In this week’s In-Ear Insights, Katie is out keynoting an event in Europe, so Marketing Over Coffee host John Wall steps in to talk about account-based marketing (also known as target account marketing or key account marketing) and applications of AI to it. John and Chris talk about uses of machine learning for lead scoring, prediction, and why account-based marketing struggles so much with advanced analytics. Tune in to find out whether ABM is right for you, and whether AI should be part of your ABM strategy.
Subscribe To This Show!
If you're not already subscribed to In-Ear Insights, get set up now!
- In-Ear Insights on Apple Podcasts
- In-Ear Insights on Google Podcasts
- In-Ear Insights on all other podcasting software
Advertisement: In the Headlights Newsletter
In-Ear Insights has a companion newsletter, In the Headlights. Every Wednesday, we publish original perspectives, fresh data about marketing analytics topics you care about, and recap some of the most important articles in marketing analytics, data science, social media, SEO, and business leadership. Subscription is free.
Listen to the audio here:
- Need help with your company’s data and analytics? Let us know!
- Join our free Slack group for marketers interested in analytics!
What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for listening to the episode.
In this week’s in ear insights, our CEO Katie is of course on tour of doing keynote in France.
So, this week, we have special guest star John Wall, who you may know from the marketing over coffee podcast.
I also do with him.
So, john, welcome to the show.
Hey, thank you.
You know, I thought you were going to get a week off and Katie was like, Hey, no, john, why don’t you just hop in? I was like, Oh, okay.
No weeks off, no weeks off.
Okay, so today I want to talk about account based marketing and AI applications to it.
But for those who are not familiar, or for those who have had heard just one too many vendor pitches.
JOHN, what is your summary of what is ABM? What is account based marketing?
Yeah, that’s a great question.
Because it you hit it right on the head.
It’s a term that vendors have put together to try and bundle themselves and get some of the traffic and it’s funny for us because it parallels very well to artificial intelligence.
You know, the differences that artificial intelligence has been around from the 50s.
But you have all these vendors saying, Oh, yeah, we’ve got that.
And then Meanwhile, the buyers are saying, Well, what is that? Like what’s there? And so you spend the next 20 minutes like digging into what is this thing.
Whereas the company’s management is at least, you know, they coined that phrase as this stuff came around.
But there’s two big ideas to it, and one, I think is smaller than the other, but they’re both important and critical.
And then the vendors jump on them.
One is in lead scoring, which is the the idea that instead of just looking at individual leads, and saying that like, Okay, this guy hit five web pages, you know, he gets 20 points, we definitely need to have somebody calling, you’re doing that across the company.
And so what you’re hopefully doing is getting, you know, you’re seeing like a team of like, Okay, everybody in this office has hit this web page one time.
So we know that there’s some action going on there.
And that’s great in theory, but then the reality of that as well.
Once companies get to be a certain size, like the fact that you have a hit from the Columbus office and the Paris office like doesn’t really mean There’s a whole lot more action there.
So, but this idea of lead scoring is big.
And of course, just to confuse it further, I would say that that’s actually a company sales, like you’re changing your sales process.
It’s marketing data that drives that.
But that’s a sales thing.
So that’s one, that’s one thing that the vendors will have.
And then the other one is the bigger idea and the better idea of you take an instead of starting your company, and just advertising and putting everything out to the world, you make a strategic change of saying like, okay, these are the hundred companies that are most likely to be our customer, we’re going to build our whole marketing strategy around getting in front of those hundred companies.
And that’s a big huge idea because for a lot of these companies, and I lived through this and bubble one where you could say, okay, we could throw $2 million at our marketing budget, and maybe, you know, we’ll get 10 prospects out of it if it gets in front of the right people.
Whereas with the same money, you could buy an RV and drive around to those hundred companies and you know, have somebody banging on their door with a bag full of $20 bills.
Get in front of them, you know, and it’s literally it’s the same amount of money, like, you know, instead of doing six months in Wired Magazine, you could have the money to travel around and visit the top hundred, you know, personally and shake their hand.
So that’s a huge idea.
And I think, and this way, there’s some AI applications on both sides, but that’s the big thing.
