October 6, 2016

Demystifying Predictive Lead Scoring

Written by

Marketers love to predict things. Heck, people do.

That’s why we enjoy pointing at line graphs in front of our co-workers and why we turn on the news every night to check the presidential polls. It’s why we love to say, glibly, “I told you so.” The notion of a future that might contain mystery and miscellany beyond our control is too much to stomach.

predictive analytics stock photo

“According to this graph, money. Because lines.”

It should’ve been no surprise, then, to see predictive lead scoring catch fire like a summer blockbuster, even before its marketing automation forebear had secured widespread adoption. Above the din of thought leadership blogs, webinars, and conference keynote sessions, a new cohort of SaaS vendors emerged—among them, Infer, Lattice Engines, Everstring, 6sense, Fliptop (acquired by LinkedIn), Mintigo, and Leadspace. These vendors all offer some variety of predictive lead scoring, predictive marketing/sales analytics, or one as part of the other.

ALSO READ: Why Drip Campaigns Work and How to Build a Good One

And of course, it’s never too soon or too comical to entertain predictions about predictive technology. The global firm MarketsandMarkets, for example, sized up predictive analytics as a $2.74 billion market that could grow to $9.2 billion by 2020.

Most of this early growth has been concentrated in and around Silicon Valley. According to a 2014 SiriusDecisions report, 78 percent of B2B companies using predictive lead scoring were in the high-tech industry. That percentage may have since shifted, but if I had to guess, I’d say predictive lead scoring is mostly being used by B2B tech companies to target other B2B tech companies—the usual tomfoolery.

I for one plan to be living in a climate-controlled space dome on the surface of Mars in 2020, but supposing MarketsandMarkets’ prediction about the predictive market is even partially correct, people will eventually need to know one thing: 

What is predictive lead scoring?

It’s Not What it Sounds Like

The first thing you should know is that predictive lead scoring (PLS) isn’t quite what it sounds like. It sounds like a way to assign a lead score before you’ve collected any real data or recorded any interactions with the lead.

That’s partially, theoretically true, but it’s misleading to think that you or any software tool can guess a lead’s level of qualification without empirical data. The distinction, rather, is in the source of that data and the way it gets interpreted.

With traditional marketing automation, you define a lead’s score based on demographic information and on behavioral data from their interactions with your site, emails, content, etc. Thus, all of the data feeding into your lead scores comes from your own media and your own systems. In most cases, this data becomes a number by passing through some arbitrary rubric created by a few members of the marketing team who, if you’re lucky, asked sales for input.

Traditional lead scoring is an excellent tool, but it does have drawbacks:

  • Your assessment is limited to known, encountered leads
  • Scores can be arbitrary and inconsistent

Predictive lead scoring doesn’t replace traditional, but it does add another dimension, and that dimension is based on value. Instead of assigning a point total to describe all of a lead’s known actions and known attributes, predictive scoring compares smaller samples of lead data to the statistical patterns of your closed customers. You discover not just which current prospects are a good fit for your product, but which ones are a good fit and don’t know it yet, and may have never interacted with your brand. Lead scoring meets B2B prospecting.

Kerry Cunningham of SiriusDecisions explains it this way: “Values . . . are generated based on their statistical relationship with outcomes of interest, rather than an assumed but unverified relationship between the attribute or behavior and the outcome of interest.” The predictive aspect here isn’t some uber-advanced, fortune-telling AI engine. It’s actually a simple concept: if you have enough data, you don’t have to wait for a prospect to take action before you can evaluate their buying potential.

Here’s an example, from EverString, of what it looks like to set up your “outcomes of interest” in a predictive marketing tool. EverString calls this a “predictive segment.” Other vendors use different terms like “ideal customer profile” or “Customer DNA.”

If you haven’t figured it out by now, a lot of the “magic” of predictive lead scoring involves an outside data discovery component — i.e. pulling and analyzing prospect data from various sources on the web and proprietary databases. In fact, for many vendors, the discovery component probably accounts for the lion’s share of their value proposition. That may not dissuade you from trying PLS/predictive marketing, but it should influence the way you compare vendors. Beyond features and integrations and the power of their “machine learning” algorithm, find out where and how the vendor is sourcing their data.

In summary, here’s how predictive lead scoring works:

  1. Identify common attributes and behaviors of closed customers
  2. Mine data from external sources and from your own database
  3. Identify and score leads with high buying propensity

Then, of course, your marketing and sales teams engage these leads and take necessary actions to nurture or convert them.

It’s a solid approach, and especially appealing if your current scoring process is underperforming. That’s why 90 percent of current users say predictive lead scoring provides more value than regular lead scoring.

PLS Is Not for Everyone

As they say in the pharmaceutical commercials, predictive lead scoring isn’t right for everyone. To be specific, it’s a bad choice for any business that isn’t already proficient with marketing automation. That means not only should you have a platform in place, but you should have a proven-effective lead nurturing strategy.

Why?

For starters, how can you expect an analytical tool to find sales and up-sell opportunities in your database if you aren’t keeping that data in one place and connecting marketing activities with sales outcomes? Secondly, PLS is good at discovering net new leads with buying propensity, but most of these leads will be cold. That’s the whole point of being predictive—to assign a score before you engage or even observe. If you’re looking for BANT-qualified leads and easy wins, you will be disappointed.

If, on the other hand, you set reasonable expectations and have a plan in place to engage what you find, PLS could become your newest “secret” weapon.