The PPC Show, Google Ads

The Rise of the Technical Paid Marketer [Podcast]

Welcome to episode #72 of the PPC Show, where we interview the best and brightest in paid search and social media advertising. This week we're joined by Gianluca Binelli, founder of Booster Box Digital.

In this episode, Gianluca joins us to talk about the rise of technical paid marketers. Stay tuned to learn:

  • How he got started working at Google
  • Why he decided to learn how to write scripts and Python
  • Four products he built to solve and scale client issues using code (Zuul, Caligola, Cherry Picker, Pincer Move)

Listen to the Episode

Gianluca Binelli

Gianluca was at Google for 6​.5​ years working on different team within AdWords. First as Product Specialist, he then went on to manage online marketing for Google’s products in EMEA. Gianluca also served as ​Capital G's Advisor, helping start ups with their online marketing in Google’s own late stage investment fund. He is the founder of Booster Box, a performance agency specialized in scientific marketing.

Follow him on Twitter.

Show Notes & Podcast Transcript


JD Prater:                     Gianluca, welcome to the PPC Show.

Gianluca Binell:            Thank you, JD, for having me.

JD Prater:                     Yeah, I'm pretty excited that you came all the way from Italy-

Gianluca Binell:            All the way. All the way.

JD Prater:                     Just to come on the podcast.

Gianluca Binell:            Just for you. Just for you, yeah.

JD Prater:                     No, I mean Gianluca was in town here in San Francisco visiting some clients, and reached out on Twitter. And I was like, "Oh, man, yeah. Come hang out." And so, we hung out on Wednesday. And now we're here recording again. I got more of his time on Friday. So, I'm excited. Thanks for coming back.

Gianluca Binell:            I'm very excited. Thank you for having me.

JD Prater:                     No, we went and grabbed some coffee as you do with an Italian.

Gianluca Binell:            You must.

JD Prater:                     And so he started explaining some of these really cool tools that he's got working on. So, I'm gonna let him kinda share some of these things with you. But first, Gianluca, tell the good people more about you. Maybe add some, the CliffsNotes version.

Working at Google to Booster Box Digital

Gianluca Binell:            Super. No, no, very, very happy to be here. This morning I was on the phone with my Mom. And I told her that I was going to be interviewed by a PPC podcast. She got all excited. And then she was very, very excited, but after a while she understood, BBC, the British Broadcasting Network. It was a little bit of a bummer then when I told her. "No, it's not BBC Mom, it's a little bit of a narrower audience." But, I am excited anyways, even if my Mom doesn't appreciate the PPC community as much as I do.

So, I've been in the PPC game for I keep saying 10 years for the last 12 years actually. But, can I stay at 10? I am originally from Italy as my accent, appearance, and name give out. And I worked at Google, always on the other side in different roles. First, I was product specialist. The guy that is whispering in the ear of your account manager what is the product for you, which I think is one of the most frustrating job in the world because you need to tell people to tell people to tell people to do something.

So, after two years now, of probably doing that, I developed all these theories. This product is perfect for this type of client. We should definitely do that. And more than that. And then, I rolled up my sleeves and moved to the marketing team where I was actually seeing the scope of my theories. And challenging some of the ideas that I had. And probably 90% of the things I theorized were wrong, but it was the 10% that worked out, which got me, they didn't fire me. So, they kept me on the team for a while. And it was very exciting because I joined the marketing team. You always see from outside. The online marketing at Google, at least I felt it was a super mega sophisticated machine. Turns out there are humans as well there. So, nothing is perfect.

And when I joined the Google team we were three or four people. When I left, we were 54.

JD Prater:                     Oh, wow.

Gianluca Binell:            Yeah, yeah.

JD Prater:                     That's massive growth.

Gianluca Binell:            Very exciting ride. And in that journey, Google does this 20% thing, which basically is from six to nine and weekend. That's 20%, at least it was for me. And I worked with the guys at Capital G, the VC firm here in San Francisco. And so, that was very exciting because I helped them in figuring out the growth thing there. Figuring out what our opportunity for paid media for some of the companies that were investing into it.

