Increasing Marketing Effectiveness Thanks to Text to Code - with Google's Richard Lee

Richard Lee, Global Head of Marketing for Games and Media at Google, talks about the magic of text to code for marketers, highlighting how automation and quick data scanning can free up time for the "fun stuff."

One of Richard's top tips for students envisaging a career in marketing is to take database management classes to understand data structure and SQL. As he puts it, understanding this is "like understanding the difference between how to build a house versus what are actually, the atomic pieces of what makes up a house."

For experienced marketers, he suggests harnessing the ability of AI to use text to code to enhance efficiency. He shares his experience in automating image and mood board creation or curating lots of pieces of news from across the web for a newsletter. In Richard's view, AI can free up time for more strategic or creative work. Will it let us have more free time? No, but it will make work more fun, as long as we continue to provide human input.

Listen to all his tips on the podcast.

The following gives you a good idea of what was said, but it’s not 100% accurate.

Richard Lee 0:00
When someone asks me if I'm data driven, I actually reframe that and I say like I'm not actually data driven. I would say that I'm actually customer driven, and I use data as that signal and not the driver of what I'm doing.

Speaker 0:18
Are you a fitness loving foodie? Then defy snacks is for you. Defy offers real high quality chocolate you love, combined with functional ingredients, giving you as much as 26 grammes of protein per bag. So you can be both a little bit bad and a little bit good at the same time as a female owned brand, defy dedicates a portion of its profits to other women owned businesses currently sold in the US. You can find us on defy with an i snacks.com and follow us on social at defy snacks.

Tom Ollerton 0:51
Hello and welcome to the shiny new object podcast. My name is Tom Ollerton. I'm the founder of the creative effectiveness ad tech platform, automated creative. And this is a podcast, called shiny new object, that is looking at the future of data driven marketing. Every week or so I have the pleasure and the privilege of interviewing one of our industry's leaders about where they think this whole industry is going. And we've done like 300 episodes of this. So there's been quite a lot of shiny new objects, classic marketing stuff. So I am on a call this week. Let me get your job title, absolutely. Job Title, absolutely perfect. It's Richard Lee, who is Global Head of Marketing games and media and entertainment industries at Google. What a guests! I'm very excited about this podcast, but Richard, some people watching this, listening to this, might not know who you are. Could you give us a bit of background?

Richard Lee 1:41
Sure. Thanks. Tom, yeah, my name is Richard. I've been in product and marketing throughout my entire career. I've worked at large technology companies, small ones like pre IPO, both on the product side and on the marketing side. So companies I worked at you probably heard of are like Boeing, Amazon, some smaller companies such as Unity, which is a gaming company, as well as Niantic. They're best known for a game called Pokemon Go, and I'm currently at the Big G at Google, where I do industry stuff, as you mentioned earlier, but then also starting to work at AI, because, hey, it's Google. Gotta work on AI these days.

Tom Ollerton 2:19
That's a bit of a humble brag, isn't it? You know, smaller companies like Niantic or something, there you go. We've all had different careers, haven't we? Richard, so first question, so someone's looking at your career thinking, right, okay, I wouldn't mind following in his footsteps. So what like? What advice do you have for a smart, driven student who's trying to get ahead in this industry.

Richard Lee 2:43
Yeah, students. That's actually pretty timely, because, like, right now is like, the end of summer break. So like, this is a time when, like, I get random outreach, of like, hey, Richard, I see your background. Can you tell me, like, what classes I should take and depending on the person I might index one way or the other. But the thing that I like consistently come back to is like, take a database management class, not like an analytics class, which is pretty common among like a marketing curriculum, but like, you know, go into a little bit more sciencey, more like information systems to understand database management, because that's the type of class where it's not just like reading report or working on a spreadsheet, you're actually understanding, like the structure of data.

Tom Ollerton 3:24
So, Richard, take, take one step back. What is database management for somebody who doesn't know exactly?