The one is targeting all of your market marketing dollars at the prospect specifically, and then the whole other idea of this whole lead scoring thing and kind of trying to watch the whole team at once.
Yeah, cuz target account marketing is, is an older version of that, right? So it’s key account marketing.
They’re both essentially, what account based marketing became, or tried to take over most of the application of SAS technology for target account marketing and account based marketing.
Is it mostly b2b or is it as a combination of b2b b2c?
Well, this stuff, it totally lends itself to b2b right? I mean, in fact, you’re kind of stupid if you’re not doing this for B to B because it you know, I’ve seen this where you You have some kind of solution, let’s say, in fact, I can give a live example of what we did back at Tibor soft.
We had e commerce solution for food service distributors.
So if you’re selling food to restaurants, instead of having the count guy go around and go to the restaurant, say, Okay, how many bags of chicken breasts Do you need today? You have an Amazon like interface where they just go in login by the chicken breasts, and they show up? Well, you know, yeah, that could be a multi million dollar business, but the odds are, you’re only selling to about 45 people that matter.
So you actually have to do that.
Can it be done at the B2C level? No, I mean, you’re really targeting accounts.
So unless you, you know, the example that would work is if it was still a b2c product, but companies just have to buy it.
So if you’re selling coffee supplies to companies, like that would be a great one.
You know, if you’re the Keurig regional rep or your company uses kirik stuff, well, then yeah, you could use similar things of being like, okay, we know that businesses kind of need to be over a million bucks to Ford the full coffee rig in their cafeteria.
And so we can go that route.
But, you know, it’s not something that Coca Cola is going to directly do to customers,
right? Although I guess you could make the argument that targeting the customer or the customers who are likely to buy or can afford to buy, you know, good old market segmentation is probably still valid.
I know real estate agents do that a lot at the personal level where they know who’s been in their house for how many years and you know, you get the effort, what it’s called, but it’s basically at the seven year point, people are looking to make some kind of change.
On average, obviously, some people stay in their homes much longer than others.
Some people stay in the house much shorter than average car dealers same thing.
But in both those cases, those are fairly large transaction.
So it might be more focused to say, from an ABM and target market, target account marketing perspective, it really applies to complex or high risk transactions as opposed to like, you know, like PayPal doesn’t need a BM everyone’s gonna use PayPal.
I b2b b2c doesn’t matter.
Chewing gum vendors, not going to Need a whole lot of ABM because pretty much any store that sells chewing gum is an eligible customer.
It’s really those higher risk transactions.
So just think,
yeah, that’s a great point is it? It’s definitely low volume, low volume transactions, and even more so if it’s high dollar, right.
So let’s talk about so the techniques in AI that can apply to ABF.
As you mentioned, lead scoring being one of them.
Fundamentally artificial intelligence, machine learning and specific classical machine learning is really good at two things, regression and classification, which is to say like what numbers matter the most, right, that’s regression, and then classification, which is sort things into buckets, and logically for ABM classifications kind of, to a degree not as important because you’re not trying to classify is as good customers as a bad customer.
You know, based on your list of target accounts, these are all good customers, they may you may be able to classify as to likeliness a readiness to buy, but you don’t have to say like is this company worth having on a Because they’re on the list to begin with, right? Yeah.
you know, building the list is a very important thing to there’s definitely different degrees of effectiveness with that, you know, some companies just like look at the two customers they have in the bin and try and clone that.
Whereas doing a little more data analysis against that to see who buys and who doesn’t buy could really change the quality of your your shortlist.
But yeah, once the list is done, then there’s no more work to be done on that front.
And that’s where the regression really helps being able to take a look at all the characteristics of a company number employees annual revenue, things that company sells, who it’s connected to as much data as you can get a hold of, when you run a regression analysis against, you know, essentially saying like, what is the most important factor in all these characteristics that matches to the response variable, which ideally, you would do it against your existing customer base in ad revenue would be their sponsor variable who has made us the most money and then you look at the characteristics and tried to do a regression against that to say like, yeah, these five variables are the ones that likely are going to be the the things that that, you know, maybe if the company happens to have a picture of Andy de Frankel in the lobby, you don’t know this, but there’s a possibility that that’s a dissident extinguishing variable.
