Eric Westerkamp, CEO at CaliberMind, joins our host, Camela Thompson, Go-To-Market Thought Leader and B2B Insights Expert, in this episode of the Revenue Marketing Report. Eric shares why using advanced logic beats out traditional intuition-based profiling.
Today, we are talking about machine learning and how it can help us with ideal customer profile analysis. What are you seeing in the market?
“The concept of an ideal customer has been around for a long time and a lot of organizations have been using very standard mechanisms to try to figure out who they are. Often, they are pretty simple: industry, size of the company, and maybe some technographic information or data. However, pretty simple things. It is a way for the organization to determine who to focus on, and who is most likely to buy my product or service.
“Then also, to be honest, how do I divide that up amongst sales and marketing teams? Who do I go after? We have seen that as a pretty standard mechanism in the industry. I think what we are finding is that the most advanced marketing organizations are trying to give nuance to that ICP. There are a lot of stats out there that say all the companies you go after might be this big, but only X percent are in the market right now.”
Yeah, it is something like 3%.
“I think I read 5%, but you’re right on that number. They may be in the market right now and so you have a slice that you want to go after and then you have a slice that you need to be aware of and bring them in. Then, at some point in time, you want to move some of those down here. However, others have more intent, maybe projects and moving, how do you use ICP?
“There’s a term we use internally called ACP, ideal customers and acceptable customers. Here are the ones you want there. Then, here are some that could be good customers, but from an expense perspective, I’m willing to spend almost anything to get these customers. They are good customers, but they might not tend to have the same LTV. They may come in at middle to lower price points. You want them, but you’ve to make sure that you are applying the right amount of resources to acquire them. They can’t be the same as your ICP.
“How do you identify the two different buckets? Then, how do you broaden out the features so that when you are looking at a customer base, you can differentiate between the two using things that are perhaps a little bit more nuanced like I said, more than just the size of the company and the industry.
“I am thinking back to some of the early-stage startups that I worked for early in my career. What I saw a lot and still see today, to be completely honest is we had some executive founders who were very technical. They built the product for a reason. They thought they knew exactly who would want to buy and why. Then, we dug into it and that wasn’t quite right. Alternatively, we hear from the sales leader. They’ve got this great big deal and that is the profile they want to chase from then on, not realizing how expensive it was to keep up with that kind of customer early on in their journey. So you are not only talking about moving away from intuition-based ICP analysis.
What Eric is talking about is very advanced already and we even haven’t dug into the machine learning aspect yet. I would love to talk about how we have products that are well-suited to a profile that is fairly easy to identify, but we’ve got some products where I’m thinking analytics. I’m thinking attribution, where it is not so much about the size of the company or the makeup necessarily of the company.
It’s more about the mentality and adoption of data-driven decision-making. How does machine learning help with that sort of nuanced profiling?
“Here at CaliberMind, we are playing around with leveraging machine learning to help us pick our ICP, our ACP. Our Head of Customer Success, Nick Zangre, we tasked him with this and told him, hey, go play around. You have access to all this data in CaliberMind. You have access to some machine learning frameworks that we have embedded on top of it from Google. We added, why don’t you experiment with how we would create a new ICP score? It was very interesting. What we came up with was a new model and it ran on top of all our data and it recreated our ICP.
“The neat thing about some machine learning algorithms, what we did was that it doesn’t have to be just industry and size and some technographic. We embedded those features in it, but we also embedded some other interesting information like how many data analysts are at the company. There is this other set of features and you can broaden out your feature set and then look back at it if you’ve access to it. Successful data. The best ICP score doesn’t just take your intuition, but it looks at over the past three to four years, what has worked. What companies have been engaged with you? You have access to the engagement scores or anything like that.
“That’s great because you can look at here are companies that got engaged and that perhaps worked, that probably worked or that maybe didn’t meet the profile out there for technographic and firmographics and some of those things. You can maybe look even deeper at what were these individuals, what were their roles. Who were the people? You can embed all of those different features and you can start to run ICP and start to tune it and start to come up with a new mechanism that allows you to score accounts.
“The great thing about something like this is when you automate it like this, your ICP might change, but over time, your ICP drifts. Your product is more mature. You start going up the market and that changes. A lot of companies, I have seen them create their own ICP methodology, and they embed it on Salesforce. It is some simple math and boom! Everything is scored that way and it drives everything. You see that all the time, right?”