Richard Lee 3:30
Good question. So within database management, so if you think about like being a data driven marketer, where does that data actually live? How does that data actually structured? So that's where you get, like, database management. Things that you'll learn about are how data is organised, like different like normalisation forms, number one, just like, what actually is a database? It's, you know, it's what exactly is a database. You'll hear things about, like, hey, what's SQL? What does it mean to be what's Spanner, right? These things of like, and you understand like, the different components of how a data system is actually created, what's the difference between storing data versus querying that? And you'll learn a list also a little bit about how to actually do the query, such as, like SQL, which a lot of people hear and all this rolls up into a lot of the things that we as marketers use today, which are like dashboards and reports. So it's really like the if I'm using example of, like, science, it's like understanding the difference between how to build a house versus, like, okay, what are actually, like the atomic pieces of what makes up a house?

Tom Ollerton 4:35
That's a great bit of advice. But what would a smart, driven student get out of that. Why would that separate themselves from the herd?

Richard Lee 4:43
Yeah. So first thing is just being able to write your own queries, to know SQL is going to make you a lot more self sufficient. You don't need to rely on an engineer or somebody that's more technical to for you in order to do your job. The second thing is that you get like this all. Almost like an intuition of, like, how data is structured, or like if something looks off in its dashboard or report. So for example, like, one time I was looking at a dashboard and I was cutting it different ways, and I noticed that when I looked at a certain way, the the result I was getting was cut and cut in half by, like, exactly 50% and that's not normal for that to happen. It's not coincidental. Ultimately, what happened was, it turns out the data was it turns out that the dashboard was double counting that data. And that's not something I would notice without having that basic understanding of, like, how these queries work, and then ultimately, to also talk to the data analysts of like, hey, let's actually troubleshoot and figure out what's going on.

Tom Ollerton 5:43
And I think the wider trend that you're picking up on is self sufficiency, right? I was interviewing the other day like there's a very rare set of marketing people who have set up businesses outside of their jobs. Now, whether someone's built their website for their kid's baseball team, or they've like, have a coffee delivery company, not to make money necessarily, but to understand the machine from a non marketing perspective. I think what you're talking to there is like, be self sufficient. Be data self sufficient, because that would absolutely separate a student from everyone else. So kind of moving on. I mean, that's one great bit of data marketing advice. But do you have like, another silver bullet bit of data driven marketing advice for someone who's maybe in the industry now and not in that junior position?

Richard Lee 6:30
Yeah, thinking a lot about my career, like I used to be on the product development side of things. So what that means is that I interacted a lot with, like customers, and I had to understand them, not just from a qualitative standpoint, but then, how do they interact with the products and services I've been using? So a lot of times, like when somebody says that they're data driven, or when someone asks me if I'm data driven, actually reframe that and I say, like, I'm not actually data driven, I would say that naturally customer driven, and I use data as that signal, and not the driver of what I'm doing. And there's a few things that I think that that's important different framework to have in your head. One is that it forces you to think through the entire customer journey, like, what? How are they signalling their intent. Like, one idea is like, hey, how do I know if a customer is happy with my product? Right? One way is like, they actually purchase it, or they keep using it, right? That's a proxy for how happy they are with it. But as you start to think of that, you also think about like, Oh, I'm actually starting to recognise the limits of data, because happiness is not just like a purchase, right? It's also like, what are they? How do they actually feel? I do a lot of, like event marketing, or I have in the past, and there's always this, like, common problem of, how do I calculate the ROI, the return on investment of a big booth that I have had, like, a conference, right? You know, somebody walks past my booth, they see a logo, and then two years later they might come back and buy my product. There's, there's no way to actually attribute that, right? So once you start recognising that limits data, you start to be, when I started to recognise those limits data, I found that I've had to use both, like the qualitative and quantitative to really round out that flavour of of that report, right? We're talking a little about food earlier, prior to the podcast, right? It's, you can't just have some salt. You need to have a little bit of pepper into it. And sometimes those those qualitative things, it really helps. It really helps get that impact that a number wouldn't normally do, right? Like, one time I was working with a product team, and there was like this, like onboarding process, and there was like a part of the funnel where it's like, hey, here, this is a place where 20% of drop offs occur, right? It was one thing to like show that number, but it's another thing but say, like, Hey, I also interviewed like, 10 customers. Actually went through it. I actually walked them through the process, and one of them just said, like, hey, this product looks sketchy as shit. You should get a designer, right? And that resonated way more than saying, like, hey, there's 20% less drop off at this particular portion, right?

