<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=1324445537888680&amp;ev=PageView&amp;noscript=1">

Podcast: What is an ID Graph?

NetWise Jul 21, 2021 6:54:10 PM

Listen:

Click to Subscribe on Your Favorite App!

 

Show Notes:

This week we're pretty much back in the weeds. I guess that's how we roll now. This week we're talking about graph databases, and specifically what we mean when we say "ID Graph" in the data-driven marketing context.

We start with a broader look at the difference between a "graph" and a regular old SQL database, and why it's a pretty big deal, and how it overlaps with marketing.

Then we specifically get into what we mean when we say "ID Graph" and why you're hearing about it more and more in the digital marketing context.

We wrap up by chasing down what deploying an ID Graph in your marketing stack would mean for a few specific channels including website visitor resolution tools, and connected TV.

 

graph-database-neo4j

Links:

https://en.wikipedia.org/wiki/Graph_database

https://en.wikipedia.org/wiki/Routing

 

More Netwise:

YouTube | Twitter | Facebook | Linkedin | Web I Blog+Newsletter

Transcript:

Adam Kerpelman:

Yeah. So I informed you that there was a new Zelda game. And then almost immediately found out somewhere else that it's not new, per se. It's an HD reissue of the old Wii game, but so that you can play it on Switch now and it looks good. That's a brilliant strategy. Actually make games so rich and enjoyable, that people who miss them on one platform during a period of their life can be like, "Oh, I mean I'll play that. It's a Zelda game. I'm sure it's good."

Brian Jones:

Yeah. That's a fun shift that happened maybe, what? Five to 10 years ago. Because if you go back and play an original Nintendo game, like major nostalgia, right? My heart is happy but the games are really lame. I wonder this with music sometimes. I'm kind of like, do we need any more new artists? We've got plenty of music to listen to. You ever have that thought? No offense to the musicians. I'm a musician. I love music but I don't know why [crosstalk 00:01:05] more love songs.

Adam Kerpelman:

That might just be we're getting old. Or, really but though I have had a running theory about this for a long time. That it's a new age of recorded media so there is an interesting paradigm now where it's like, "We have the recordings from the '80s, in perfect digital fidelity."

Brian Jones:

True. Yeah.

Adam Kerpelman:

So it's not like we're listening to Mozart but like there's no recording, and so it has to be played. And so you have quote definitive versions of things.

Brian Jones:

Yeah. It's the same thing that we were just talking about the video games. It's the level of quality is a different metric in video games versus recorded music. Right? Different podcast, man.

Adam Kerpelman:

Completely. Longest cold open ever. Okay. Let's do this thing. Hey it's the Data Driven Marketer, sponsored by NetWise. I'm Adam.

Brian Jones:

I'm Brian.

Adam Kerpelman:

Welcome back, for another hang in the data basement. What are we talking about this week, man? Thanks for joining us. 

Brian Jones:

I love that. I'm so glad-

Adam Kerpelman:

Thanks everybody for subscribing. Smash that like button, or whatever, wherever. This goes out on so many platforms. You know what I mean? 

Brian Jones:

I'm so glad we use database, man. It makes me happy every time you say it. 

Adam Kerpelman:

What are we talking about?

Brian Jones:

We're talking about ID graphs today. And the concept of graphed data. Data stored in the format of a graph. 

Adam Kerpelman:

Right. So marketing people are probably closer to understanding the idea of a graph. Or let's to be fair, having bumped into the terminology at the very least. I think your average person though, has probably heard of graph more so in the context that like, we're not talking about like what you make in the statistics class when you're learning how to chart data and basic math classes. You've maybe heard Facebook talk about their social graph. Sometimes they brag about that but that's probably unlike the more technical presentations that I go to. But yeah, so we say ID graph a lot. It comes up in the context of marketing, but more broadly just what is a graph, in that context?

Brian Jones:

A good visual is to picture your own network of friends, and who's friends with who. So, I don't know. You will have, say you have 10 friends, right? You're a little circle, and you've got 10 friends around your inner circle and draw the lines out to them. It looks like a little web. Each of them have a bunch of friends, right? And maybe you have some in common and maybe some you don't. And so all of a sudden, as soon as you get to that second little web on your first friend, you've now got all these like sub webs. And it's all these little lines drawn between connections. So a graph is like, it's network information, right? That's the technical term for computers networks to each other, but it comes up all over the place in our world, right? It's how the universe works. So it's like social, how you're connected with your family history. We often think of lineage, which is like a downward facing move through time.

But really, it's network too. Right? Your cousin spawn off into people and they spawn off into people. It happens with business supply chains. Our graph in nature. Right? Like who's doing business with who? And who supplies stuff for their products and services? And who buys stuff? It's all over the place.

