The use of predictive modeling to score online consumer sales leads has been embraced by interactive marketers across a wide range of product and service categories, from education, insurance and financial services to telecom and automotive. Real-time sales lead quality scoring has fundamentally changed the dynamics of lead pricing, budget allocations and conversion ROI for advertisers and lead providers alike. So what are the secrets behind these algorithms, and what makes them effective for lead scoring?
Predictive scores have consistently proven to be the best way to pre-judge lead performance. From our experience, it is common to have between 100 and 200 predictive variables in a lead quality scoring model, with each variable contributing between 0.3% and 4% to the make-up of the final score. The variables are typically a mixture of individual, household and neighborhood-level attributes depending on business objective.
Some of the predictive variables in scores are intuitive, some are surprising, while others are quite esoteric and obscure. As a result, the variables and the weighting applied to each variable may be hard to interpret individually, but when used together in a score, makes perfect sense for predicting which leads are most likely to result in a sale.
From our experience, there are generally three categories of predictive variables that are important for lead scoring. At a macro level, roughly 20% of the predictive variables are related to data validity, 20% capture some notion of user intent, and 60% of the variables are based upon the individual consumer. Let's take a look at each of these factor groups and what a lead generator can do to boost lead scores and the quality of their leads.
Data validity is the accuracy and persistence of the personal information associated with the lead. Data validity answers questions like, is the lead "Mickey Mouse at 123 Main Street" a real person? Is the email address from a domestic or foreign IP address? Is the landline phone number accurate and within the same ZIP code as the address?
These factors and many others relate to the "contact-ability" of a lead. Leads with falsified, missing or inaccurate data will generally score lower in a lead quality score. If the lead buyer only operates in the U.S., leads that have a foreign IP address or email address are likely to get scored lower. Phone numbers also provide important scoring variables and clues. Is the phone number mobile or landline? Empirically, mobile numbers tend to produce a higher contact rate and, eventually, a higher conversion rate. Verification services exist today that can assess in real time the validity of the contact information; some can automatically correct typos in phone numbers and address information. A "mis-key correction" feature can reduce subsequent verification filter rates by 3% or more.
Predictive factors related to user intent is the second category of predictive variables in lead scores and provide clues about the mind-frame or motivation of the individual who completed the lead form. There are a number of factors that are good proxies for user intent. The time and day of week that a user completes a lead form sometimes provides valuable clues. Is the lead coming from a work IP or home IP address? A lead submitted at 3 p.m. on a Tuesday is in a different mind-frame from a lead submitted at 3 a.m. over the weekend.
Sometimes user intent is captured overtly on a lead form. For example, a typical lead question is: "How soon do you intend to buy?" The media channel where a lead comes from also impacts user intent factors. A search lead generally has higher user intent than a co-registration lead. Sometimes the factors are correlations of different data points. For example, in the online education sector, if a prospective student (a lead) indicates that she wants to attend a campus-based university that is 89 miles from her home address, it is a fairly safe bet that this particular lead will take a hit on the quality score. It is rare that a student will drive 89 miles to attend a university.
Finally, some lead forms have optional fields. Users who choose to complete the optional fields (such as a secondary phone number) consistently score higher and have a higher sales rate. It stands to reason that someone willing to give you more information about themselves is a more motivated prospect with higher intent. However, it is not recommended that form fields be manipulated in an effort to boost a lead score. What lead sellers can do to boost user intent is to make sure that all prospects have a good understanding of, and a legitimate interest in, the product or service. Sellers can also take steps to encourage prospects to complete any optional form elements. This may hurt form conversion a bit, but it can also help weed out some of the lower intent leads.
The remaining 60% of predictive variables typically come from variables that are specific to the individual user. In other words, how much does the lead "look" like others who have purchased the particular product or service in the past? When a lead is more like a previous purchaser in terms of where the person lives, what the person owns, what interests the person has and what past purchases they've made they are more likely to score higher. Most of these factors have been used individually by marketers. but effective scores assign values to all of these variables collectively to create a useable index for lead management decisions. There are also more esoteric predictors, like whether the lead has a credit card associated with a frequent flyer program and how much they spent on dining out last month. The combination of who you are demographically and what you do behaviorally forms the basis for comparison to the "ideal" lead prospect.
As with all marketing activity, it is crucial for lead providers to target the appropriate audience and craft campaigns that speak to the hearts and minds of their ideal prospects. Since 60% of the predictive power of a lead score comes from individual user information, attracting the right audience is essential.
All of this data is not useful unless you can easily access it and act upon it. There are real-time, web-based visual dashboards that provide lead quality insights and detailed demographic data at the lead source or campaign level. Such solutions reveal which campaigns are producing the best quality leads and why certain sources outperform others.
In the end, targeting the right audience with the appropriate message, ensuring accuracy and completeness of form information, and establishing legitimate user intent will not only produce higher-scoring leads. It will produce better-quality leads.