One of the most common elements of any business is leads. These are potential customers who have either made contact with you or you have contacted them, but you haven’t yet started moving them through the sales process (or conversion funnel). The bigger the business and the lesser the human resources, the harder it is to spend as much time with every lead you like.

A traditional business has hundreds and thousands of leads coming in. Talk to any sales guy, and he will admit only 2–10% leads convert, so why are businesses wasting time and money on the other 90% plus leads? Simply because, as the Marketing and Sales team, we are yet to adopt AI.

Introducing Lead Scoring

Lead scoring is a methodology used to rank prospects against a scale representing the perceived value each lead represents to the organization. The resulting score is used to determine which leads will get a receiving function (e.g. sales, partners, teleprospecting), in order of priority.

Lead scoring models incorporate both explicit and implicit data. Explicit data is provided by or about the prospect, for example — company size, industry segment, job title or geographic location. Implicit scores are derived from monitoring prospect behavior; examples of these include Web-site visits, whitepaper downloads, or email opens and clicks. Additionally, social scores analyze a person’s presence and activities on social networks.

After you’ve determined what you think are the best attributes that describe your leads, you can allocate points to each one, giving more points to the most important attributes (and you can even give negative points for “bad” behaviours, like unsubscribing from a mailing list). Once the points are tallied, sort your list in descending point value, and you are left with the “hottest” leads at the top of the list. These are the leads that you should pursue now, while others lower on the list need more cultivation.

‍

Only 2–10% leads convert, so why are businesses wasting time and money on the other 90% plus leads? Simply because, as the Marketing and Sales team, we are yet to adopt AI.

Lead Scoring allows a business to customize a prospect’s experience based on his or her buying stage and interest level and dramatically improves the quality and “readiness” of leads that are delivered to sales organizations for followup.

Problems with traditional Lead Scoring

why are businesses spending time and money on the other 90% plus leads? Simply because, as the Marketing and Sales team, we are yet to adopt AI.

In a traditional lead scoring, you have predefined criteria multiplied by weighting factors to deduce the leads’ total score. The problem is two-fold. First, you are using your past knowledge to give weighting factors; for instance, you could give a lot more weight to the type of email address (whether a company email or personal email). Secondly, the criteria that you are using may not be even predictive of lead conversion. There is just no room for intuition in the world of marketing.

Another issue is — this whole process is not automated and every time a lead comes into the database, people are manually giving them a score, making it more error-prone.

Further, every time your intuition goes wrong, you have to go back to old data and make adjustments. For example, earlier, you were rewarding leads from California, but this time the leads were not so good from California, so you go back and make adjustments and reduce some score for California (again using some manual calculations and intuitions)

Further, every time your intuition goes wrong, you have to go back to old data and make adjustments. For example, earlier, you were rewarding leads from California, but this time the leads were not so good from California, so you go back and make adjustments and reduce some score for California (again using some manual calculations and intuitions).

How AI can automate Lead Scoring and make it 10 times better

Though the traditional scoring mechanism is helpful, it does not provide you a prediction as to whether the lead will convert or not. Instead of defining attributes and their corresponding weight factors, predictive lead scoring uses existing sales data, along with data mining and analytics techniques, to build the ‘ideal’ lead. Subsequent leads are compared with this lead and correspondingly labelled qualified or not.

In predictive lead scoring, the marketer defines the KPIs used in the analytics so the model algorithm can create a formula for automatically ranking leads so the marketer can quickly identify the most qualified ones. The model is continuously fed with data on leads that successfully converted to customers or those that failed, eliminating the need for ‘run and check’ processes.

A combination of predictive and prescriptive analytics can be used to take actions and measure deliverables. Leveraging predictive and perspective analytics gives marketers a solid, robust scientific approach to effectively increasing conversion rates. Read how a case study of how a company saw a significant increase in conversion.

Predictive modeling is the focus of this post, but it’s also worth mentioning causal modeling, which focuses on determining which factors effectively influence outcomes. This could help determine which kinds of sales strategies are most effective, but it does not help predict new leads’ conversion. We will cover causal modeling in a future post.

Many organizations build a collection of these heuristics over time, and they use them as a strategy to triage new leads. A set of heuristics can often be represented as a decision tree diagram or a series of true/false statements that someone can follow to make a consistent prediction about a new lead outcome. Using our example variables from above, we might end up with a decision tree like this:

What are the limitations of predictive scoring?

Predictive scoring is not for all companies. Here are the shortcomings —

• Need lots of historical data: Since models depend on historical data, you need to have some sales history. You shouldn’t try this until the leads become unmanageable.
• Need lots of data points for current leads: More the data better, collect as much as possible, for instance, their web activity, social media activity, join it with third party data, for example, their credit score, annual spending etc.
• Most predictive scoring systems are black-boxes: The simple models like logistic regression and decision tree can easily be explained but more complex algorithms like ensemble models, lightGBM are harder to explain. We use more complex AI models because they are more accurate.
• Your target market and product should be stable: If your target market changes, your sales data loses its value for analytics. However, if you know the critical junction after which your company decided to focus on a certain market, you could take data from that point on to have training data for the prescriptive model.

‍