Adam Kerpelman:

And it's a fairly new... It's not a new way of modeling data, it's just because everything... We think about a normal database or graph, it's kind of like, "Okay. How does this on the x-axis relate to this on the y-axis?" The graph in the context that we're talking about is, how does every point on this chart also relate to every other possible point, if they do relate that way? When you actually map out a graph, like a social graph or a network that looks like what we're talking about. It looks more like a galaxy. You have that thing where there's some things over here and all these wispy lines running between them. It looks like a bunch of stars being closer to one another. Or how iron filings behave when you put a magnet in them. It's an interesting... Look, it's a completely different thing than, "Hey, this is a chart. With an x and y-axis." That said, that thing is also very necessary for how computers and the internet and data and stuff are processed. 

I just realized as you were saying that, as it relates to marketing understanding this distinction here. Like a graph versus a standard database, is maybe the key to understanding the difference between multichannel marketing and omnichannel marketing. You hear them and you hear them-   

Brian Jones:

[crosstalk 00:06:20] interesting.

Adam Kerpelman:

... used interchangeably, but they're not the same thing. One is saying, "Okay. We have one account here, and we're going to target all of these channels at once and see what happens." Omnichannel also takes into account the interconnectedness of those channels. Because the reality is, none of the channels in the digital marketing ecosystem actually exist in a vacuum. They all talk to one another. People share the videos on Twitter, and the Twitter people chat about the videos, and then TikToks get shared on Instagram, and it's all interconnected. And so when you think about that way, you're growing your data stack exponentially with every connection you make. And so that's kind of core to the whole graph idea, right? Every new node in that network has a value proposition that just isn't represented if you're just sort of like, "Here's a list of [crosstalk 00:07:14].

Brian Jones:

Totally. This is a funny one to do without having visuals for people. So there's going to be a lot of-

Adam Kerpelman:

[crosstalk 00:07:20] while me for sure.

Brian Jones:

... us trying to describe visual things over [crosstalk 00:07:23].

Adam Kerpelman:

I mean, we'll put some pictures in the show notes.

Brian Jones:

Yeah. You're totally right. This is where I don't, I've never heard anyone differentiate multichannel and omnichannel in that way. I like thinking, I like kind of where you're going.

Adam Kerpelman:

I can't take too much credit for it. I bumped into that from someone else's post recently.

Brian Jones:

When you first said it I was thinking the difference between just marketing in one channel versus marketing across multiple channels.  As soon as you do multi, as soon as you're doing many different channels, you start to introduce all that extra complexity, right? Because you start to see the same people multiple places, but they might be different ideas. Right? You might only have a cookie from someone over here, you might have a form fill from someone over here, you might have a page view over here, you might have a login over here, you might have a phone number over here. So you very quickly, not only have the channel complexity, the channel network that is like a graph, right? You also then have the communication methods or like the IDs of the people of the person, that is a graph. There's a lot of layering, that happens really quickly and it gets very complicated.

Adam Kerpelman:

Right. Right. So, I mean and that's why you end up with the ID graph, right? Which is just like, there's a cat and mouse game that you have to talk about when you're talking about cookies and stuff which is kind of, people are doing stuff on the internet and then we're trying to figure out how to reach them while they're doing those things. And if the things that they're doing are interconnected, of course the data stream coming off of them is also unbelievably complex because it's also interconnected. But I think for a long time, we have been trying to wedge that data into ways of thinking about it that isn't really that. And so it's not to say that TV is linear, so you don't need graphs for it. We just left a bunch of data on the table. We tried to connect it because it was like, "Well, let's figure out what household is watching what shows and what other shows they're watching." And now we know that people who watch this on CBS, are also watching this on ABC. 

We've always been doing this networking association thing. We just didn't have the computation and the sort of architecture to talk about it in this way. Until computers and digital networks, and now it's so voluminous that when you look at one of these charts you're like, "I can't, what? What is this? It's just looks like a galaxy. This means nothing to me except interesting, nodes and strangely biological looking things."

Brian Jones:

Yeah. And it's important to recognize too that, although the concept of graphs and the technical structure of it and the mathematics and stuff behind them, that's kind of always existed, right? That's core to the universe and core to how we look at information. But to actually be able to process information, in that sense organizes a network on a computer. That's relatively new, because that was one of the big technical hurdles to building social networks. To be able to process who your friends are, and how to send messages out when you communicate. This stuff doesn't get talked about anymore. But like 10 years ago, whenever Twitter was founded. Being able to scale to keep up with their adoption, so that you could like see your friends tweets. When you send a tweet out it might go to a million people, it might need to go in their feed. There's this crazy computation to understand what are the networks look like? Where does this information have to go? Who do we ping with these messages? Who do we ping with replies?