And that was kind of putting all the dots, connecting the dots backwards. I figured out that even big companies, or unicorn, or supposedly unicorn, not necessarily were taking the full advantage of the huge amount of data there is out there. So, I bridged the two things together. The expertise I gained in the last 12, well, we're gonna stick with 10 years in the industry. And I managed to be part of the team that became more and more sophisticated on all of our advertising in India. And the need in the market for helping companies making a better sense of the data they have available.

So, I went on into building Booster Box where that's what we do. So, we are a scientific performance market agency. We're focusing on large, medium-large international clients with big problems. So, they are more challenging to work with. And by big problems, I mean big data sets really. And we are based out in Tuscany, where else. Silicon Tuscany, the middle of nowhere.

Where the most, again, the biggest challenge that I face, we've been doing that for 18 months now. And definitely the biggest challenge I had to face was setting up the internet there in the office. But, now we have internet connection. We're good to go. And there's a very exciting ride to be, grab a glass of wine in the evening, and having an impact working on big problems. So, that's what we do.

JD Prater:                     Nice. I mean, it's kinda hard to beat working on big problems, especially with a glass of wine in your hand.

Gianluca Binell:            It helps.

Rise of the Technical Marketing

JD Prater:                     It really helps, yeah. Well, that's really cool. So, we got you on the show here today to kinda talk about the rise of the technical marketer. And then, talk about some of the tools that you have built because of that. So, in particular, I've noticed within PPC world, in particular, the amount of coating that's almost needed now. Not necessarily needed, but everyone is getting more savvy. And so, if you don't know how, it's almost like you're getting left behind, right? And so, you see Fredrick Valley, you got Daniel Gilbert writing these amazing scripts that we're all using and some people are like, "I don't even use scripts, right? I don't even know what that is. What's going on?"

But, you're able to take that even a step further. So, what have you seen within the community with people getting more technical? Do you think that that's true? Yeah?

Gianluca Binell:            Yeah, absolutely. There were a few articles coming out in the last few years in this direction. The rise of the technical marketers. And I think overall, that's a reflection that marketing has changed in the last 10, 20, 30 years. And the fundamentals probably are the same. Meaning find the user, find the magic, connect the two, right?

But, if you want to use a sport analogy. I was watching the Viking game yesterday night. It was super exciting. That was the first football match I saw in my life. If all the matches are like that, man, I'm hooked. And probably the fundamentals of football 30 years ago, 40, 50, are the same I guess. I don't think the rules change, right?

JD Prater:                     Mm-hmm (affirmative).

Gianluca Binell:            But, the technology that is there now, right, even the materials, the shoes, the thinking, the process, the food, the training they do. Everything is changed. So, it is a different sport. And I think the same analogy, it works for marketing, right. Of course, the goal, it's kicking the bowl in a certain direction anyways, but everything has changed. And I think the complexity of data that we have on the user, and the complexity of data that we have on the product makes the same job of finding a user, finding an edging, connect the two, much more exciting.

In all that complexity, I think that unlocks the need for a different type of skill set. So, just being the best and wining and dining probably is not anymore the keying element. You also need to be good at making sense of a complex data set. Or, knowing a little bit of statistics in order to get your head around on which one is the winner in A/B test, etc., etc.

So, I think definitely there's a trend that is happening. It's been happening for the last few years. And Dan and Frederick, the people you mentioned, and to be honest, I'm just very grateful to be playing this game now in a contest where we have so much access to data. Like you just said, "At the end Twitter will eat you up, and will be nerding out thinking about naturally, everything in the world."

And I think we're so lucky to be allowed this time where we can reach out to anyone in the world and we can learn statistics from MIT for free online. Tell that to our grandparents or our parents, right?

JD Prater:                     Yeah.

Gianluca Binell:            So, I think there's no excuses, right?

JD Prater:                     Right.

Learning Python and Writing Scripts

Gianluca Binell:            The world is going in that direction as independency game, as marketing. I think it's our responsibility to just embrace the fact that technology is rising and it's just like gravity is happening. We're just not gonna fight it, we just need to embrace it. And make sure that we are riding that wave.

JD Prater:                     Yeah. I think part of that too is you're unlocking different abilities too. So, we're going to get into some of the tools that you've built using technology, and using this code. But, how did you get started writing scripts or learning Python? How did you get involved?