Richard Lee 6:30
Yeah, because I've just sort of published a book called using creativity and data in marketing, where I interviewed, regrettably, not your good self, but 40 or so marketing professionals. And that was like one of the really strong things that that came back, is that our tendency is to overvalue the data that's easy to get a click through rate data that's super easy to get, to get on the phone or organise a focus group or speak to a human being at the point of purchase or the point of use, that is so much effort and it's so expensive, but yeah, if you don't do that, then you can the data is blind to the things that you can't actually measure to your point.

Richard Lee 9:59
Yeah. Absolutely, and there's a little bit empathy there as well. You know, people talk about customer empathy, it's really a lot. I find the job a lot more satisfying when I actually talk to somebody and they they're telling me, like, Hey, this is where it's good, this is where it's bad, versus, oh, here's an aggregate of 10,000 people, and I'm up like, three basis points in terms like, performance.

Tom Ollerton 10:21
Right? Yeah. And it's the more senior person you speak to in data driven marketing, the more they talk about qual I find.

Richard Lee 10:28
Yeah, yeah. They're like, tell me about the customer. The more junior.

Tom Ollerton 10:31
They're like, I've got dashboards. I've got like, APIs are gonna crank everything in and make it all ones and zeros any. But then, actually, once you get up there, they're like, no, just, just go and talk to someone. It's quite interesting.

Richard Lee 10:41
Exactly, all of our job is so digital, but yeah, it comes back down to like, oh, physics and people.

Tom Ollerton 10:54
So your shiny new object is text to code and why marketers should care, right? So what is text to code exactly? Yeah. And why should we be really passionate about this?

Richard Lee 11:09
Yeah, so I work at Google at the moment. I didn't want to come on and be like, Oh, I think AI is, like the shiny object. What does that actually mean at this point? But, you know, I imagine a lot of people that are going to be listening to this podcast have probably used AI at some point. You know, whether it is Gemini, ChatGPT, Claude, and you've probably used it to write an email, right? That's text to text, or maybe you've done some of the more advanced features, like, if you're on LinkedIn, you're definitely seeing, like, a lot of AI generated imagery, text to image, right? There's been a lot of announcements and kind of flash with the new text to text to video stuff, right? From a prop you can create, like, an actual video with like audio, right? And as a marketer, you look at sites like, oh, wow, I can do so much, like creative stuff. I can do so much content. But I think marketers are really sleeping on text to code, and this is something like developers already doing. You might hear of services like cursor oak or copilot, but it's this idea of like using AI to actually not, you know, as a marketer, not at the point is not to help you become a developer, but it's actually to have more like a developer buddy or developer manage, developer mentor to help you do more technical tasks that you didn't think that you could do before, and just that, like five, 10% that you're scraping with the help of AI can make you so much more effective, I believe.

Tom Ollerton 12:35
Right. Okay, so someone's listening to this podcast thinking, Yeah, all right, cool. Let's have it. How'd you what do you get started? What might, am I gonna do? We're talking about, like, launching a new can of beans, and you can do a text to code landing page, or what's a, what's a good place to start?

Richard Lee 12:51
Yeah, and this actually came up recently for me. I'm doing a side gig where I'm doing, like, a direct to consumer brand, right?

Tom Ollerton 12:59
Ooh tell me about that.

Richard Lee 13:00
A little bit. It's like, not some fancy AI startup. It's just simply like, hey, let's create some some jewellery, actually. So I've been working with a friend on this one, so a little bit of a side gig.

Tom Ollerton 13:11
Well, if you have a URL to share, we'll whack it in.

Richard Lee 13:14
We're still working on the name, but we're also working on, like, What's the concept and what's the mood? And this is where this first case comes up. It's like, I just need to create a bunch of images, and I don't want to use documentary. It's actually pretty hard to, like, search through all that stuff. It's like, what if I could just create a bunch of images with AI to help me design that initial mood board, right? So show me different textures?

Tom Ollerton 13:37
And then with the mood board for the brand its visual identity, right?