And solving that was a major technological achievement of different networks that were running this information. And so the concept is common now, right? A lot of companies are using information in the form of graph because it's a really handy way to access information, when you have all these nodes, right? All these people and all these systems that have arbitrary number of connections to all the other systems.

Adam Kerpelman:

So the piece I want to hit, if only... Because it's like, I don't know. Nerdy shit. Talk to me about the difference between relational databases and graph, like architectures. This is really the thing we're talking about, right? It's like Excels way of thinking about the world, versus the modern database way of thinking about the world. Facebook doesn't live in Excel. I mean, it kind of does. It's a giant database with things you could look up for each person. But then there's this graph thing. Honestly, understanding this is fundamental to even figuring out how to break these companies up right, if we want to do that-

Brian Jones:

Totally. Totally. 

Adam Kerpelman:

... with anti-trust. The graph is just completely different thing from what people envision if you say database, which I think is mostly like an Excel.

Brian Jones:

Yeah. Excel is a good place to go when people ask me like, what is a database? And database means a lot of different things these days, right? Because we have more complexity to how these things work. But if you think about rows and columns in Excel, that's a good approximation for how a traditional database stores information. You've probably heard of like MySQL or Postgres. MySQL is really common. It's an open source database. It's used all over the web. And that's designed to just look up information, right? The same way you use Excel. If you say you have all your customers in Excel, and you just scroll down the screen until you find the customer you're looking for. Maybe you sorted by the customers last name or sorted by their company name. That's kind of how a traditional database like MySQL database works, and what it's for. It's to look up information quickly that you have in lists. A graph database, is specifically designed to navigate the graph to do queries like, how close is Adam Kerpelman to Kevin Bacon? Right?

And to be able to quickly ascertain how many people you need to connect to. Is Kevin Bacon really, seven degrees away from everyone on Earth? That would be a very, very hard thing to do, to look up in a regular database. But a graph database is specifically designed to store those connections and make exploration of the graph really easy. How will that?

Adam Kerpelman:

So, you know it was great. I was just trying to think, "Okay, so..." Because I remember the non-relational database part, which I guess was a phase that gets us to the graph software that you're talking about. Because there's two different things at play here. And this may be a rabbit hole too deep, I'm looking at the clock. I think we're good. There's storing all that information effectively. And then there's navigating all of that information effectively. And it's kind of like, it's easy to imagine navigating an Excel spreadsheet. Because it's new coordinates, right? Column and row. And you can find anything that you want. Once you have a graph you're like, this is what you need the network traffic All the signal routing stuff, we had to figure out just to build the internet. Then gets applied to these databases, right?

Because it's also, it's like how can we find the fastest answer to the question that you're looking for? Inside of this database it's just completely organized differently than rows and columns. It's really more like a blob if you want to... It's like a cloud shape if you want to think about how the stuff all actually arranges itself.

Brian Jones:

I've got a real world example for marketers. If you're in LinkedIn, and you are looking at a contact and you see the little piece underneath that says like, Are they a first order connection, a second order connection, a third order connection, or more. And then they'll also tell you like, you have 27 double degree connections or two person away connections. That's a graph system. It may not technically behind the scenes be a graph database, but it actually probably is. Because  that's really hard to compute unless you put it in the format of a graph, so you can see your network. So that's literally like looking up network stats about your social network on LinkedIn. And then displaying it in a really functional way for you, right? It's helpful to know, when I'm looking at someone I might want to try to reach out to, that either I know someone immediately or I have three friends in common. 

Or maybe 50 friends in common, you're like, "Oh, I definitely should know this person. I totally should be doing business with them or be networking with them or whatever." It gives you really interesting context. So, having that information linked to that way and being able to see it that way, is really powerful. 

Adam Kerpelman:

And the reason we lump it together and just say, "Okay. You got the graph." Which is all of that we just talked about, right? Not just the database but also the relevant information for how to use the database. It's important to know... The reason we say, okay, here's this... We're going to call it graph database. Or whatever technical term. This is why it's weird to talk about... If I can just take my data with me out of Facebook, like the government kind of wants to do. I'm not sure that solves the problem. But understanding this graph piece is like this fundamental thing that's super. I mean, I say that just because it's like... It's not just the level you and I talk about and write about. But also, it's a critical break in the understanding of how the world works and how marketing works in that world. Because it just, it's a paradigm shift and the ability-

Brian Jones:

Totally.

Adam Kerpelman:

... to handle this information.