Gianluca Binell:            Okay. First of all to give you an idea of how good I am at Python, every time I make a script and I give it to my guys they're like, "Aw, that's sweet." They trash it, and they rewrite it from scratch. So, on a scale from one to nine, I'm definitely at about four, which I'm very generous with myself.

JD Prater:                     Well, I'm at a zero. So, you're doing better than I am.

Gianluca Binell:            I don't have a calculator now, but four is more than zero.

JD Prater:                     Yes, that's right.

Gianluca Binell:            So, when I started at Google I was literally in between two guys that were two of the best people I know in Google Analytics. And they were talking to each other. The only option not to learn was listening to music 24, seven, which I didn't. And so things just sunk in. At one point, there was this tool, a script that they were featuring in Google spreadsheet. And then I started playing around with that. I remember the first thing I did with that was a stupid thing to automatically send emails to instead of taking notes for phone calls for my colleagues. When you call your account manager, it's not his desk. And somebody else is picking up the phone. As I was lazy, taking notes it wasn't part of my particularly favorite mundane task. So, I start playing with the script.

So, that was the get-go that got me hooked in finding. And the first time you write something from scratch. And then you works. It's just magical, right? The first time you see a little word. So, I start with that. And then, I thought to myself, "Hopefully, it'll be more on theory, a little bit on code academy." And I try also to brush up a little bit more my sophistical skills as I mentioned that MIT has an online course. And there's plenty of stuff online. I mean, everything, it's everywhere for free. So, that's what I'm saying. No excuses on that.

And yeah. So this is where the two main elements knowing, getting a little bit better in coding. The theory was for me, Code Academy. Then the second bracket was definitely playing with amazing resources that are out there in the community. So, on the Google developers side there's so much, so much stuff that it's unbelievably cool. And then, Frederick, all the people that you mentioned earlier, are sharing great stuff online.

But, I think the theory is the first pillar. The second one is figuring out what are the scripts that are out there and start playing with it. But, the third pillar, for me is people. Just for instance, at Booster Box, my company, now we do lectures. So, on a weekly basis we invite an external guest that is sitting with us. And it's just explaining their job. And these people are people, investment bankers, or data scientists. They're not necessarily are in the PPC game.

But, building that connection and talking with someone that is solving a similar problem in just a different environment, right? It's really helping me and helping my team accelerating the level of knowledge. Even when it comes to scripting, right? So, having the opportunity to reach out to someone, a human being, a mentor, a person, just to bounce off ideas and say, "Listen, I'm using this approach. It's not working. What do you suggest?" That is definitely key. And even on that one, we live in a fantastic time. We can reach out to anyone on Twitter, and Instagram. And the worst case, they're just gonna say, "No," or not pick out that message, right?

So, definitely feel free to reach out. And if you have tough questions on Python, reach out to JD. If you have easy questions, reach out to me.

JD Prater:                     That's right. So, I wanna kind of rewind some of that because I think one thing you didn't intentionally, but you glossed over. So, you actually started with a problem. And then you, "I'm lazy, right, and I wanna solve this problem." Right, whatever it is. But, I don't necessarily think it's even laziness. I mean, you actually took the action to go solve the problem in a programmatic way, which is fantastic. So, that's what I love about your tools, which we're about to get into.

It's really starting with, "Man, wouldn't that be cool if ... " Or, "I have this problem, the client wants this," and then it's, "What can I do to mitigate that, or solve that in a way that's scalable so people aren't necessarily doing this manually every day, every hour, whatever it is." But, there's something running in the background, right?

And I think the other part to that too, is you also took action to go and learn. I think the people that you might say is lazy, I think it's the opposite. I'm lazy. I have the exact same access to these resources that you do, but I'm not out there learning, right?

Gianluca Binell:            Well, I guess, yeah.

JD Prater:                     And so you're gonna push me to go and learn it this year, 2018. I'm gonna do it.

Gianluca Binell:            It's happening. This year is mobile. I think, look, I think it's important that each one of us is focusing on the right priorities. And it depends at which point people are in their career or start with a basic example, right? If you are an agency or if you are solo, then definitely the skill set that you need to build is entirely different, right?

