The Unconventional Approach to Programmatic Advertising With AI  

Olin Hyde

Listen now on these popular platforms

Share this podcast

Highlights from this episode

Podcast Transcript

Artificial Intelligence as Applied to Programmatic Advertising: A giant leap forward in B2B targeting has occurred, and today we're going to talk about the experiments that led to this breakthrough. LeadCrunch[ai] uses artificial intelligence to drastically improve the performance of B2B demand generation campaigns through account-based "lookalike" modeling. Click the link for more information. https://leadcrunch.com/solutions/

Posted by LeadCrunch on Thursday, April 18, 2019

Hosts: J. David Green and Jonathan Greene

Topic: Programmatic Advertising

Subtopic: Artificial Intelligence

Duration: 20 minutes 

In this episode of the Green & Greene Show, the LeadCrunch B2B podcast, seasoned marketing experts discuss programmatic advertising – and going from ho-hum results to a triple digit lift. 

TL;DR

  • Dave Comes Clean With the Audience
  • AI in Programmatic Advertising & Rad Results
  • How Gartner Helped Us Key in on our Real Target Audience
  • Jonathan Runs an Advertising Experiment
  • … and Shares the Results of Using a Lookalike Audience
  • Define Best Customer 
  • Real Talk on Artificial Intelligence for B2B 
  • Top of Funnel vs. Middle of Funnel 
  • Why Conversation Leads Are a Truly Valuable Metric 

Podcast Transcript

[INTRODUCTION]

[0:00:05.1] ANNOUNCER: Live from deep in the heart of Galveston, Texas all the way to the gleaming shores of Jacksonville, Florida, it’s the Green & Greene Show. Here are your hosts, Dave Green and Jonathan Greene, ready to unlock the mysteries of scaling demand gen. The Green & Greene show is brought to you by LeadCrunch, which has reimagined how to find B2B customers at scale.

[INTERVIEW]

Dave Comes Clean With the Audience

[0:00:23.8] JG: Dave Green.

[0:00:26.8] DG: You know, we have been dishonest with our audience, Jonathan. I am no longer in Galveston.

[0:00:36.0] JG: I know.

[0:00:37.0] DG: I am now in beautiful San Diego, and all you people in the other parts of the world can eat your hearts out because you don’t have the best weather ever.

[0:00:45.7] JG: It is. It’s like I get off the plane in San Diego and my body automatically feels better. No humidity, it’s ridiculous.

[0:00:55.2] DG: I have actually started to age backwards now. It’s incredible.

[0:01:01.6] JG: We’re going to have to have new intro music made. I kind of want to get a hypeman, too like a hip hop: “The Green & Greene show, what? Yo!” Something like that. Have a little something for the kids, I guess.

[0:01:15.7] DG: Yeah.

AI in Programmatic Advertising & Rad Results

[0:01:17.0] JG: Anyway, we’re going to talk about programmatic advertising today, and specifically, we’re going to talk about artificial intelligence in programmatic advertising. We have done some experiments and the results are pretty rad, so we might want to just go ahead and throw that out there. What do you think?

[0:01:35.8] DG: Absolutely. Just to clarify, you know, AI has been used in programmatic in order to do bidding. This is using AI to do targeting, which I think is way more important. 

[0:01:50.6] JG: I think it’s way cooler anyway. Everybody has this problem. Unless you’re using a very specific ABM list, you sit down at your computer terminal to do top-of-funnel programmatic display advertising for awareness campaigns the top of funnel and the targeting options that are available to you are somewhat uninspiring. 

With your general stuff that most people use, like the Bombora segmentations and stuff like that, you have general programmatic data. This is, as far as I know, the first real true artificial intelligence application of targeting for top-of-funnel programmatic, which is how it’s conceptually meant to work.

Instead of doing all that segmentation and stuff, you would just bring a list of your best customers and we would feed that list into our artificial intelligence machine, and it would go out and find the data, algorithmic commonalities between those companies that you had success with already.

Then it applies that to a large data set, in this case, the universe of B2B businesses and in North America, not necessarily B2B which is businesses in North America. It compares that data set of businesses in North America to the algorithmic pattern that you developed from your best customer list, and it returns a list of people to whom you should market based on those you’ve already had success with. That’s pretty freaking cool if you think about it.