Brian Jones:

It's a really good distinction. And I would say, this probably is not always true but I would say you're rarely using a graph database by itself to do anything. You probably also have the Excel type database. Where you're looking up information, and that would be... And that's you just had a really great example, right? When we're looking at data portability on social networks, and something the government cares about. Right? Can we take our information with us? They tend to only be looking at the Excel data. They're looking at my data that's like lists of things. Messages I've sent and pictures I've posted. And they're ignoring the graph. And they're ignoring the graph because that's the hard part and that's the shared part, right? My social network doesn't belong to me, it belongs to all of us. We created it together. 

So the network is stored in a graph shape, and the information about me is stored in kind of an Excel shape. And they're distinctly different. You can do kind of either thing with both, or you can force the graph into an Excel thing. It just doesn't work very well. So, technically behind the scenes doesn't matter but the concepts are very interesting like that. 

Adam Kerpelman:

So then, let's dig a layer deeper. Or to use the one of my [crosstalk 00:17:57] favorite, but of obnoxious corporate speak terms. Let's double click on that. 

Brian Jones:

Oh, man, that's awful.

Adam Kerpelman:

Isn't it? 

Brian Jones:

I've never [crosstalk 00:18:06].

Adam Kerpelman:

Every time I hear it I'm like, "Oh man, that was good." And then it died in like a week. 

Brian Jones:

Yeah. It was too good. It was like a really, really good pop song. 

Adam Kerpelman:

it's too good to be. It's to good to be.

Brian Jones:

What was that pop song that came out? Oh man, with the phone number.

Adam Kerpelman:

Call Me Maybe?

Brian Jones:

Yep. You knew right away. There you go. You're Too Good. It crushed everybody. 

Adam Kerpelman:

Too Good. 

Brian Jones:

It broke everybody.

Adam Kerpelman:

It broke the internet, so to speak. Okay. So to double click it. ID graph. Now what is one layer deeper? What do we mean when we say ID graph? Which is the thing that comes up in NetWise. That's what we do at NetWise. So, what?

Brian Jones:

What do we mean? You're asking me?

Adam Kerpelman:

What do we mean when we say ID graph? Yeah.

Brian Jones:

It's essentially what we do, is connect businesses to the people that work at those businesses and then to the consumer information about those people. And we're B2B marketing data. So the consumer data is only used because of some technical abstraction of the ad tech industry. It's very helpful. We'll get into some of that. So, it's essentially the graph of people that work at companies, and all this information about the people in the companies. Like IP addresses and emails and phone numbers and physical addresses, web domains. It's tying all of these pieces together. So it's actually a little different than a social network exactly. People tend to think of it more as a web of information about these entities. But there is the connected piece too. Right? An arbitrary number of people are associated with the business. And we have historical information too so there's people connected to multiple businesses over time.

You get the network web really quickly of all these pieces of information. And we say, ID graph. We use it in the way everyone else uses it but I also mean it differently. As well as the way that everyone else uses it. So a lot of people use it as an identifying graph. Like I can identify a person using your tool. I actually mean, it's a graph of IDs. Which is different than I think what most people mean when they say it. I mean, both things we actually maintain in our business all of the IDs that are necessary for you as a marketer, to interact with audiences successfully across all of the different advertising channels. So what I mean by that is like, if you only have a company domain, we can get you to that company and then to all the people that work there. If you only have a consumer email, we can tell you where that person works. Because it can come in and link into their business email, link it to the business, understand that.

If you only have an IP address, we can tell you is it a business or is it a person. And we can tell you, so you can come in... And these IDs exist in the ad tech industry, right? 

Adam Kerpelman:

Right.

Brian Jones:

Depending on the platform you're in, you see different signals, cookies, [inaudible 00:20:59], IPs. 

Adam Kerpelman:

Yeah. So when you say IDs, you mean, literally the technical ID that comes off of a cookie or off of LinkedIn or off of Facebook or whatever that represents your-

Brian Jones:

Totally.

Adam Kerpelman:

... account. Your identifier in their system so that we can, [inaudible 00:21:15]-

Brian Jones:

Yeah. And it essentially mean anything that lets you identify a business or a person. Right? Because that's what ad tech and martec, and that's what marketers are trying to do. Right? You're trying to expose your business's messaging to other businesses, and ultimately to the people that work there. Right? So, you're trying to run an ad on Facebook... I'll just give a real specific example. To target people on Facebook, you upload the hash of their personal email. Right? And what that means is you apply this mathematical function to their actual email. Like my personal email, something something something@gmail.com, right? I do something to it. And it turns into this long string of characters and I upload that to Facebook. And now Facebook knows, to show ads to me whenever I'm on their site for whoever paid for that ad space. And so, every ad platform is doing something different, right? We don't think of sending an email as ad tracking, but it is. You just upload an email and send an email, and that's an ID you just used. You use my email as an ID to get an email to me.

