Predictive Lead Scoring: Get to Know the Basics

By now you’ve probably heard the buzz about a relatively new marketing technology called “Predictive Lead Scoring.” Marketers are excited about the prospect of using big data and machine learning to score leads more accurately than the usual rules-based techniques used in Marketing Automation.

This technology is developing fast and has enormous potential to improve efficiency and generate revenue. Right now it is in the early adoption phase, but eventually it will probably become a standard component of the marketing technology stack. Now is a good time to familiarize yourself with the basic idea of this technology, so you can assess whether or not your organization should consider adopting it.

Predictive Lead Scoring

Predictive Lead Scoring is a scientific method of predicting the probability that a particular lead will convert. It takes historical data from your CRM and behavioral data from Marketing Automation systems, and combines that with “big data” attributes gathered from multiple sources. The method uses this data to build a model of what a good lead looks like for your organization, then scores new leads against this model to determine how likely those leads are to close. That way you can correctly prioritize your sales efforts and focus on the leads that will generate the most revenue.

In other words, Predictive Lead Scoring uses data science and machine learning to uncover the hidden signals that predict the behavior of your prospects and customers: a dream come true for marketers.

Predictive modeling has been around for a while. However, using it in a business setting has generally involved hiring a team of data scientists to build a custom modeling system, assembling the required data themselves from whatever sources they can find. What’s changed is that advances in data science, computing power, and data collection have enabled vendors to offer Predictive Lead Scoring as a product you can buy “off the shelf.” Now you don’t need your own team of data scientists. You simply give the vendor access to your data. They build and test the model for you and insert lead scores directly into your CRM or Marketing Automation system with very little effort or expertise required on your part.


Traditional lead scoring can be a very effective way to increase conversion rates and make your sales operation more efficient. However, it has some drawbacks:

  • Guesswork – The rules used to score leads are determined manually by the guess-and-check method. Though we can make highly educated guesses as to what factors are likely to influence conversion and iterate to check against results, this process is slow and difficult in practice and not always satisfying. Predictive Lead Scoring automates the creation of model parameters, using statistical methods to give a rigorous, scientific result. This makes the building and testing of scoring models easier, quicker and more effective.
  • Incomplete Data – Traditional scoring only looks at the data you already have in your Marketing Automation and CRM systems, and usually only a tiny subset of that data is actually used in the model. Predictive Lead Scoring brings in thousands of data attributes to get a complete picture of the prospect: attributes such as financial data, social media presence, job postings, website technology use and sophistication, demographic and firmographic data and much more.

By replacing guesswork and incomplete data with scientific accuracy and big data, Predictive Lead Scoring can deliver better scores resulting in greater efficiency and more revenue.


There are several factors to consider when evaluating whether and when you should set up Predictive Lead Scoring:

  • Do you have enough leads? Predictive lead scoring helps you by identifying which leads are more likely to convert, so you can prioritize your efforts. These could be new leads coming in from your website or ads or other channels, or they could be existing leads already in your database. Either way there need to be enough of them that it is not possible to dedicate your full sales effort to each one. If you are short on inbound leads and you have a small house database and Sales is calling everyone with a pulse, you probably want to wait until you have a lot more leads before you worry about how to prioritize them.
  • Do you have enough historical data? You need to have enough examples of successful lead conversion to teach the scoring model what success looks like for your organization. Some modeling can be done for new companies and products, but it works best when you have an established product and a consistent sales process for that product, and you’re ready to make that process more efficient and effective.
  • Do you have traditional lead scoring, and if so how is it working? If you already have lead scoring in place, you should assess its effectiveness. If it’s working well, don’t mess with it! A simple way to check the effectiveness of lead scoring is to look at the conversion rate between MQL and SAL – the higher the conversion rate, the better the scoring model is working. According to SiriusDecisions (see this webinar on predictive lead scoring), a good conversion rate for B2B would be around 32%, leaving lots of room for improvement through Predictive Lead Scoring. If you don’t already have traditional lead scoring, you might want to implement that first and see how it does. However, in some cases, particularly if you have a very large volume of data and a complex sales process, it may be easier and simpler to skip straight to Predictive Lead Scoring and you may get better results more quickly. It mostly depends on the complexity of your existing processes and campaigns.
  • Are you prepared to make use of scoring? Predictive lead scoring only benefits you if you change your marketing and sales process to make it more efficient as a result of feedback from the scoring model. Is your organization ready to make those kinds of changes? Can Sales and Marketing align around shared objectives and a common plan of action? Can you quickly tweak and optimize your processes consistently across your organization? If not, you may find yourself unable to respond effectively to the information that your scoring model is giving you. Also, if you are planning major, unrelated changes to your sales and marketing operations (for example, introducing a new product) it’s probably not the right time to implement Predictive Lead Scoring, as the radical changes you are making might break your scoring model. Wait until you have a stable, consistent process that you can easily test and tweak.
  • How far along are you in your demand generation journey? Here at Intelligent Demand, we use a “crawl – walk – run” model to move our clients smoothly from “Marketing 1.0” to “Marketing 2.0” in a series of smart, manageable steps. If you are still in the “crawl” or “walk” phases, you are probably not ready for Predictive Lead Scoring yet and there are likely lots of other big wins you can get more cheaply and easily. Predictive Lead Scoring is a good fit for organizations who are already in the “run” phase – they have used a solid marketing strategy to build a successful revenue engine with good sales and marketing alignment, and they are looking for ways to make that engine run more efficiently.

If you don’t meet these criteria or if you’re uncertain, it is probably best to focus on the basics – get your demand generation engine built and running – before you try an advanced technique like this. As always, Intelligent Demand can help with that. But if your demand generation is already advanced and you are looking for ways to take it to the next level, it might be a good time to explore this technology further.


To implement Predictive Lead Scoring, you will need to select a vendor. However, before you start looking at vendors, you should do a thorough assessment of your requirements. What exactly is the event you are trying to predict – is it conversion all the way to closed/won, or some other event like conversion to SAL? Do you need a single scoring model, or do you have multiple products and/or processes that would have to be modeled separately? Are you trying to score new inbound leads as they come in, or score the existing leads in your house database, or both? What benefits do you hope to attain? Do you need to just score the leads you have, or do you want to bring in net new leads as well? Do you need to append the data attributes used by the scoring model to your contacts? How much visibility do you need into the scoring model and how it works?

Once you have a clear understanding of what you need, you’re ready to start evaluating vendors. Looking at Lattice, Infer, FlipTop and Mintigo would be a good place to start, but shop around because this is a hot market and there are several promising new vendors who are just now entering it. Look for competition to drive innovation and price reduction in the near term as the market heats up.

So what do you think? Are you excited about this new technology? Are you considering using it, or have you already tried it? How do you think your organization could make use of the ability to accurately predict your prospect behavior? What would you do with that kind of power?

If you have questions about Predictive Lead Scoring (or any other demand generation innovation for that matter), or if you need help with assessment, evaluation or implementation, please visit our contact page or reach out to us at

Eli Snyder

Eli is an old-school geek who fell in love with marketing technology. In addition to writing code and setting up advanced configurations in marketing cloud applications, Eli helps ID clients navigate the ever-shifting martech landscape. At some point, Eli will probably ask you for admin-level access to all of your systems (if he’s working on a project with you, or even if you just meet him randomly at a conference). It’s OK though, he knows better than to touch anything without permission, and with permission, he’ll make them hum like a Tesla roadster.