So, I don't think that developing a certain skill is mandatory for everyone. I do think though, that not being afraid of that, that's it, right? If you're not afraid of seeing a script and playing with it, you're already 99% better than anyone else, right? So, I think the key's just not being afraid of testing something new, and whatever. That's probably applicable in anything in life.

Tool #1: Zuul the Gatekeeper

JD Prater:                     Well, cool. Yeah, we'll lets ship into some of the products that you built. So, we're gonna talk about four of them specifically because as we were getting copies, talking about them, I was like, "Whoa, whoa, whoa, whoa. We gotta have that on the PPC Show." So, which one do you wanna start with?

Gianluca Binell:            Yeah. I think the first one I really like is what we call Zuul the Gatekeeper. In honor of the marvelous movie, Ghostbuster, that I strongly encourage everyone to watch again.

JD Prater:                     The original, right?

Gianluca Binell:            The original.

JD Prater:                     Of course, okay.

Gianluca Binell:            So, Zuul the Gatekeeper was a very, very, nasty monster. I don't know how to define it, probably, whatever monster, I think is a good definition. Our Zuul is much friendlier. And he's basically a tool that helps us mitigate the risk of adding ads of bringing in the wrong placement. Let me start by the problem as you mentioned.

JD Prater:                     Yeah, yeah, yeah.

Gianluca Binell:            Everything starts from there.

JD Prater:                     It's a great problem.

Gianluca Binell:            Usually, tools start from people nerding out in the mini kitchen or getting a wrong email from the client. This is a story to start with a second scenario. So, classic email, Friday night. It was summer last year. And I still remember where I was. And I get an email from a CEO of a very large company, which is still one of our clients, fortunately. So, it's a good story.

JD Prater:                     Good. Zuul's working.

Gianluca Binell:            Zuul's working. Of course, somebody took a screen shot of one of our ads appearing in the wrong place. And you know, Murphy's law is true. We've all been there, getting one of those type of nasty email. And there's nothing that you can do, right, to avoid. To put yourself in a situation where you're 100% in control of every single placement in the world.

So, you might remember a few months back. I think, it was May last year. There was this backlash on YouTube. Everybody was against Google. We shouldn't buy GDN.

JD Prater:                     Pulling money out, yeah.

Gianluca Binell:            Google's not doing enough to protect us. I strongly disagreed with that approach because I think it's our responsibility to, as an agency in my case, but as players in the market. It's our responsibility to make sure that we're using AdWords, and Facebook, and all the amazing tools that are out there responsibly. These are self-serving platforms, right?

So, it's how well the self-service element. There's a very beautiful Brainlabs article, Daniel Gilbert wrote something similar sometime ago underlining that the self-service element in the word self-service tool implies that it's agency's responsibility to take that ownership.

So, I think that pulling spend out of GDN and Facebook, and just stop playing, probably is not the best idea. So, from one side definitely, you cannot control fully 100% where your ads will appear, and probably, our own ads will appear in the wrong place maybe in the future. But, we started putting together a methodology in order to make the game this risk. So, from the basic things is loading, adding a blacklist of websites where you don't to appear, adding a list of negative keywords that you want to exclude, excluding side by topics, etc.

What do you do if you don't know who the bad guys are, right? So, what we did was literally, taking a list of all the bad keywords, a gigantic list of negative keywords. So, imagine tax fraud, pornography, keywords you would never want to be associated with. And we build a script that goes and pulls the placement where your ad appeared yesterday.

So, at midnight, close of business, the scripts calls AdWords. And it's, "AdWords, can you please tell me the placement where we appeared yesterday?" That's phase one. Phase two, we go and scan every single element on the page of each placement. This is not really every single element because we cannot scan images, we cannot scan videos. We are only scanning text.

We tried to scan videos, to be honest. We used the Google Cloud Video API, but I think technology is not there yet. So, the results that we got by submitting videos there where it works like this. You submit a video to the Google Cloud API, the machine analyzes it, and it tells you, "In this video, you find A, B, C."