Adam Kerpelman:

Right. So the messaging is-

Brian Jones:

It's totally no different than showing it on Facebook.

Adam Kerpelman:

[inaudible 00:22:18].

Brian Jones:

It's the exact same thing. So it's getting all those pieces while you identify it. Same thing sending a piece of mail. It happens to be going through snail mail, and a person is basically delivering a letter but it's the same concept. It's an identifier that identifies a business or a person. 

Adam Kerpelman:

And so that's what we've done is organize that as a graph instead of as a giant spreadsheet, essentially. Which gets to the interesting thing, which is this is the multichannel, omnichannel thing. Why does this shift and the power of an ID graph matter, for data and ad tech? Okay. So that's an ID graph? One step higher? Why does this shift in the way of organizing all this stuff, matter in the data and kind of ad tech context?

Brian Jones:

Yes. So it matters because to target audiences with ad campaigns, right? To get your messaging out to your potential customers, you're reaching out into networks. Right? And a lot of cases you're running your ads on networks, like on social networks and stuff. Right? 

Adam Kerpelman:

Multiple networks.

Brian Jones:

How you do... This is funny, because there's like networks on networks on networks, and there are graphs on graphs on graphs here. You have to be thinking in this way to realize how communication is happening. And, in our sense specifically for our business data, we can help you navigate the business graph to get business information. For example, if you come to us with a list of companies that are in your market, and you want to advertise against them. We can expand those companies out very quickly into people that work there, right? We could give you everyone to do an ABM campaign against everyone at each company. If you want to do like big blast and programmatic. Or we can help you target that down very, very specifically to people with marketing in their title, who have experience with, I don't know HubSpot. Right? And then from that, that kind of helped you navigate the graph from a business to the people that work in the businesses. We can then connect all of these data points that identify people. 

So depending on where you're trying to advertise, Programmatic uses people's addresses, uses people's emails. And then it also has the concept of advertising to households versus advertising to individuals. And so there's, again, you're like more network concepts are keeping layered on here. So, there's just there's really rich, deep component, intrinsic to marketing, modern digital marketing that is just how you're navigating graphs. And I don't just mean data wise, right? Like conceptually how your messaging is going out to people and where it goes and how it gets there. It's all tied up in this concept of graph relationship, network relationships.

Adam Kerpelman:

Well, and I think the interesting thing about building an ID graph and stuff like that. Again, this goes to the multichannel, omnichannel thing is, there's a graph of graphs. Right? We talked about it in the context of Facebook's graph or Twitter's graph. Right. But there's also a graph that connects all of those, like everything zooms out a level and it's just another galaxy shaped the graph of how these things connect and interplay, and you share between the things. So I think, for the next chunk of the podcast, we'll dig into like some specific channels. But the real thing is the why it matters of trying to do this ID graph thing that we work on here, is if the end game is to ride the attention that is traded off between all of those connections, so that we can put the marketing messaging there. You have to understand that graph of graphs, right? Or else you're just stuck inside of the silo of any given channel, and then not doing multichannel.

And you're certainly not doing omnichannel, but like omnichannel is eventually being able to thread together a system where you are aware of, this ad gets pinged on Facebook, it gets on Twitter, it gets shared from Twitter to Facebook. And it doesn't matter that you cross that wall, because you have your graph in place that can freely associate those two off of both of the graphs. And then know that that connection happened and give it some weight, if it's valuable, right? Like an engaged user, as someone who's sharing your content, well knowing where they're sharing it is important. And then yes threading that back into making sure you catch them with the right message and in all the compounds, as you link this stuff together. 

Brian Jones:

You just hit on an important point for us to for our company. The teams that are getting started with us typically use our ID graph more in a lookup sense, right? They're looking up companies, and then people and they're going to run an ad against an audience. Or a more sophisticated customers license our entire database and tie all their marketing data back together after engagement using our ID graph, right? So you might see signal coming to your website, you might see cookies, you might see emails, you might see people's filling out form fees or phone calls coming in your business, but not be able to connect those right? So they'll license our database to put on top of their like CRM, in a custom format. Only larger companies are able to do this. They'd have the resources, but then our ID graph allows them to get that clean signal and say, "Oh, I'm seeing the same person across these 15 different marketing channels." Which there are, that's extraordinarily challenging to do, right? They're very few platforms that do that effectively, if at all really, without kind of doing a custom that's still a very advanced [crosstalk 00:28:14].