Richard Lee 13:41
Okay, exactly, yeah, because we're trying to figure out, like, what's the right name, and we need to figure out that mood, because as a jewellery project, you know, a lot of it is in that brand, right? So I started, like, just putting these prompts into chat GPT, and it just became onerous, right? Because I had to create, so I would create a prompt, I get something back, and that would take like 30 seconds, but I needed to create like, hundreds of images, right? So I asked chat GPT, how can I create hundreds of images in a programmatic way? And it actually responded. It said, like, oh, you should write a script. And it was like, Python. I was like, what is that? What's that actually mean? And it and it was very patient with me and explaining exactly what all these terms were. What's Python, what's API? Why should I set up and how, what's it mean to call an API from chat GPT to get an image back, right? So it's giving me like, it's explaining me what the Coding Problem is, and then it actually helped me create and run that initial set of code from like, bare bones I might I don't know anything about coding. Explain how I do this setup, right? So it's explaining me. This is how you instal Python. This is the code that you're going to copy, paste. Uh. Know, here's the information that you need to retrieve, to copy paste into it. And at the end of the day, I was able to, you know, create a relatively simple script, but it would take, it would take him, like, a an engineer, like, maybe an hour to do, but it would have taken me, like, months or weeks or months in order to figure out the basics of like, okay, how do I write this loop? How do I learn Python? Right? And it was able to compress that into like half a day project, which, you know, it's not nothing, but compared to the alternative of like, months of this stuff that would just turn me off, it was actually pretty effective. So at the end of the day, I just had, like, a directory full of images I can just, like, quickly sort through, right?

Tom Ollerton 15:43
This episode of the shiny new object podcast is brought to you in partnership with madfest, whether it's live in London or streamed online to the global marketing community, you can always expect a distinctive and daring blend of fast paced content, startup innovation pitches and unconventional entertainment from Mad fest events. You'll find me causing trouble on stage, recording live versions of this podcast and sharing a beer with the nicest and most influential people in marketing. Check it out www.madfestlondon.com.

Tom Ollerton 16:20
Right, so that will be applicable to someone who's launching a brand or, or potentially in a DTC environment. Perhaps, maybe I'm simplifying a bit here, but what about, what about, if I'm a, I'm the CMO of Samsung or something? How would it work for someone who's in a more strategic level or, is it? Is this more tactical you suggested?

Richard Lee 16:42
And that one is pretty tactical, because it is like a, you know, it's like our own little side gig, so a little bit earlier of terms, this idea of, like, self sufficiency. But if, like, I'm working at, like, my day job, and actually, this happens fairly regularly, you know, I have teams of data analysts. I have more data that I know what to do with. How do I, like, manipulate and analyse all this stuff. And AI has been helpful for that in the text to code standpoint, because it helps me with things like SQL, or even more tactically, like, even if you don't like mess with like the database stuff, imagine just using like Google Sheets or Excel, and you just have like, this really wicked Excel formula that you want to do. Or here's more practical example, I've been working with, like, another company for for leads, and at the end of the day, I get like, a data sheet back, right? It's a CSV, comma separate value file, and it's not in the right format, right? So old me would have been like, oh, I need to go back and write some weird query to extract the email address out of the single line, right? And but with, with something like, like, these new text to code bots, they're able to create those for me. So I'm able to feed that CSV into the code and be like, hey, write me a formula to extract this particular step part of the data. So at the end of the day, what this means is, like, I'm able to clean the data a lot easier and faster by myself, as opposed to, like, talking to data analysts, like, Hey, I have this sheet. It has 5000, 20,000 rows. Can you process it for me? This is something now I can do by myself, right?

Tom Ollerton 18:31
Okay, so it's starting to become super clear to me now. So I think what the pain point you're solving here is that, like, throw it... when that throw your hands up in the air moment: I don't know how to do that. How does that work? But actually you're just going, right, well, I'm gonna, I'm gonna go into it as a complete newbie and go, right, okay, I... Excel terrifies me. I'm not an Excel person, but now you're giving me the confidence so I could just go right. I need this data to do this thing, write that for me, and then you can go with it, as opposed to being like, right? Okay, I'll see if I can admit to someone that I'm not very good at Excel, despite running a business.