The level of variety in the output that you get is two probe probably, because you only get, "In this video, you find a man with a white T-shirt." Well, thank you. So, if you don't have men in white T-shirts in your negative exclusion, that's not really gonna help you. So, videos definitely are a big limitation.

But, on text we managed to figure out that by first of all counting the appearance of specific keyword in the text contained on the website, that was already giving us an indication if the website was about, lets say, tax fraud, or pornography, or something you don't really want to appear next to. And then the second element would be in the regression model that identified if a certain word is repeated more than once within a space of 2,000 characters. Then, we overweigh that element. And that really helped us identify with a certain level of accuracy websites where we appeared yesterday, and we don't want to appear tomorrow.

So, what the script does at midnight, goes and calls the placement. Then, from midnight until four in the morning start measuring. If analyzing the website, and trying to identify if these sites were bad guys or not. And then by four in the morning, takes the decision and push that back in to AdWords. So, that the following day, your ads are not appearing on that website.

Now, this is definitely not perfect. It doesn't capture 100% of the website, and is ex-post. So, your mistakes already was made. And somebody could have already taken a screenshot on the previous date, and sent it to the CEO. So, definitely, this is not a 100% guarantee of success. But, is a step that is helping us mitigating the risk of operating in the wrong place.

So, that's why Zuul the Gatekeeper is the protector of brand for us. And then the icing on the cake on this story was the fact that we're able to put the script into a tiny raspberry pie. To be honest, if the script works in the Cloud, it's much faster. But, the fact that it's on the raspberry pie, sitting in your machine, right? So, you plug it on your computer is giving us people, humans, a concrete sense of protection, right? So, at midnight the tiny raspberry pie in a very noisy way will start wiring out, calling the placement, and that's where you feel protected. So, that's really something that is helping me sleeping better at night. Thank you, Zuul.

JD Prater:                     Again, you started with the problem. The problem was it, we're talking about brand safety. This is more and more important that we've been hearing this specifically, on YouTube in the last year. And so, I thought it was a really great way of using code. Making a tool that is doing something programmatic to where you don't necessarily have to have someone in there manually checking every single placement. And of course, there is that additional layer that goes in and reviews.

Gianluca Binell:            Of course.

JD Prater:                     But, being able to go in and say, "Hey, check out all the placements." And then you're crawling the site for the text. You've assigned weight to it. And this is all done while you guys are sleeping. So, by the time you guys wake up you guys have an answer, right?

Gianluca Binell:            Exactly.

JD Prater:                     So, fantastic. I love that story. I love the tool. And I haven't even seen it, but I just think it's a great way of using code to benefit PPC or so. Yeah, good work on that one.

Tool #2: Pincer Move

Gianluca Binell:            Absolutely. A similar logic we try to replicate a singular approach of scanning content in order to recycle existing keywords. Let me start by the problem again. So here we often work with clients that are an established business. They've been in the PPC game for a while, maybe four, five, seven, years. So, you've plenty of data, right? And sometimes these businesses are maxed out. So, it's not as easy as . We forgot to add extension. That means with demand you need to figure out what are, in a way you can call it, lateral thinking. And try to figure out, "Okay, we have this type of information. How can we reuse this information in a different context?" So, what we did in this case was try to figure out, we know where these keywords that are working very well for us. How can we use this keyword in a different way and not simply buying ads on word search.

So, we call this tool Pincer Move. I'm a geek on Roman history. So, the Pincer Move is this move that Hannibal and the Romans were fighting. I don't know the year to be honest. So, the Romans developed this technique on laterally attacking the enemy. So, not where the enemy was stronger, but on the sides.

So, we started by that type of approach. Already is the following. Let's say you have a keyword that's working super well on Google search. Let's say the keyword French Press café it's working amazingly well for your coffee commerce. And you're winning first position, mazotoff, great for you. But, you really want to squeeze out more traffic from that keyword.

So, we take that keyword. We crawl the top 10 results on Google. And then, we automatically submit these results as part of GDN. So, we will be targeting specific places, the specific placement, sorry, as GDN placement of the results that are appearing on the top page.

In other words, when somebody is clicking on that keyword, is searching for that keyword, if they click on your ad, great. If they click on your competitor's ad, there's really nothing you can do about it. But, if they click on our organic results, this is a zero sum game, right?