How Gartner Helped Us Key in on our Real Target Audience

 [0:03:34.5] DG: Not to get too technical because, you know, part of my audience is in California, where they might not really get it but… my point is that, to build lookalike model is when you put your best customers in there. They’re doing kind of what a sales person would do. They’re going out and looking at sites and looking for keywords that are unusual which seem to be indicative and part of this profile of customers.

What’s close? We do this for ourselves and did it on this experience and found that, if the customer had the word “Gartner”, for example, the technology giant, that tells you what kind of CRM to get and stuff like that, it was a good indicator. Why? They’re licensing content probably from Gartner and they tend to be good prospects for us.

I don’t think that even our best sales guys would have necessarily come to that kind of a nuanced conclusion, and every company is going to be different. That probably has nothing to do with what’s important to you, but there probably is a key to all of those sides from which a really good sales person could say, “Yeah, these guys look like they’re a good prospect.” That’s sort of the essence of that modeling.

We have other kinds of models, too, but I just thought, in this case, that was the type of model that was used. I thought it was so different than, “Give me all the people that are this size in these industries.” You know, that’s like trying to cut bread with a river rock. It’s just not very precise. 

On top of that, this was an internal experiment, and Jonathan, if you didn’t know this, it’s like he’s not just a black belt, but there’s some darker color after that in digital. I was thinking, “Well, this is going to buy us a test.” It’s got to be done for the average Joe. Jonathan was way ahead. He actually outsourced it to an agency that did all the creative and didn’t put all the Jonathan magic to it and just to make it like what everybody else would get. 

Jonathan, do you want to walk people through the experiment and what the results were?

[0:06:40.0] JG: You make a fair point not to espouse the idea that I’m like some sort of magenta belt or something, I don’t know.

[0:06:48.5] DG: You are. 

Jonathan Runs an Advertising Experiment

[0:06:49.9] JG: I don’t know what the next color would be, but anyway, the point is that I’m good at digital, so if I just ran my creative against somebody else’s creative, I would win nine times out of 10. I’ve been doing it for so long, so I didn’t do that. I set the test up and I held static the creative. I had the creative display banner ads generated by an agency so they were identical. Roughly identical number of impressions, that’s really hard to hold static, but we held static budget. We held static bids, everything that we needed to do.

This is a head-to-head test between two targeting segments which is what we wanted to get at. I took the LeadCrunch-generated artificial intelligence look-a-like audience segment for our business, for LeadCrunch. I took our best customer list and submitted that to the AI. The AI spit out a list of lookalikes to go after and then I used LiveRamp to append cookie and device ID to that and hung it in TradeDesk. Then we ran a head-to-head against a typical top-of-funnel targeting segment, like the Bombora segment that we use, which is essentially a segment of advertising and marketing executives in North America.

This is what a lot of people do with their top-of-funnel demand generation. This is a proxy for what most marketers would do right. I ran that for 30 days and came up with the results. 

… and Shares the Results of Using a Lookalike Audience 

We ended up getting a 285% lift over the control with that particular experiment, which is tremendous. I mean, that’s three times the click-through rate in terms of the click-through rate. That’s everything to do with targeting. The creative was identical between both targeting segments. That initial experiment was very successful. We’re very excited about potential here. That’s what we did.

Define Best Customer 

[0:08:59.6] DG: Yup. You know, just so people know, best customers is kind of an amorphous term, and you can actually be very precise with your definition of a best customer. These could be best customers for a product in your product line. These could be big spenders. They could be from a segment of the market that you want to go after.

However you want to define best customers, we can use the same approach and find people to look like that side of best customers. I think the possibilities with this are really spectacular, and I’m pretty excited. I think you’re just getting started. There’s a whole series of tests you’re going to be doing around this same framework over the next several months.

[0:09:47.5] JG: Yeah, absolutely. You can categorize marketers into buckets, in a sense. At the low end of the spectrum, you have people who are sort of upstarts and would grab a general targeting Bombora-style segment and run that. They don’t really know what else to do, so that’s the initial test. If that’s you, if you’re just getting started in programmatic, come to us. Our targeting is going to revolutionize what you’re doing. 

