Hypertargeting with Buying Signals: The Next Generation of ABM
B2B hypertargeting is a method to identify prospects who are most similar to a business’ best customers who also have the highest propensity to buy. Buying propensity is the sum of buying signals which are specific traits that indicate a need for a solution based on a company’s business focus and how they operate. Evaluating buying signals is significantly different from traditional methods of targeting prospects because it evaluates the uniqueness of every company and each person working with that company. It replaces the view of a company belonging to an industry with an understanding of what a company does (business focus) and how they do it (operational profile). Unlike traditional marketing methods, hypertargeting can tell you who to target within a large enterprise and why they are interested in a particular solution.
In a side-by-side comparison conducted with six demand generation campaigns, hypertargeting identified 28% more “best fit” prospects than account-based marketing (ABM) lists and 217% more than traditional filter-based lists using ZoomInfo data.
LeadCrunch evaluated data from 7 demand generation campaigns using three different targeting methods in each campaign:
- Filtering ZoomInfo data to target companies based on firmographic data such as industry, headcount, revenue, etc. This was considered the baseline measure as it represents the methods used by approximately 86% of B2B marketers.
- Account-based marketing (ABM) lists generated by our customers to identify ideal prospects sought out by internal sales and marketing teams. ABM is used by approximately 14% of B2B marketers with a heavy concentration among high-growth, high-margin SaaS businesses.
- Lookalike audiences generated from artificial intelligence algorithms that find prospects that are most similar to a list of “best customers” provided by marketers. This is a new method used by less than 1% of the market.
Baseline, ABM, and Lookalikes were then compared to determine which method would find the largest available audience of “best-fit prospects.” A “best-fit prospect” was defined as a lead that engaged with content marketing that was subsequently verified by our customer as “accurate and actionable” to use as marketing qualified leads (MQLs) for further lead nurturing. Passing the MQL test is the first step in qualifying new business opportunities so it is also a directional indicator for conversion into closed-won sales. More MQLs equals more marketing-generated revenue.
The six campaigns reached approximately 90,000 prospects in demand generation campaigns for solutions to business financial planning, cybersecurity, cloud communications, and enterprise marketing. The campaigns generated 1,815 marketing qualified leads from these prospects (2%).
Each MQL went through a 17-step data verification process to verify people’s names, titles, and contact information. This process allowed us to compare the data accuracy of each of the three targeting methods. Customers then verified the accuracy of the data. 99.5% of MQLs were accepted after verification.
Although firmographic filtering (baseline) is used by 84% of B2B marketers, it was by far the worst-performing method to find new prospects. It had the highest data error rate (20%) as it generated the smallest audiences. LeadCrunch identified many cases where the primary data source, ZoomInfo, had grossly misclassified a company’s industry — such as a consulting firm being listed as a software company. The data on headcount and revenues was wrong in more than 35% of the cases; presumably because of changes in the market due to the COVID Recession.
ABM outperformed baseline by 189% and resulted in far fewer errors (<1%).
Hypertargeting with lookalikes performed the best by identifying 217% more MQLs than filtering using ZoomInfo firmographics (such as industry, headcount, and revenue value) and outperformed ABM lists by 28%. Errors were the lowest at (0.3%) due to data freshness by continuously harvesting the internet for data from public sources (e.g., company website, LinkedIn, government sources, etc.).
The conclusion is clear: Hypertargeting makes ABM more effective by increasing the available market by finding ideal prospects that are missed by ABM lists.
This is an important finding because we live in a world of resource-constrained marketing. Now more than ever, marketers are expected to turn marketing dollars into high-margin revenue. This explains why account-based marketing (ABM) has become the dominant targeting method used by the most successful, high-growth companies over the past decade. ABM focuses all marketing efforts on a list of companies that fit the ideal customer profile to fulfill a company’s strategy. ABM lists enable management teams to align efforts to capture specific opportunities that fit within the overall corporate strategy. Moreover, by focusing on a limited number of opportunities, marketing dollars are more efficient because they are spent on directed efforts where return on effort can be approximated by the capture of specific accounts. This is far better than the marketing to segments defined by firmographics such as industry, headcount, and geography. Yet, ABM is coming to the end of its lifecycle as artificial intelligence (A.I.) finally starts to deliver on its promise to predict demand with hypertargeting.
Marketers have long heard promises about A.I. transforming how to target ideal customers. Countless startups have received billions of dollars of venture funding to build the next generation of marketing artificial intelligence. Almost all failed. Why?
We believe there are two main reasons why A.I. has failed marketers (so far).
The Targeting Problem
Almost all marketing technology (MarTech) starts with the premise that industry classifications are accurate ways to determine if one company can sell to another company. Sure, companies sell into industry verticals. But accurately defining the industry for any given company is remarkably hard. Take our company as an example. LeadCrunch The problem of industry classification gets even harder as companies get larger. Take Apple Inc., the most valuable company in the world with a market capitalization of more than $1.5 trillion (as of today). Are they a phone company? Retailer? Music publisher? Software company? Yes to all of the above and a lot more.