Adam Kerpelman:

It's very advanced, it's very new. Yeah. I mean, you still need a data science team to do that. Grab all the data and use it on your platform thing. But to pluck our platform, so that's where you don't need the data scientist and you can do all kinds of cool stuff.

Brian Jones:

Yeah.

Adam Kerpelman:

So channels. So let's talk through some actual use cases on what we would think of as the channels-

Brian Jones:

Totally.

Adam Kerpelman:

... in multichannel and kind of how the ID graph... How something would play out. So, let's everybody knows email. 

Brian Jones:

Well, I was going to say start even simpler, start with the totally analog ones. Phone calls and snail mail? Right? 

Adam Kerpelman:

Yeah, perfect.

Brian Jones:

What are the IDs for those two things? 

Adam Kerpelman:

You phone number and your address.

Brian Jones:

Yeah. So like the classic, the oldest of the IDs, right?

Adam Kerpelman:

Right.

Brian Jones:

And this is an interesting note to make. But people, I don't know if people even bother to think about the US Mail Postal Service. But we have like an extraordinarily sophisticated postal system in the United States, just from the sense of the data. They're not very many countries in the world that you can actually get really good, clean data about all the addresses, Shipping and mailing things, in a lot of the rest of the world is very, very challenging. Nobody has a clean address system all the way down to like every single household in the country. Very unique-

Adam Kerpelman:

The postal service is even amazing because they came up with a way to not have it be entirely tax funded. Which is a hard thing to pull off. That said it also file it under, this is what your taxes pay for. We just take this for granted that somebody can get mailed to us.

Brian Jones:

Right.

Adam Kerpelman:

But there's an interesting way to jump from that to one of the other players here that we talk about a lot, which is publishers, right? One of the fascinating things behind the idea of publishers who are users of data and have rows of data... Think about what's in the Washington Post's database. Your email address, your phone number, and your mailing address. Them knowing your address, that's a piece of first party data that publishers have that's like valuable for retargeting, without even doing this whole graph outreach thing.

Brian Jones:

Totally. 

Adam Kerpelman:

And they are already doing [crosstalk 00:30:37] the less.

Brian Jones:

And conceptually, the idea of you signing up for a magazine or newspaper with your home address, and then getting a physical magazine delivered that then has ads on some of the pages. That's the same thing that's happening with a website, right? Except instead of it being your address, it's like your IP address or it's a cookie that someone dropped on your web page. And then because it's digital and it's real time, we can do really powerful things now in marketing, like display a unique ad to you based on who you are, rather than having to have been printed on a piece of paper and shipped to you with the same ad to everybody. 

Adam Kerpelman:

And so publishers also know that it's you because you logged into the website, right? If you're signed up, and you're really signed up, right? I mean, so you log into the website they know it's you. Not really any different than knowing your address to send you a paper. In fact, you could argue with like less invasive, because all they did was drop a cookie if they didn't get your phone number and your mailing address. But same deal. They are a conduit for papers to get ads to you because that's how they make their money. Ad supported. Not entirely subscription supported, usually.

Brian Jones:

Yep. Yeah. So in terms of our ID graph, right? Offline, we call these things offline. I think we call some... It's funny terminology. It's not always totally accurate. But sending mail, direct mail, or phone calling, right? With a phone bank, or your sales, whatever. Those are kind of analog solutions right? They don't have the same sort of digital signal. And if you want to get that you can get that information out of our ID graph by plugging any other ID in, right? If you put someone's email in, we can tell you the business address that they work out, if you want to send some mail. Or we could give you the phone number of the business to call. So there's a lot of information connected that way. And that's the same across all these channels, right? Put one input in and get the other input out. That's part of what the power is. 

Adam Kerpelman:

Right. Well, and that's where they'd look alike audience idea comes into it. Right? Facebook, you talked about previously, Facebook's look alike audiences. The idea that you put it in an audience and you can grow it, is one of the value propositions of Facebook. And it's also the idea of an ID graph. Except the ID graph is a little more blunt because Facebook's thing is trying to extrapolate your interests, and then grow your look alike profile based on a guess at interests. We are literally just chasing other touch points on the web and saying, "Okay. Based on what we've got, these are the targeting IDs to make sure you reach this person on all of the multichannel channels. right?"

Brian Jones:

Yeah. And that's a good differentiate-

Adam Kerpelman:

So you can do a look alike audience on an ID graph. You can say, "Okay. I've got this one touch point for this person. Give me 10 more." And the ID graph ideally can go, "Here you go." [inaudible 00:33:40]. Sorry. 