JD Prater:                     Mm-hmm (affirmative).

Gianluca Binell:            So, the rest of the CTR goes to someone else. So, if they're clicking on organic result, and if that organic result is selling ad inventory, you wanna be there. Now, of course, some of this placement will be Wikipedia, or competitors, or placement that are not part of the Google Display Network. But, if someone is selling that inventory, you want to be there. So, the Pincer Move is a tool that is helping us. Once you have figured out that these keywords are working well for you on search, you can try this approach to scale that beyond search exactly with the same keywords.

JD Prater:                     Yeah, I think this is another really smart tactic, right? So, if you're not able, or I should say maybe you're not ranking organically yet. Maybe you have an SCO team that's really trying to get you in that top 10. And this is something that again, you can use as long as that organic result is part of the GDN network. And again, it's a programmatic way. You're scanning it going, "Okay, cool. Two of the 10 are. Fantastic. Automatically add them to this ad group. And we're gonna be bidding on them at this price," right?

Gianluca Binell:            Exactly.

JD Prater:                     Fantastic.

Gianluca Binell:            If you're lucky enough also to have direct media buying.and the client has that type of contact, or you as an agency are covering that as well, this is also a great way to form your direct media buying team to pick up the phone, and start calling those placements.

JD Prater:                     And this is, I mean, I think this is another good one, reviews, right. So, French Press coffee reviews, right. And we all know there's a ton of review sites out there. And we know that a lot of those review site do sell inventory. And so, again, maybe you've got the top position. And now, they click on the first organic, but you're also showing an ad there, right?

Gianluca Binell:            Yeah.

JD Prater:                     I mean, you're just ... Great job on that one. I really like that one. So, that one is called Pincer Move.

Gianluca Binell:            Pincer Move, yeah.

JD Prater:                     So, you've got Zuul. You've got Pincer Move. What else you got?

Tool #3: Cherry Picker

Gianluca Binell:            We also have something that we call Cherry Picker. It's funny you mention reviews. The place for reviewing the word is YouTube. So, you guys, if you've ever been in touch with someone from Google talking about YouTube. YouTube is the second largest search engine in the world. We heard that sentence. The problem is when you try to buy YouTube, okay, fine. You can buy remarketing. You can follow people when they are watching a YouTube video with your French Press coffee product. Great. And it works from a performance perspective, right? But, trying to get YouTube to work beyond branding from a performance perspective aside for marketing is quite difficult.

JD Prater:                     It's tough. It is.

Gianluca Binell:            Now, there are 46 million, trillion, zillion, thousand of videos uploaded every nanosecond on YouTube, something like that. Probably a little bit more than this number, by the time you are finishing this podcast. So, definitely strictly from a Google perspective, but from an advertiser perspective, trying to figure what is the content of the video, right?

JD Prater:                     Mm-hmm (affirmative).

Gianluca Binell:            So, what we do is, what Google offers, is the opportunity to buy keywords on YouTube in terms of what is used to be called, through the in search. I don't know if it has a new name. But, the problem with that approach is that these keywords are broad. So, definitely broad keywords are something that were scrolling 2006. And since then, we all figured out what we shouldn't do.

So, what we are doing is actually trying to cherry pick this specific placement. So, very similar to the Pincer Move logic. Once we figure out that reviews are a thing for e-Commerce selling French Press coffee, then we will take the keywords. We will match them with unboxing, where were you, etc., etc. And then we'll have a bunch of keywords that are suitable for YouTube. We go and plug into YouTube every single keyword and then we'll extract each specific YouTube video. We take the YouTube videos and then we target them specifically with a TrueView in-stream.

So, at this point from a user perspective, you're about to watch video on skiing in Aspen. And if you are skiing in Aspen, you want to be in front of the eyeballs of a user that's about to watch that video. So, definitely this is something that for our performance, advertising, perspective, has been proven to be very successful for us. And it's something that we're putting out there in the world. Giving back.