And then you build prediction.
So every time you talk to a prospect, you look at their scoring on those variables say, Yep, we need to do more of with this customer.
And this customer is exhibiting you know, weakness in these numbers, maybe they’re not ready to buy it.
The tough part and the part of the reason why this is you know, companies who are using a BM have not already made a billion dollars overnight is that the data is ugly, it’s difficult to clean.
It is difficult to run these analyses.
But the hardest part is actually putting it into into people’s hands to say like, okay, in your sale CRM, you need to call this person in two days at noon, with a picture of a cat.
Most CRM software doesn’t have that
Yeah, no, you totally nailed it right there.
That’s the biggest thing, it’s very easy to come up with ideas for what these tools should do to work.
But the dark secret is that Yeah, you just don’t have the data.
You know, and, and people don’t want you to have the data like in that food service example that I talked about, you know, the top five people get pounded by vendors every day, they will not even give out their phone number.
You know, I mean, you can’t even you have to become an insider for a year or two, just to find out who these people are.
It’s not like there’s just some database where you’re going to run an algorithm against and suddenly be like, Hey, here’s the six people and here’s their home address, I can stop by it’s, you know, they’re, they’re going out of the way to hide it.
One other thing with that, too, we should for SEO, bingo, you do have to stay fit scoring is the term that a lot of the vendors use for this stuff.
why do they call it that?
Well, it’s just, you know, trying to come up with a fit.
It’s the stuff that most the characteristics that match the model of what you’re looking for it, it’s totally but it’s just like you said that, you know, Most of them have about 20 variables and 80% of the time they have four of them, you know, so that’s just not statistical, statistical relevance is your answer to that.
Those terms every time.
It’s just like, well, how relevant is it? And yeah, pretty much in all the b2b situations I see it’s like, well, you’ve got 15 of them.
So good luck with relevance.
Oh, just it’s interesting.
You say that because there is an actual statistical term goodness of fit, which is, is classical statistics, how well, if you had a curve or a pattern, how well does this the data you have fit to that? We were looking at last week at you know, how do you explain gradient descent and gradient boosting in terms of the language people would understand it really boils down to if you had a scatterplot and you drew a line thread at the most likely line, how far away is each dot from that line that is essentially your goodness of fit.
So if you have all these variables, how close to the prediction line whether it’s your number of employees at annual revenue, whatever.
How close is that to the line and that gives you that that goodness of fit.
Now, I from what I’ve seen in CRM software, lead scoring.
lead scoring very often is is completely unscientific.
I people just arbitrarily assign points to things.
And have you seen anyone who’s got a really solid formula for like, yeah, this lead score actually, actually correlate strongly to this question will eventually buy something, or is it really just a bunch of well, we kind of guessed.
And then and these are the numbers were just made up?
Yeah, no, you’re totally on the market, that it’s all back scoring.
It’s like, okay, let’s come up with a model.
And let’s look at the end of the quarter.
How many of those how many of those actually closed? And then the big one is, let’s look at a bunch of the leads that are in the pipe that did close and what did they do and can we build a model around that? And that’s one thing that I’m to kind of to your horn, you have changed that process completely in that so many of that is come up with a high Pass hypothesis and test it.
But you’ve actually taken the approach of using machine learning of just take all the data and run the models and let the model come to you and say, Hey, here are the 15 things that are important.
That’s Nobody does it that way.
But that’s the way where you can actually see if the data is relevant or not.
Otherwise, you’re just making it up.
And one thing that the vendors will talk about with this is fit scoring is different from behavior scoring, behavior.
Scoring is another thing, it’s like, you have your list of, you know, demographic and techno graphic things that okay, they have this tool and they’re in this city, and they have this much money.
And then you have the behavior ones, like okay, they go to this trade show, or they hit this website, 14 or 15 times.
And those are just more traits in their, their quest to hope to hit statistical relevance that seems,
again, somewhat unscientific, like the example you’re talking about, which you typically use something like either a structural equation model or Markov chains to to essentially identify, you know, the propensity of certain characteristics.