More sophisticated people who have the skills to go out and curate an ABM audience, for instance, are using traditional programmatic-plus layered data, like maybe DiscoverOrg or something. I want to test against those audiences next and see what the list looks like for our artificial intelligence against an ABM list. 

I suspect we will still get a lift. I don’t suspect it will be 300%, but it may be 150% or something in that range in terms of improved click-through rate, even over ABM list. If that happens, obviously, we’ll be very excited to go to market with that.

[0:10:57.4] DG: Incredible story, Jonathan, incredible success. Is there anything else you think the audience might want to know? I have one thing, but I’ll defer to you.

Real Talk on Artificial Intelligence for B2B 

[0:11:11.1] JG: Listen, not all artificial intelligence is created equal. A lot of people are throwing the word around, you see it on just about every B2B services website now. If somebody is claiming that they have artificial intelligence, you need to research what they mean. Predictive analytics and what we’re doing are not the same thing.

We have devised a completely different way of analyzing businesses that includes but does not hinge upon programmatic data. This is like an entirely new way of looking at things, and it’s not been done. Our artificial intelligence really is intelligent in a way that most people cannot approach.

I think it’s an important point of differentiation.

[0:11:59.0] DG: Yeah. By the way, I think the word “artificial” is the operative word in some of the artificial intelligence claims and I agree. I think you have to really peel the onion on that and find out exactly what they mean and how they’re doing it. 

You know, they may have some wonderful application of artificial intelligence, but I think it has become such a buzzword, it’s being grafted onto everything, whether it applies or not. Even small services business that don’t have the data scientists or a programmer on staff are going to be using artificial intelligence now, I just saw one yesterday.

The thing that you’ve done, for anybody who is trying to do programmatic advertising, I thought was really interesting, Jonathan. I wonder if you could just take two minutes, but it’s the idea that, often, you’re not driving somebody to a landing page with a lead capture form. 

That’s not the objective, but you’re trying to educate them a little and cooking them and following them with other ads that ultimately get them to convert. Can you talk a little bit about that philosophy? I thought it was really fascinating the way you were doing that, and I suspect other people might find some value in learning about the thinking behind it.

Top of Funnel vs. Middle of Funnel 

[0:13:20.3] JG: Yeah. From a strategy perspective, people spend an awful lot of money at the extreme top of funnel, trying to get a lead capture, which if you think about it, is really more of a top-middle of the funnel position activity. I didn’t think of lead capture as more of a middle-of-the-funnel activity.

Top of funnel, all I want to do from my strategic perspective is to begin to familiarize people with my brand and how my product operates, what the value proposition of the brand is and what our specific operates are. I think that, especially when you’re launching a new or a revolutionary product, people are not quite ready. The value is not there yet for them to exchange their data for whatever it is that you have behind your lead gate. 

The only way to move them to the point where there is going to be an appropriate value exchange is to begin by having a conversation, by curating information that’s interesting and moving.  There’s no burden to it, so people can engage when they learn about your brand and then, look, when people click through my programmatic ads, they land on a story. It’s a page called “A Tale of Three Marketers”. It’s illustrated, it looks like a cartoon, basically, but it’s about this marketer named Jen who is having a very specific problem that all marketers can relate to. You hit this landing page and you get engrossed in the story and you find yourself clicking through.

The hero of the story ends up being the artificial intelligence that we have available and what it could do for Jen. Her problems are like, “I can’t prove my ROI,” or, “My targeting sucks,” or just, “My sales team hates my leads.” 

[0:15:07.0] DG: Hey, Jonathan, there is no one in the audience who has experienced those kinds of problems.

[0:15:12.4] JG: Yeah, right. I bet you every one of them has at least one of those problems right now. I know I do, and everybody does as well. 

Anyway, the point is I’m familiarizing. I’m telling a story. I’m engaging people rather than converting them. When you get to the point of conversion, I don’t know what everybody here pays for brand leads, somebody who has decision-making authority and who is well-indoctrinated on the brand and who is ready to buy, but by the time I’m done with my nurture process right now, it’s costing me $72.