The Targeting Solution
“We can not solve our problems with the same level of thinking that created them.” — Albert Einstein
Accurate targeting requires an understanding of how companies make buying decisions from two perspectives. First, we must understand what a company does from an economic perspective. This is their business focus. How do they add value to their customers? This is a question of differentiation: How are they different from their competitors? This cannot be answered with traditional data sources that list companies by industry. By grouping unique companies into a general classification (such as a SIC, NAICS, or proprietary industry code such as from ZoomInfo), we lose the ability to see what makes companies unique.
Our approach is to look at what are the array of activities that define a company’s business focus. This can be thought of as the degree to which a company participates in multiple industries. We do this by looking at the language companies use to describe themselves. The website for a business makes it clear why they exist and why they think they offer the best solution to a particular problem.
The second perspective necessary for accurate targeting is understanding how a business works. This is their operational profile. Businesses are nothing more than a collection of people who are organized to perform specific tasks to achieve a business focus (as defined above) using technology. So operational profiles must understand the people, systems, and technologies a company uses to create value. Again, the answers are hiding in plain sight on public sources of data, such as a company’s website, LinkedIn profile, job listings, etc.
Our approach is to constantly scrape data from these public data sources to gather a complete view of every company. Next we append data that we collect form running demand generation campaigns to create a more complete picture of that company, its employees, and technologies.
The final solution is to evaluate the operational profiles and business focus for millions of companies to predict who will form meaningful relationships. Our customer provides a seed list of either ABM targets, or prior customers. Algorithms then look for similarities between the seed list and the universe of possible prospects. The most similar emerge as lookalikes.
The following example illustrates how LeadCrunch uses our hypertargeting technology to drive revenue.
First, we are a demand generation company. So if we evaluate a universe of approximately 4 million businesses with operational profiles (sic how they do business) to find those that use language related to “demand generation” activities on their website, we find a weak signal (as indicated in the diagram below).
The poor results of looking for “demand generation” are easily explained by the dispersion of the list of seed companies (represented by red dots). To get good results from hypertargeting, the A.I. must find a buying signal that tightly clusters the seed lists (as represented in the “Final B2B Score” below). Here we see that the red dots clump together in the top right corner of the graph. They look like each other in terms of business focus and operational profiles to form a high “lookalike score” and they emit a strong buying signal as only selling to other businesses (Final B2B Score).
LeadCrunch uses this approach to target and engage prospects which is why LeadCrunch is one of the fastest-growing companies in the USA (as reported by Inc. Magazine).
The Data Problem
B2B data is notoriously difficult to keep current and accurate. This explains the rise of customer data platforms (CDPs) as a way for enterprises to organize, maintain, and manage their data from both internal and external sources.
We believe CDPs will fail to deliver high returns on investment because, like all data warehousing efforts, they are expensive to maintain. CDPs focus on consolidating data from many sources, such as the visitors to a company’s website, third-party data vendors, and data collected from sales- and marketing automation platforms. Of these sources, only the so-called first-party data from inside the company is likely to be current. The rest of the data needs to be constantly cleaned and verified.
The poor quality of data is the Achilles’ heel for CDPs. We conducted a study in early 2020 to evaluate the data accuracy of the top data vendors. The best were publishers such as IDG, Ziff-Davis, and TechTarget. The worst were the oldest and most established, such as D&B and Hoovers. We were surprised that ZoomInfo scored below InsideView, and DiscoverOrg (which acquired ZoomInfo in February 2019 but continued to sell data separately at the time of our study).
Contact data changes rapidly. Prior to the COVID Recession, several studies show that contact data accuracy decayed by 3% per month. This data volatility compounds over time, suggesting that 100% of the data will be inaccurate within 2 years of collection. The spike in unemployment in recent months makes this data even more volatile. It is safe to say that with at least 13% unemployment in Q2 20120, the data decay rate is at least 20% per month.
The Data Solution
The only way to keep data accurate and fresh is to constantly collect it at the speed and scale of the Internet. Although ZoomInfo, D&B, and other third-party sources claim to collect data in this way, we see little evidence that it is accurate enough to use in artificial intelligence targeting applications.
By combining the process of data collection with the process of identifying buying signals, LeadCrunch is able to collect and refine data as fast as the data changes. This approach requires an infrastructure of specialized machine learning algorithms that detect suspicious patterns that indicate data inconsistencies. Once found, the system automatically searches the suspicious company’s website and related data sources on the web to update the information. If the automated process fails, then it is flagged for human review and confirmation. Fortunately, only a small fraction of percent requires manual work.
Both ABM and lookalike targeting are emerging technologies — with ABM at least 10 years more maturity. We expect that the fastest-growing companies will migrate from ABM into hypertargeting by using both methods to A/B test results. We believe this so strongly that we now include hypertargeting in every campaign we run.