Yeah. I am cringing since if you’re not looking at your ICP every six to eighteen months at the most in a younger company, you are not doing it often enough.
“Exactly! By using a model, you can rerun it every month, if you want to do it too often, every quarter you might see something. Every six months, definitely! You want to rerun it and then you rescore accounts and then it tells you what your ICP is. Then, of course, there’s still a human element, okay, once you get the data. Let us look at it.
“For instance, we ran our ICP scores and it did a good job. We embedded the machine, then we ran it back by our sales and marketing teams and said, let’s look at the data. And we decided, you know what? It is doing a good job, but here are some industries that it is selecting that we just know aren’t for us. So even though the model is pretty good, it selected some stuff that we probably don’t want to target.
“Then, you go back and add a layer of filtering on top. Okay, let us take out those that are done. Let us give the sales team a way of saying, I am getting access to these customers. This is working. That will feed back into the model later on. It enables the model to adjust over time and lets you operationalize that testing and that work on the model. If you do it right, you can have the model create the scores and then push those scores into Salesforce and rerank all of your accounts on a quarterly or biannual basis.”
I like talking about this particular use case since ICP can help define so many things throughout the go-to-market team like your sales territories. Do you want to dig into that a little bit? I know marketing doesn’t usually design territories, but it impacts them.
“It absolutely impacts them. I’ll be honest, it’s the beginning of 2024, we are doing all our territory planning right now, for our salespeople. We are using the ICP scores. What we are doing is we will say, okay, we have reps, we’ve assignments and we have regions. Okay, now based on the ICP. Well, go back in. What is great about it is I can say by rep, how many accounts do you have that are at ICP Level A? How many accounts do you have that are at ICP Level B for your region? We then start to balance it that way.
“An interesting nuance that we are working on right now is we said, okay but we know that certain regions over or underperform for our technology in different companies. That will be different regions and things or perhaps different nuances. We are looking at that and we said, okay, we know that certain areas of the US tend to perform way better than other areas, even though there may be similar companies in those regions that match that ICP.
“The question is do you take that data and try to get it back into the model and rerank the ICP or do you add a layer on top of your ICP that weights it? That is more of a data science question. However, it’s also how you operationalize this. We decided to reweight the ICPs based on region. If you’re trying in this region, your ICP, the value of companies, that ICP A, in this region, are maybe more valuable than the ICP A down here. That is based on historic trends that we’ve seen right now.
“The reason we don’t want to embed it on the model is some of those trends may be self-reinforcing. The reason those regions are performing is that you may have sales reps that are focused in there or on the ground. We do more marketing for some of those things. We decided for our own to layer it out so that we could perhaps adjust it over time and change it and keep an eye on it and keep the identification of companies that are strong since now, you still have your ICP score. There are companies in these regions that match that score.
“You do want to find out what it takes to get them. That is what we’ve done with ours. We are blending machine learning models to create that original score with some context that we put on top of it that’s specific to our business. Then, we use that. We push that into our system. We do our sales and territory planning that way.”
Hopefully, with a sophisticated model that can pick up on the signals that indicate data adoption and maturity in that way, you can start to pick out where perhaps you aren’t marketing enough or you don’t have enough in-person interactions going on.
“This starts getting into the combination, I think of three key terms that help marketing and sales teams know where to focus on your ICP, and what kind of accounts they should go after. You have intent which essentially says, okay, there are some signals out here coming from intent vendors and stuff like that. Different industries have them. We use things like Bombora.
“Then, you have engagement. How engaged are they with you and your brand? Those three numbers govern tactics. Is it marketing that should be engaged in doing something? And what should they be doing based on where those three numbers are? IT sales, that should be engaged. A heavily engaged customer that is interacting with you and your brand a lot, sales should be picking that up and working on it, but maybe a customer pulls dead on for ICP and had dead on for intent, but isn’t engaging with you at all.
“That is marketing’s job. Go get those engaged with us. Then, when they get to a certain level of engagement, sales starts to pick it up. Then, marketing has a job in that too. That is helping them stay engaged, bringing all the other buyers in that buying group into the process after they start to get engaged, things like that.”
For more content on B2B marketing trends, listen to the full Revenue Marketing Report episode at the top of the article or anywhere you podcast.