I’m looking at my Facebook, $72 to capture brand leads and give them to the sales team. That’s fantastic in the B2B space. Most people are probably paying a multiple of that, I imagine. That’s because of this top-of-funnel-like nurturing engaging strategy.

[0:16:01.0] DG: You do have the people who are fooled by volume and by very inexpensive leads because that’s what they can measure, the cost per lead, and that’s how they think they’re extracting value from the spend. 

If and when they’re able to close the loop, not in every case but in most cases, sales quickly learns that those aren’t worth following up with and there’s no ROI whatsoever. It’s actually even more negative because you wasted sales capacity which is very expensive on something that doesn’t convert.

[0:16:36.8] JG: Yeah, I should clarify those $72 leads are people who are actually saying, “Let me talk to sales.”

Why Conversation Leads Are a Truly Valuable Metric 

[0:16:44.0] DG: Yeah, we have a nomenclature here we use. Nomenclature is a big word for a phrase. We used to distinguish between leads that really just introduce themselves and are willing to give us some information about themselves in exchange for a white paper or whatever, and then we get people who we believe are ready to talk to sales.

That’s a much smaller subset, we call those ones conversation leads because that kind of describes, in our mind, what people do. I know there’s all the Sirius Decisions lingo out there about NQLs and SQLs, but I always liked language, and this comes up with something that’s easier to remember and descriptive of what it does for these stages.

[0:17:33.8] JG: Yup. Best to just keep it simple. I tested this targeting technology at the extreme top of funnel for that sort of awareness campaign. I suspect that, when we get to the middle of funnel with the same technology, we’re going to find that the conversion rates in terms of lead capture are higher as well. 

It makes sense. Think of it this way. You can start with the Atlantic Ocean, trying to catch a specific kind of fish or you can start with a stock pond. We’re starting with the stock pond, so it’s much easier to catch the fish that you want to catch.

[0:18:08.0] DG: Yeah, there’s also another benefit. If sales people have to go find people who will buy it from them, their bias is towards people who will buy, not whether the customer is going to be a good fit. Yeah, there are really professional sales people who will, in fact, steer people in another direction when it’s not a good fit, but a lot of sales people don’t.

Get the free eBook “How to Win the Love of Sales “

The downstream problem with that is that your customer service department or customer success team is getting a lot of grief from the customer because they’re unhappy with the decision they’ve made. It’s hard to really build a robust business if you’re leaking customers out the back end, and so you’re doing something to try to find out, “What can I do to replicate more customers that tend to be happy with us?”

That’s a really under-appreciated part of demand generation. Rather than, as you said, trying to boil the ocean, really getting very focused and very targeted, I’ve always found that, in a lot of aspects of life, focus matters a lot. Whether it’s getting your product portfolio focused, the way that Apple did when Steve Jobs came back and took more than a hundred products and reduced it to three or four, or targeting your audience in a more precise way, focus really does help with efficiency.

[0:19:45.6] JG: Absolutely. That’s it, you know. That’s what we wanted to talk about today, but I wonder if anybody has any questions. There are a few people on, you go and drop them. We’ll give it a minute or two. I guess while we’re doing that, Dave, parting thoughts? Anything else?

[0:20:02.2] DG: Here’s a question. They’re asking why we don’t wear hats.

[0:20:07.0] JG: Man, listen, I make this look good.

[0:20:12.4] DG: A polished man.

[0:20:14.2] JG: No. Look, I got the sides tightened up not long ago. Beard looking robust, it’s fluffy but soft. Feel this, it’s like… 

[0:20:23.4] DG: I’ll take your word for it.

[0:20:28.3] JG: It’s fantastic, I don’t need a hat.

[0:20:30.3] DG: All right folks, thanks so much.

[0:20:32.5] JG: All right, no questions, we’re done. It’s been The Green & Greene Show. Get yourself some top-of-funnel programmatic advertising mojo. Give us a call; we’d be happy to help you. 

It’s been real, we’ll talk to you next time.

[0:20:46.5] DG: Bye.

[END]