Brian Jones:

Yeah. And for us in particular, we tend to do explicit look alikes, right? We're able to say, these people all work at the same company. And this is an important differentiator for us actually as a product line. Facebook does a black box look alike, but their consumer data. I'm sure they would argue something different, but I believe pretty firmly for my experience, their black box is really good for consumer products. If I'm selling t-shirts, right? They'll be all to tell me other people that are into funny t-shirts. They do a much worse job at black boxing, like I sell some complex B2B product and I want other people that look like my current customers. They'll do a very bad job of that. And so will everyone for fairness, right? Google will do a bad job. So what our data allows you to do is be very explicit with what you're trying to do right? Come in, put into our system customers that you've had great success with, expand that out to all the people that work with those companies, right?

So to do upsells for your products and cross sells into different departments. And then also you can look up other companies that are similar, right? Target other companies that are in the same space. And you're getting like a very explicit audience expansion, as opposed to kind of meaningless expansion. Because most social networks just aren't built around business data.

Adam Kerpelman:

So what's an example I guess, of how you would do this? We actually did a video recently, it was a good one for this which is, how would we work with a website visitor resolution tool? There's a bunch of different tools that you can use to sort of raise some information on the people that are visiting your website.

Brian Jones:

Totally. That's a great one. So this is interesting because this is uniquely useful to business to business marketing. Because you're not... I don't know exactly what the regulation is, but you're not supposed to identify like individual web traffic, anonymous web traffic to your website. I don't know who regulates that or what the deal is but-

Adam Kerpelman:

Which means it's hard to do.

Brian Jones:

Mostly, it's hard to do.

Adam Kerpelman:

Individual web traffic.

Brian Jones:

But there are a lot of services for businesses where they'll monitor your web traffic, and they'll tell you what companies are visiting. And they'll tell you, at what scale? Are they seeing a bunch of people visit for one company? And they do this essentially by knowing the IP addresses. Like Microsoft employees are accessing things from the same IP block, the same IP addresses that our networks are connected to. This is not really true as much anymore, because so many people are remote especially this last year. So you'll subscribe to the service and they'll say, "These 25 companies visited your website, here's the company domain, which is the company ID." And then you can bring those domain reports to our business, to our ID graph. Jst like you can with technographic information or intent data. These are all systems that give you signal about what companies are looking at you or engaged with you. Put those domains into our ID graph, and then pump out the people to market to.

So you can then say, "Okay. At these companies, I know the person who buys my product is a software engineer." So then you'll pull out all the software engineers and all the companies that you were just told that visit your website, and then you can go run ads multichannel, right? You can go send them emails, you can go run ads on Facebook, you can go run ads in programmatic. So you'll see them across the web, and it will target a specific population that's really relevant to you at those businesses. So it's another way of how to kind of navigate the graph and get the data out that's then needed to go pumping the systems to reach those audiences.

Adam Kerpelman:

Yeah. And then, the end game I think. You can certainly imagine ways that this stuff could be used. At crazy, massive scales, which is why people license the entire thing for a lot of money. The end goal though, is to not use existing tactics to spam a bunch of people that don't need your message. 

Brian Jones:

Totally.

Adam Kerpelman:

And so if you can based on your web traffic resolve like, "Oh, a lot of people at Microsoft seem interested in our tool right now. We are a tool for marketers. Let's get the list of the 500 people that are actually in the marketing department and we'll start there, with a targeted campaign that's just... We'll run ads and make sure they see them on LinkedIn." And if nothing test out the idea that Microsoft is interested in us." Instead of bugging everyone at Microsoft. Or instead of, just getting a phone book and calling all of the people which, maybe you want to do that too. But that's kind of not how modern marketing. That's the shift here, which is, well now let's ping that more specific audience to see if our software can actually help them.

Brian Jones:

Right. Yeah. I mean, and that speaks to like your marketing budget and your marketing effort and the size of your team and your sales team. Because you don't want to waste time marketing to people that aren't going to buy from you or aren't relevant. Right? Everyone knows that's a classic, right? But what it really speaks to is kind of where's this headed. Right? Which is, ideally, just the classic marketing saying of, reach the right people at the right time in the right place. That actually would be nice, right? And I see that happening in modern marketing, right? Once in a while when I'm shopping for a very technical product where I know all the companies are well financed by VCs and they're running really advanced marketing programs. I'll be shopping for a product line, and I'll all of a sudden start getting emails and seeing ads for a bunch of competing products. And as much as the channel is often annoying, I don't necessarily need work ads while I'm on Facebook. It's at least moving in the right direction, right?

I'll learn about products that actually are relevant to me and will help me at my business and help my business do better because all this targeting is working. So lots of ways to make this work better down the line. But it's working when you do it right. 