Richard Lee 19:07
And then sometimes it surprises me how well it does. I've done things like, I've taken a screenshot of like, Hey, this is in my sheet. Now, this is what I want it to look like, and it's able to write a formula to intuit like, what is it that I'm trying to do? And I'm just being, like, literally, JPEGs.

Tom Ollerton 19:24
What are you doing next with this? What do you want to try that you haven't had the chance to try yet?

Richard Lee 19:28
Yeah, I'm more on the geeky side of things. So I've actually been doing a lot more thorough things, just to, like, experiment more on the on, like, where... let me take that back. So I work at, like, a large AI company, and a lot of this buzz is around this term, like AI agent. AI agents. Have you heard of this thing?

Tom Ollerton 19:51
All I hear about is AI agents. It's a label that I think everyone is very happy to to use.

Richard Lee 19:59
Yeah, and. And I have no idea what it actually means either, or at least, what's your guess for me, actually, you go first.

Tom Ollerton 20:07
It's personification, right? Like you say, LLM, you're like, what's a large language model? LLM, and large language model mean absolutely nothing, right? It is classic TLA, three letter acronym stuff, and so agentic or agent based users of AI make it, it's a personification or humanization of what is actually just an LLM or a or a, you know, text to speech, text to visual, text to chat to go, All right, I've got a I've got a menu writing agent. No, you haven't. You just got a mini GPT, whatever. So it's, it's a simplification, or a re pitch, or a spin of what is actually just a derivation of one of the larger LLMs.

Richard Lee 21:02
For me, an AI agent is simply an automated programme that uses AI, or that's pretty much, I know that that sounds like overly simplistic...

Tom Ollerton 21:17
But agent's like James Bond, isn't it? It sounds so crafty and clever.

Richard Lee 21:20
It sounds so cool, but it's actually a very simple concept. Think about like all the computing for the last 20 years. It's just an automated programme, right? And now you're just sprinkling AI over it. And I was able to come to that conclusion because AI was able to help me build my own AI agent, like that was like, the ... So like, what is actually an AI agent? Well, what if I actually just, like, built one, right? So the way that I did this was, you might have heard these workflow automation platforms, if you might have been bombarded by like these ads, but they're things like Zapier or n a n, right? These are things that. Example of what these things will do is somebody liked my comment on LinkedIn that triggers something in Slack. So then I get a message in Slack about it, right? So the more advanced case that I want to do was I wanted to create a newsletter creator. And the way that I wanted to do this high level is I want to scan a bunch of interesting websites that I think are interesting, and I want to summarise them on a regular basis. So the basic workflow of this is that AI was, text to code was able to help me do was set up this, this automation workflow. So I had these, hey, chat GPT, I want to do this workflow. Help me break it down into discrete steps that I can use with Zapier or an ADA that was like my prompt, right? And it's telling me, oh, identify the different sites that you want. Oh, here's how you can actually pull the data from those sites using something called, like an RSS feed, which has been used for a while, right? Yeah. Now you can use, now connect that to the node that is a Gemini node, and tell that Gemini node, get, feed it this prompt. You are a newsletter agent. Your job is to summarise content so and then output that into a spreadsheet, right? So it's it's helping me break down all these discrete steps, and helping me, like, troubleshoot and actually interact with these tools, right? So at the end of the day, I now have, like, this programme up on Google Cloud somehow, and it's just scanning these sites that I like. It compiles them into a database and basically a spreadsheet. And then every week, another AI note summarises that information and writes me an email saying, like, hey, here are the top 10 things that you need to know about media entertainment this week.

Tom Ollerton 23:59
So what I love about these chats about agents, ah, I hate the phrase, is that every, every time I speak to someone who's experimenting like yourself, you're like, yes, why am I doing that? But I had a really odd moment today. I was reading a website called performance marketing world, obviously, we operate in that world. And I was reading an article coming what it was about, and it was so obviously, wasn't written by any AI at all. And it had critical thought behind it. It had art and effort and taste put into it. And you can... it was so obvious now, like, I've started to see these two kind of worlds, you know, like just critical thought and craft and thought is so obvious now, unless I'm an idiot and it's as obvious as generated stuff is being very different, and obviously I I tend to occupy, hopefully, the space in between the two. But I'm curious to know in your little journey of RSS and nodes and spreadsheets and cloud and Google Cloud and so on, how are you, how are you making sure that you're putting critical thought and love and craft into this? Or is that dead to you?

