I can make a couple of dollars off of it. So got an online nonfiction bookstore now and that is, like, My favorite when I'm not product managing, I'm fulfilling book orders, looking up new books, writing about new books, and having a field day there. That's cool. You could do some data science on that on your own market. I could. So our 2nd complete to sentence is, I think the coolest thing in technology today is blank. So

I Simultaneously, the coolest and in some ways, the scariest is, how rapidly things are changing and evolving. You know, over the last couple of years, we've seen just like a couple of, like, flashes in the pan, on the technology landscape where people have said, like, Oh, this is the next big thing. Web

3, the next big thing. NFT is the next big thing. But I think we are actually at the point where we're encountering the next Big thing, which is all the different ways that ML and AI are influencing, you know, numerous industries. And I think that's really exciting, especially to be kind of right in the midst of that, To be able to, you know, chart these waters and figure out how these tools work together, how we can use them to, improve people's lives and hopefully not just, like, make their

jobs redundant. Yeah. That I think is is is really cool and really exciting. Very cool. And our 3rd and final complete the sentence, I look forward to the day when I can use technology to blank. Oh, let's see. I look forward to the day when I can use Technology to, create a dashboard that lets me automate all the side projects that I have running. I am a perpetual tinkerer and doer. I'm

always building something new. And as a result, I have a A ton of spreadsheets and notion spaces and Google Docs and everything else just floating through the ether. I have an

Eisenhower matrix on the whiteboard behind me. I have a poster note Kanban board on the wall behind me, and I would, you know, Love to have, like, a a smart board or something where I can take all of these different projects that I'm constantly Throwing ideas down for, you know, everything from home improvements to side hustles to day job stuff, and and Create a better sense of organization than having Post it notes and docs everywhere. That's a good product idea. I like it. I like it. Yeah.

So we asked our guests, to share something different about themselves, but we remind our guests also because we're all geeks and wise acres That, to remember, it's a family show. We wanna keep our clean rating. So we have to throw that out, you know, just just as a condition. Sorry. Well, I was going to talk about my love of profanity. That's No. That as a joke.

That is a joke. No. You know, I I think something, different about myself, would probably go back To what I just mentioned about being a a chronic tinkerer and a chronic experimenter and doer. I I read a book, Many, many moons ago called the 10% entrepreneur. And there was so much about it

that I, didn't particularly like, But there was also quite a bit about it that I did. And part of what really stuck with me was this idea or I guess this question of, like, What would it look like to allocate 10% of your time and resources to, you know, entrepreneurial endeavors or, you know, Anything that kind of scratches that itch of

wanting to do something more. And so, you know, through my own side project, out of curiosity, And through, the numerous other projects I'm always in, you know, in the process of juggling. I've I've really leaned into that 10% entrepreneur approach and it it is so fun. It's often a time sink and a money sink more than it is, you know, a a a revenue channel, but I just I love it. I love trying

new things. I love learning new things, and I love, forcing myself to stretch my skill set beyond where it's currently at. Very cool. And I I love that, and I'm hoping that the transcription will pick up how to curiosity.com. I love that, you know, you're not Just throwing that in out of nowhere. It's definitely a passion, and it shows up and it keeps showing up. So and I would I I just I was trying to think of some clever way to say how much Ifranksworld.com that

that way the way you're working it in. I'm just Sorry. That's funny. Picking on you a little, Blake, but it's I saw you laugh. So if you're not if you're not watching the video, and I don't think We'll have the video available if you're just listening. Blake laughed at that, so you should too. And you mentioned a book which kinda leads into Frank's next point, I think. Yeah. So Audible is a

sponsor of the show. We we love Audible. Thoughts and prayers go out to the folks who were laid off in Audible this week, but, they are still a sponsor of the show, and hopefully, our domain works to go to the date driven book .com. Can you recommend any good audiobooks or books other than what you've already mentioned? Yeah. Absolutely. So I think, for me, the litmus test of a good audio book Yes. I listen to it and get so excited about it that I go and buy a print copy immediately.

And, recently I did that with 2 different books on Audible. The first being The Long Game by Dory Clark, and the 2nd being Deep Work by Cal Newport. Interesting. Interesting. That is a mark of a great book where you listen to it, and you're like, I have to have this on paper. You know, there's something about as a book lover, I think you can appreciate, you know, there's something about dead trees, that just makes something magical. Totally agree. The Go ahead, Frank. No. Plus, you

can listen and not get distracted by notifications. So Exactly. You know, It's a big issue for me. Sorry, Andy. I cut you off. That that's okay. Go ahead. Oh, no. That was it. Oh, so where, sorry. Where can people learn more about you, Blake? And, you you've already mentioned your side hustle. And I'm gonna check that out because I'm a I'm a book geek too. So where can people learn more about you and and all of the things you're involved in? Yeah.

So you've queued me up to name drop my side hustle for, like, what, a 4th or 5th time? Exactly. Right. I'll get close to the microphone. That's Howdy. Curiosity.comhowdycuriosity.com. See, I mispronounced it earlier. I thought it was how to, And it's how d. And as a Combination of of COVID science COVID sinuses, excuse Me and, southernism of just kind of dropping vowels. I'm not sure you can relate.

That, my my business's website is probably the best place, and then folks are also always welcome to connect with me on LinkedIn. I love connecting with, you know, other folks in the industry, especially the data science industry. You all are some of my Favorite flavor of nerds. I say that with love. So, yeah, either my my business website or LinkedIn. Cool. Awesome. And we'll let the nice British lady finish the show. Thanks, Frank,

Andy, and, of course, Blake for an outstanding show. Alright, you lovely lot. You've somehow endured another episode of our delightful ramblings, and for that, we're eternally grateful. We've got a tiny, almost insignificant request. You know where this is going, don't you? Pop over to Itunes, Stitcher, or your podcast platform of choice. It's just a click away. Even an advanced AI could do it. We need those shiny stars and charming reviews.

Why? Well, it appeases the almighty algorithms, elevates our status in the digital realm, and brings more unsuspecting listeners into our fold. More ears, More joy, and frankly, the world could do with a bit of that right now. So unleash your inner critic, rate, review, and help us conquer the podcast universe. Go on. Make an AI proud.