Adam Kerpelman:

Yeah. I was going to say so, a lot of ways to clean that up. And I think there's a good example in one that's popping up more and more as this stuff evolves, which is CTV. Which stands for connected TV. But they're just talk... [inaudible 00:39:44] we've got a brand everything as a different thing. When they say CTV, they're just talking about, "We have finally gotten to a substantial shift of people using either TVs that are aware of what's being played on them, or some version of a streaming service that has better data stream and Nielsen ratings or whatever." So now we can use it for targeting, and I can stop watching pharmaceutical ads for people that are 70. Which is-

Brian Jones:

You can watch pharmaceutical ads for people that are almost 40. 

Adam Kerpelman:

Right. Yeah, it's fine. That's I got issues-

Brian Jones:

That's more relevant pharmaceuticals at least.

Adam Kerpelman:

Right? That's fine. It's not the pharmaceutical ads that are the problem. It's the fact that I'm watching this and going this is so clearly not for me. That I don't ever get to have that sensation that we've talked about before of going, "Oh, man. That's actually a thing I'm interested in that you're talking about right now. So I'm going to pay attention." It's really like, I guess I sometimes imagine a childlike state around ads, where think about how you felt about the commercials that would play during the block of TV that we used to watch on a Saturday morning growing up. Because you knew exactly what that slot was for, Saturday morning cartoons. You could really make it so every ad was like, "You know what? That toy is awesome."

Brian Jones:

Zero Gravity cliffhangers. I still want them.

Adam Kerpelman:

I want some bubble gum. Yeah.

Brian Jones:

I want some bubble gum.

Adam Kerpelman:

Crossfire? I'll play Crossfire with my friends. 

Brian Jones:

Right? Right, right. And connected TV is a really interesting one right now too, because TV is still I think the largest ad market in the world although it may not be anymore. It's really close to being eclipse. But the targeting is awful, right? Like you said it's Nielsen's. It's people filling out forms and then you're approximating populations and geographic areas. It's awful.

Adam Kerpelman:

And then guesses based on contextual kill clues. 

Brian Jones:

Yes. It's terrible.

Adam Kerpelman:

Here are the people that we think like the next generation reruns they're playing on BBC America right now.

Brian Jones:

So you get mostly junk ads, right? It's either junk ads or companies who are paying to just be top of mind. Kors wants you to buy their beer over Bud Light. And so they just pound you with marketing that makes you remember Kors when you're at the grocery store. That's stupid marketing. Well, that works great for them but that doesn't work for almost every business selling to other businesses. No one... 

Adam Kerpelman:

Well, also it wouldn't work if you ran it during the next generation. But it works if you run it during a football game.

Brian Jones:

Right. 

Adam Kerpelman:

In different ways, right? 

Brian Jones:

Totally. So connected TV is interesting because they saw, connected TV is kind of new, right? The concept of streaming services. Those companies all saw how ads worked on old TV, not very well. And then they saw how it works on the internet really well but then there's really convoluted complex technical architecture. And so all streaming services make you log in. And it's not just for a payment. It's so they know who you are. Right? When you see the connected TV platforms that are select, who's watching? Yeah. That helps you like make lists of playlist, but that also then lets them know that you're watching versus your wife's watching, and they can sell ads differently. 

Adam Kerpelman:

But it is a very good example of what I would argue is an improved experience.

Brian Jones:

Absolutely.

Adam Kerpelman:

If you're not paying for the ad free subscription. Which is like, if I have to watch ads as the way that I pay for this product, then I would at least prefer that there are ads that are relevant to me.

Brian Jones:

And they're good ad. There exists some good ads too, right? Like the ads on Superbowl commercials, people like them. So do more ads that are... The more you can target ads that are relevant for people, and the people like. Right? We'd solved the problems that people, more people say they hate advertising, right? Nobody wants their show to be interrupted by ads. Right? I don't know how you solve that. Right? You have to kind of hop in where people are going to watch the ads, but I don't mind an ad that's-

Adam Kerpelman:

You just can't. The open web is built on ads.

Brian Jones:

... relevant or [crosstalk 00:43:46] of educational. 

Adam Kerpelman:

Yeah. But yeah, ultimately ID graphs and graph software. More broadly or certainly a core part of this. Hopefully, this sort of exploration of the channels and all that kind of stuff is helpful. 

Brian Jones:

This is a tough one. There's a lot of depth here. I feel like we just kind of... We rumbled through the surface level of it, but yeah I hope we-

Adam Kerpelman:

I think, ultimately we'll get guests and stuff like that to explain certain aspects and things like that. So like, subscribe, stick around. This has been the Data Driven Marketer. Sponsored by NetWise. I'm Adam.

Brian Jones:

I'm Brian.

Adam Kerpelman:

Take it easy, everybody.