That’s that you do that just For two to speed up a hypothesis testing like okay, we can rule out these 14 things because there’s there’s no association whatsoever.
And now let’s build hypotheses about the five remaining variables that actually do matter.
But when you’re talking about something like a fit and behavior unless you’ve got clear sequences of events like this person, you know, downloaded our white paper, then read an email then we talked to the trade show and we bought them a steak dinner, and then and so on and so forth.
You have these clear chain of evidence really, if you have that.
Anything you make up on those behavioral scores is going to be wild guess at best because you’re not looking at propensity scoring from action to action.
If you’re unfamiliar with propensity scoring, think about it.
We like to use examples from Sports because it’s the best, best way to convey this.
If you’re watching a basketball team play basketball.
The person who scores the basket is important.
Yes, but the person who passes who does the special This is equally important because if the ball doesn’t get there, it can’t be scored.
The same is true for all these ideas about, you know, b2b marketing, the person who closes the deal or the channel or the technique, the closes the deal may not be the most important in that interaction, that $200 steak dinner that you bought, that person may actually matter a lot to some people.
But unless you have the data and models to build that basketball passing scenario, you don’t know whether one stage is as is correct.
So when you see these implementations and you know, in marketing, automation, CRM software, does anybody ever go back and check like, yeah, our behavior scoring, our fit scoring and our lead scoring are all completely non correlated to actual outcome?
Yeah, you know, in theory they do and it gets much worse actually.
So not only like, are they kind of living that method like praying that that’s all working.
A lot of the vendors have gone a step further to and saying that a big part of this is orchestra.
of, you know, your system can do that will move these parts around and change the orders.
And you’ll find the optimum order of what to do that you know how to do that stuff.
And when those programs dropping, yeah, all of this stuff is, you know, we kind of these are the ideas and we hope and pray that they work.
And you know, unfortunately, we haven’t kind of seen anybody rise to the top as a vendor to where this differential differentiation is so great that it starts to dislodge Salesforce or HubSpot, or some of these other tools.
And of course, they’ll spot Salesforce and HubSpot do get better with their lead scoring and can eat away with that.
So yeah, nobody is emerged as King of the Hill and proven This is all work.
this all works.
And yeah, I’ve never, you know, I’ve seen case studies of kind of like, Okay, I’m this specific thing.
This sort of works and we prove this model, but there’s definitely nothing where you could do that in one place, and then just go plug and place that in some other company, right.
And, again, if somebody had figured that out, they’d be you know, a trillion dollar company by now fairly quickly.
So I guess to summarize that Then account based marketing seems to still be stuck at the data gathering stage.
Right? It’s not the data is not there.
So all the analysis insights and, and model building and applications of AI.
Would you say it’s fair that vendors who are promising the AI is going to solve everyone’s problems in this case, probably have not got the goods behind the scenes.
With you, they’re definitely stretching the truth.
But they do have their banner flying under this idea of, for marketers that are still living in the branding world that, you know, we need to advertise and build our brand.
No, you need to find 100 people that are likely to buy and go chase those people that that can change your financials and make you successful.
But then yeah, the rest of the data.
Yeah, you hit it on the head.
It’s all about the quality of the data, and most people don’t have the data,
So, takeaways, if you don’t have good quality data, account based marketing is not going to do anything better for you than your existing marketing is just going to cost you more money for more vendors and expect that you’re still going to have to Do a lot of the work yourself because behind the scenes, the tools the algorithms are only as good as the data put it.
And it’s it’s been true and software for 70 years it is true today and AI garbage in, garbage out.
As always, if you have follow up questions for the show, please leave it on the blog post associate with this over at Trust insights.ai.
While you’re there, please subscribe to our newsletter in the headlights which goes out every Wednesday, and we look forward to talking to you soon.
Need help with your marketing data and analytics?
You might also enjoy:
Get unique data, analysis, and perspectives on analytics, insights, machine learning, marketing, and AI in the weekly Trust Insights newsletter, Data in the Headlights. Subscribe now for free; new issues every Wednesday!
Want to learn more about data, analytics, and insights? Subscribe to In-Ear Insights, the Trust Insights podcast, with new 10-minute or less episodes every week.
Also published on Medium.