What are some common myths about lead scoring?

Hi Lauren! There's a doozie that says lead scoring does not require active human involvement ... that it can be entirely automated.

True, marketing automation systems can accurately track on-line actions, site visitors’ “digital footprints,” pages viewed, time-per-page, and so on. But they cannot discern the business motivations behind each of those. Absent the reasons, one is seeing raw data, but is not yet ready to place a value on it. In other words, marketing automation “scoring” is incomplete at best, misleading at worst.

Enter the inside sales / business development function. For genuinely actionable, validated opportunities, BtoB companies cannot do without it. It is not the only way to qualify and pass leads, but it is integral to an effective account and customer acquisition effort. In BtoB marketing and sales, the combination of smart humans + technology stand-alone technology every time.

I agree with Michael that human intervention -- in the form of a lead qualification phone call -- is a hugely productive way to gather more information about a lead's interests. But however you've gathered that data, it should be combined with all other available information and plugged into an objective scoring algorithm. The alternative is to let the inside sales people assign a score based on their personal judgment. One of the biggest myths I see is that people can do this accurately -- and the reason the myth persists is that so few companies actually do any research to determine whether the scores are correct (i.e., how well they correlate with future behavior).

The same problem applies to manually-built scoring formulas, created with the classic "let's all sit around the table and decide on some rules" approach. Again, the problem is that few companies actually test how well those formulas perform. Carlos is pointing to this by saying that building the model shouldn't be a one-time exercise. But, again, he seems to have in a mind a meeting between marketing and sales rather than a more objective statistical analysis.

It's rare to see a company even analyze past data to provide guidance to decision-makers about what factors actually correlate with future results, and even less common to see a statistical model that uses the data to create a model without human intervention. We've done some testing of automated methods and found they work quite well -- although we're still in the early stages of our research (and are looking for more test cases, if anyone is interested).

Another, related myth is that lead scores should be assigned once, typically when the lead first enters the system. In fact, lead scores should be continuously recalculated as the lead continues to interact with the company. The revised scores reflect the progress of the lead's own purchase cycle and the evolution of the lead's relationship with the company. Those inherently change over time, as opposed to the more-or-less static BANT characteristics that determine whether someone is a good prospect.

To make things even more complicated, a REALLY good scoring approach would actually change the scoring formula itself over time: activities that are significant early in the cycle are less important (or perhaps even negative indicators) later in the cycle. So that's yet another myth: that, for each product and customer segment, the same lead scoring model should be applied to everyone.

Piggy backing on Carlos' answer, another myth is that marketing automation as a platform covers your bases when it comes to lead scoring. While over time you may be able to factor in most demographic and behavioral criteria, there will always be knowledge outside the interactive environment that requires manual intervention.

For example, if a salesperson knows that a qualified lead has left the target company, it still requires manual intervention to adjust the score as soon as possible - and someone has to make a decision whether and how to track the individual prospect or qualified lead who left the company, presumably to another company where another opportunity may eventually arise.

Does or should some part of the person's score "carry" over to the next place? How do you even know that you have the same "John Smith" when he used to be jsmith@abc.com but now is jack.smith@xyz.com? How do you make sure that the historic data of interaction and scoring maps over? What happens to the score of the individuals or company that the person left?

How big are deals that this kind of granularity in a scoring model needs to be heeded?

So yes, as Carlos says, you must "revisit, revise, and test your model." And you must also determine how complex your model needs to be based on "outside the internet" knowledge and how critical a deal is.