.png)
Dale W. Harrison likes to run a poll: imagine a company about to sign a $5 million annual contract, renewed every year for five years. Who would be involved in the buying decision? Most people guess the CEO, certainly the CFO, perhaps a CTO or another senior executive.
Then Dale reveals the deal: Google's contract for toilet paper in its employee restrooms. Google spends roughly $5 million a year on it. Nobody's CFO is sitting in the meeting.
The point isn't the toilet paper. It's how quickly we fill in the blanks when we imagine who makes a B2B buying decision. We assume the most senior people are involved, construct buying groups around those assumptions and then build targeting and lead-scoring strategies around them.
Episode 7 of B2B Effectiveness: Evidence-Based Marketing Ideas for B2B Practitioners, featuring Dale W. Harrison and Liam Moroney, looks at what happens when those assumptions meet the available evidence.
Can better lead scoring define better targeting?
Most organizations approach targeting and qualification as two separate steps. First they define an ICP and build a target audience around it. Then they use lead scoring to decide which of those prospects are worth passing to sales.
Dale argues that the relationship can work in the other direction. Your own historical data – the accounts you've actually won and the accounts you've actually lost – can tell you which characteristics genuinely distinguish good opportunities from bad ones. Those findings can then shape the next round of targeting, creating a feedback loop that becomes more useful as more outcomes accumulate.
The important distinction is between finding what your customers have in common and finding what separates your customers from the companies that didn't buy.
Why lookalike lists keep pointing at your losses too
Consider what happens when a marketer uploads a list of closed-won accounts to an ad platform and asks it to build a lookalike audience.
The platform can identify characteristics shared by those accounts: employee count, revenue, industry, geography and countless other attributes. But unless it has information about the accounts you didn't win, it has no way of knowing whether any of those characteristics actually contributed to the outcome.
If 90% of your closed-won accounts are based in California, for example, the model may conclude that California is an important characteristic of your ideal customer. But what if 90% of your closed-lost accounts are also based there?
The pattern is real. It just isn't useful.
That's the same problem the series explored in Episode 6 with the deliberately absurd example of vowels in company names. A characteristic that appears in all your winners isn't necessarily a useful signal. It becomes useful when it helps distinguish your winners from your losses.
Buying committees aren't new, and they're mostly invisible
The problem becomes even harder when the model starts trying to identify the people involved in a purchase.
A lot of current B2B marketing language treats buying groups as though they represent a relatively recent change in how businesses buy. Dale points out that the concept goes back decades. A 1965 Harvard Business Review article discussed buying committees in large organizations, drawing on research and practices that dated back even further.
What has changed is the amount of data marketers can collect about individuals. That can create the impression that we have much better visibility into the buying process than we actually do.
Dale's experience working with Fortune 500 procurement teams illustrates the gap. Vendor targeting often assumes that senior decision-makers are actively researching suppliers online. In reality, the people handling the practical work can be much further removed from the people marketers expect to find. In some of his examples, vendor materials were being distributed and reviewed by junior employees who had never even visited the supplier's website.
The buying process may be complex, but that doesn't mean the people involved leave a neat digital trail.
How much of the buying group can you actually see?
That matters when account-based targeting starts treating inferred buying groups as established fact.
Dale puts the distinction in numerical terms: around 65% de-anonymization may be achievable at the account level, but only around 5-7% at the level of an identifiable individual. That leaves a substantial gap between knowing that activity is happening inside an account and knowing exactly who is involved in the decision.
Filling that gap with assumptions can create a very precise-looking targeting strategy built on very little evidence.
Working backwards from your own history
The alternative is to start with outcomes.
Look at the accounts you've actually won and the accounts you've actually lost, then ask what genuinely distinguishes the two groups. That requires looking for differences rather than simply cataloging the characteristics your customers happen to share.
From there, buyer-seller fit adds another important layer. An account can look attractive in terms of size, industry and apparent demand, but that doesn't mean your organization is particularly well equipped to win its business. Your own sales history contains evidence about where you perform well and where you don't.
The result is a targeting model based on what the business has actually demonstrated it can win, rather than an idealized picture of the companies it would like to sell to.
How would you know if it's working?
The test is relatively straightforward: watch what happens to the quality of the leads reaching sales.
If a BDR has to spend five minutes qualifying a lead only to reject the overwhelming majority of what the scoring model has passed across, the model isn't adding much value. The more useful question is whether the proportion of leads accepted by sales improves as the targeting and scoring become more informed by actual outcomes.
Dale uses the example of two slot machines. If one pays out once every 100 pulls and another once every 50, the second machine is twice as productive even though neither looks particularly impressive in isolation.
That is important when the starting point is already low. If only around 1% of leads ultimately become customers, as long-running B2B benchmark data suggests, improving the quality of the opportunities reaching sales can have a significant effect even without producing anything close to certainty.
And even perfect information about buying intent wouldn't solve the whole problem. Knowing that someone wants to buy doesn't tell you whether they will buy from you, or whether your sales organization is particularly good at winning that type of business. That's the buyer-seller fit question that has run through the series: a company can be an excellent prospect for the category and still be a poor prospect for a particular seller.
Data is not information
That brings the episode back to one of the central ideas of the series: having more data doesn't necessarily mean having more information.
Knowing the last hundred results on a roulette wheel tells you nothing about the next spin. Knowing the last million doesn't change that. The volume of data has increased, but the information relevant to the next outcome hasn't.
The same distinction applies to B2B targeting. Millions of firmographic and behavioral data points can create an impressive picture of an account without necessarily improving your ability to predict whether that account is worth pursuing.
The useful work is identifying the relatively small number of signals that actually distinguish outcomes, then using those signals consistently.
And that brings the series back to where it started: qualification isn't about finding a score that tells sales exactly who will buy. It's about improving the odds, using evidence from the business's own history to decide where sales effort is most likely to pay off.
Next up: Dale and Liam turn to what happens when the market itself is changing – and why a category being “eaten alive” can dramatically alter the number of buyers actually available, regardless of how good the targeting is.