JD Prater:                     Yeah, so this one with all of these two, we'll have links for all of you guys in the show notes to go in and check it out. But, this one he does have the code that we'll be sending a link to. There's also a slide chair presentation. But, this one I thought was really cool within YouTube. So, if I'm looking for French press reviews and we maybe have an ad on that one as well, but also we're scraping the YouTube organic. He's pulling out that YouTube video ID and adding that as a placement for his ads as well.

Gianluca Binell:            Correct.

JD Prater:                     So, that's another brilliant move of utilizing organic. And using it again, in a programmatic way of writing some code to say, "Lets scrap these lists. Let's pull this out. And then, lets start bidding on this without having to go in and manually type in all of these keywords." Keep a list of all of these YouTube videos because no one has time for that. I mean, that's probably hundreds of keywords, thousands of keywords. And you're able to do this in a way that's really fast, really scalable.

Gianluca Binell:            Exactly.

JD Prater:                     So, saving that time. I love the automatic part of that. The automation stuff.

Gianluca Binell:            Yeah, we really try to strive for automation as much as we can. I think we mentioned earlier, being lazy is a big portion of that. And the second thing is being process oriented as much as we can. The reality is that the way I see this job growing in the future is really all this stuff is gonna be automated anyways.

JD Prater:                     Sure.

Gianluca Binell:            So, if we drive automation better because that will allow us to move faster and closer to the strategic part. Linear data doesn't have to be figuring out what are the negative curve terms from the SQR or fixing those bits, right? But, you can sit and think on strategic stuff with your client on how you double the revenue next year, right?

JD Prater:                     Yeah, and that's all. That's how you don't lose your job to automation, right?

Gianluca Binell:            Exactly.

JD Prater:                     If your job is going through looking at YouTube placements and seeing how they are ranking, I mean ...

Gianluca Binell:            There's something wrong with your life.

JD Prater:                     There's something wrong, man, because he already has a Python script that's going to put you out of business. Let's go to the last one.

Tool #4: Caligola

Gianluca Binell:            Again, Roman history.

JD Prater:                     Thumbs up or thumbs down, right?

Gianluca Binell:            Thumbs up or thumbs down, yes. So, Roman history. We call this tool, Caligula. So, the idea here is, is the following here. You know when you set up a brand new company and you're putting the keywords. A keyword that potentially has a second meaning. So, lets say the keyword panda because you're selling panda, an Italian car. Let's say it's a car brand, right? You're selling cars, lead generation for cars. And you are very happy. You close your computer. You go home. The day after, you go in the office. You open the computer. You see the SQR and, "Oh, my gosh. Panda is also an antivirus and panda obviously, is the animal." And you forgot.

So, how can you not have to wait to make a mistake in order to figure out what negative keywords you want to add in your campaign. So here, really there are a few steps that one can take. Obviously, you can go proactively on the keyboard plan mill, plaguing your keywords, trying to run an engram on the output, and you will see that panda is also an antivirus. So, you will add antivirus as negative term. And that's easy enough to do for terms that you know are ambiguous already like in this obvious example.

But, if you have plenty of keywords, if something is not obviously, ambiguous. How could you predict, which terms will give you trouble from a search and report perspective. So what we did was literally trying to search each one of these terms on Google because we are insane. But, obviously, we did that with a bot.

So, in this case we're talking about a very large e-Commerce selling thousands, and thousands, of products. These products, the brands of these products are wines. So, a wine often in Southern Europe is a family name. So, imagine Binelli wine. How cool is that? I should branch out.

JD Prater:                     Binelli.

Gianluca Binell:            Yes. And so, how could you, especially on family names, right, try to discern if a certain name has also a secondary meaning beyond the wine or whatever you're selling. So, in this case, we went out on Google and search every single one of this term. So, at this point that for Binelli wine you have these 10 results.

Okay, so that's our first step. The second step is we will go and match the content of the H-1 and the mental tuck. So, the least thing that you get on Google versus a positive library. So, a library that contains keywords that are describing your product. So, in this case will be wine, winery, etc. So, we will at this point be able to assign a score to each one of these listings. And this is binary, right? That's why Caligula, yes or no, thumbs up, thumbs down, because we will have zero, one, scoring to each one of these H-1 and mental tuck.