Richard Lee 25:15
No, that's a very good question. I get that. I also get the annoyance almost when I do get something that's clearly completely AI generated, so I use it as a way to get me 80% of the way there, like in explaining this workflow that I just described, like most of the work is reading websites like and maybe, like, hundreds of articles a week, right? I don't have time for that. I need a summarization. So what I have to do for this newsletter is that I have it output. Here's a selection of different headlines that are interesting, right? So I'm no longer just a copywriter. I'm an editor. I am curating what is what is it that I think is going to be interesting to my peers, having all this information just consolidated for me, right? So that's part one, what's the most important information? And then two, like, especially as a marketer, less words is always better than more words, right? And I don't think that AI has been able to do that level of distillation that we still provide. You know, what is the most important thing? How can I boil this down to, like, two or three words, right? So I think that there is a lot of, when you look at newsletter, you know, there's, there's work in, like, that wall of copy, but there's so much work that went into that prior, right? So I think what this allows you to do, like, all this AI stuff allows us to do as markers, like, really focus on, like, what's on, the most important things, and all the stuff that would have been work that you did for yourself, or work that you're negotiating with, like an engineering team for you, a lot of that stuff can go away so that you can focus on what's important and what's fun, right?

Tom Ollerton 27:07
Yeah, I hope you're right. I did some deep research the other day was it was in chat GPT, and it probably saved me a day's worth of work, but I didn't, then write to my business partners and say, I'm taking tomorrow off because I've, you know, I just went did something else. I mean, like, it's, it's not, I haven't seen any evidence that any of this, AI, is freeing us up to take time off. It's just freeing us up to do more work.

Richard Lee 27:38
Yeah, yeah. But hopefully it's different type of work and more meaningful work for us, right? I don't want to spend all my time writing SQL queries...

Tom Ollerton 27:46
But then in but then I'm sorry this. I'm gonna try, and that's on a podcast. I feel I was just having a chat. But at some point you you automate or optimise tasks until you reach the next thing that can't be automated and so and then eventually your your Yes, like, it's like, you know, I can send 100 emails instead of 100 handwritten letters, right? Do you know what I mean? Like, it's that email is a giant time saving automation. But once I've done that, then it's like, well, there's a whole bunch of things that we cannot automate and process-ise... Can't believe I'm really waffling here, but eventually we're going to butt up against the thing that you can't optimise, and then that's what our diary will fill up with. And they're not necessarily fun or interesting. I don't know maybe...

Tom Ollerton 27:46
I don't know if it's optimistic or pessimistic, but I think this AI stuff is going to help sort us and distil us to what is going to be most impactful and important.

Tom Ollerton 28:23
I think it's like, I had this experience with my daughter a couple years ago. She's seven now. And then I said, Look at this thing. It's called Chat GPT. And I was like, tell it to create an image of a rainbow in space with a banana, rainbows and all that kind of stuff. And it went Bing and she didn't care. She had no interest at all. So you can have all of these people arriving in the workforce in two, 5, 10 years' time, where this is just normal. They're not going to be walking around going, Oh, I did text the code or text the voice, sort of, you know what I mean, like, it's we will arrive at that new normal, and then it will just be normal, whereas at the minute, we're the pioneers of it. Anyway, look, Richard, we've gone horribly over, because you've been very interesting. So thank you so much. And you've got me talking on one of my favourite subjects. So if someone wanted to get in touch with you about this stuff, where and how should they do that?

Richard Lee 29:30
Yeah, probably the easiest way to catch me is on LinkedIn. Just type in, it's Richard. Or if you're on Google, if you type in Richard Lee, MBA, I'm usually the first result that comes in, but yeah.

Tom Ollerton 29:43
And what makes a message that you'll respond to?

Richard Lee 29:47
What makes a message that responds to me? Ask me how I can help you?

Tom Ollerton 29:52
Okay, Richard, we'll leave it there.

Subscribe to our YouTube channel

Listen to our ‘Shiny New Object’ Podcast on Apple Podcasts, Spotify and Soundcloud.