So, lets say the first organic listing on Google is about Binelli Wineries that contains the word winery. Great, one. This result is definitely relevant to our gain, to what we're selling. We go on with the second result, the result, etc. And at this point, we are able to give a final score to each one of the terms.

So, numbers, high numbers. So, something close to 10, probably very highly relevant. Terms that are closer to zero. So, on the lower end of this spectrum, probably they could be potentially ambiguous. So, at this point you are able to isolate terms that are potentially ambiguous. And the real icing on the cake here, is the fact that there's a second part of the script that scans also the description. And then runs and engram on that. And gives you back suggestions of negative keywords.

So, let me give you the panda example. Let's say you will plug in the keyword panda on Google. Google will tell you, "Okay, the first result is panda antivirus. The second result is panda zoo whatever." So, clearly, none of these two results are referring to panda car. Panda car by the way, I'm pretty sure nobody knows in the world. You guys should definitely Google that. It's horrible car from the '80s.

JD Prater:                     Never heard of a painted car, okay. Look it up.

Gianluca Binell:            If you never heard of a Panda car, your life is good.

JD Prater:                     Okay.

Gianluca Binell:            Use that as a benchmark. So, at this point the first that we have to identify that panda very likely is an ambiguous word. And then the last part of the script will crawl the content, sorry. The H-1 and the mental tuck of the first result. And the second result, we'll run an engram on those. And then we'll suggest to you, antivirus and animal and zoo as negative keywords.

So, that is really cool because this is something you need to run once, only in the beginning, that saves us a lot of money because from now on when you set up a campaign you can go home knowing that the following day in that SQR that you're gonna check out, you're not gonna find horrible, obvious mistake that you should have known better to exclude in the beginning. So, Caligula really cool, thumbs up, and thumbs down in order to figure out which terms are ambiguous or not.

JD Prater:                     Gotcha. I really like that one too. So, I'm sure people listening are wondering, right. So, could it mess up? Does it mess up, right? So, is there an element of human review before it goes live?

Gianluca Binell:            Yeah. So, these are all, you can see them as a human assistance. This is all human assisted because we then need a human to watch the review the negative keywords. So, imagine stop words especially in known English languages are very difficult to predict. For some reason in French, the script always messes up. So, the engram sometimes could pick up stop words.

By stop words I mean, and, of, it, etc. The engram could pick up stop word as a suggested negative. And then imagine, you could exclude the keyword, for, from your entire campaign. And maybe that's a bad idea. So, definitely we need a human in order to review that. But, as I said, this is at the campaign bidding moment, sorry, the campaign building element. So, it's a phase where the human is heavily involved. And we will have a human review that.

JD Prater:                     Yeah, definitely. I like the idea of the human assisted type of automation that goes into it. But again, you've scaled their work. It's not them manually going out and doing it. It's you going out and bringing the work to them to then review and say, "Yes, no, okay, ship," right?

Gianluca Binell:            Yeah.

JD Prater:                     And that's way faster. It's way easier for me to go, "Yes, no," right?

Gianluca Binell:            Caligula, there you go.

JD Prater:                     If my decision is that binary, you made my life a lot easier. So, I think these are fantastic tools that you have-

Gianluca Binell:            Thank you.

JD Prater:                     ... put together. I encourage you guys to go check out Gianluca. He's at And where can people find you on Twitter?

Gianluca Binell:            On Twitter, my handle is K-T-Z style. And I was very young when I picked out that handle, so sorry guys. I know it doesn't look professional now.

JD Prater:                     And I'll make sure to again, link to that so that you guys can reach out at any time. We'll be linking to these docks in the show notes. And he's already said, "Hey, reach out to me," if you guys have any questions that you're across because he can help anything that may come up as far as implementing.

Gianluca Binell:            Sure.

JD Prater:                     So, really good stuff.

Gianluca Binell:            If it's hard Python stuff, JD.

JD Prater:                     Yeah, if it's hard Python stuff, I'm just gonna laugh at you. So, all right. Well again, thank you so much for flying all the way in from Italy to be here.

Gianluca Binell:            Thank you, man. Thank you.

JD Prater:                     All right. Well, thanks everyone for tuning in. We'll be back next week for another episode of The PPC Show.

AdStage Team