Taking the Lead:
How Predictive Modeling Is Changing
Lead Acquisition Strategies

White Paper June 2008

Datamark
2305 President Drive
Salt Lake City, Utah 84120
www.datamark.com













Table of Contents




I. Executive Summary

II. What is Predictive Modeling

III. Speed of Development: Weeks Not Months

IV. The Inefficiency of Working Poor Online Leads

V. What Does Datamark and eBureau Offer

VI. ROI on Marketing Money Spent

VII. Conclusion











I. Executive Summary

There is a way to more precisely ascertain the quality of an online lead: Predictive Modeling. Datamark’s partnership with eBureau’s predictive modeling service has created a cutting edge system that gives schools the capability to accurately measure the likelihood that a specific lead will convert into a start and apply that data to online media buys in the form of a lead score in real-time. Predictive Modeling does not simply score an entire category of leads; it scores each lead individually, enabling school marketers or admissions agents to make a decision about each lead as they receive it.

Schools can now exploit this opportunity to capitalize on advanced data collection, storage and analytics software to mine golden information pertaining to their students and prospective markets. Predictive lead scoring is destined to become one of the most effective and proven business intelligence practices used by the education industry.

By placing the focus of school intelligence gathering on power, speed and simplicity, Datamark and eBureau have made predictive scoring affordable and practical for school marketers seeking to improve student acquisition and retention efforts. Knowledge is power, and participating schools gain a formidable competitive advantage by unlocking the recruiting knowledge hidden within their customer data in order to make better and timelier decisions throughout the customer lifecycle.



II. What is Predictive Modeling?

Predictive Modeling is a statistical methodology that can predict a desired outcome based off of historical performance. In the education industry, the desired outcome can be many things like a response, an enrollment, a start, a retained student or a graduate. Modeling is a very clear way of separating the good from the bad, good defined as leads yielding the most desired outcomes vs. leads yielding the least desired outcomes.

The use of statistical models to predict the behavior of customers and prospects is a proven, valuable method for schools to boost enrollment, reduce costs and increase profitability. This process is similar to the concept of credit bureau scores used by lenders in their decision to grant consumer loans. Predictive Modeling uses underlying statistical models that enable decision makers to create automated policies about what to do with each lead, thereby maximizing conversion rates and minimizing costs.

Predictive Modeling begins by taking historical lead and start data, then appending thousands of additional data elements like identity data, purchasing data, credit data, etc. Then the modeling software, through hundreds of algorithms, develops and resolves the model that produces the best fit to the actual data and holdout sample. The most accurate predictors are determined in the model and their relative weights to the prediction of the desired outcome. Once the model is tested and validated, it is then able to score leads from internet sources in real time.



III. Speed of Development: Weeks Not Months

A major obstacle preventing schools from incorporating an effective lead scoring program is the lack of an established back-end data infrastructure. Designing an operational custom lead scoring system from the ground up demands specialized expertise in statistics; expensive IT investments in data warehousing, data mining and analytics tools; and procurement of expensive third-party consumer data.

Development also requires lengthy cross-functional coordination between multiple departments and repeated complex data translations between offline warehouse and online production systems. In other words, the do-it-yourself method is a budget and logistical nightmare rife with error and can take six to twelve months to complete.

After enduring the pain and suffering during the development phase, customized scores can then be applied in a production environment. However, because of the cost and complexity, many schools that can benefit from analytics lack sufficient resources to justify its implementation. Some schools have tried shortcuts, but data-poor or generic off-the-shelf scores fall short and deliver inadequate and inconsistent results. Because of these limitations, few, if any schools have enjoyed the benefits of precision scoring calculated on enough essential data to be highly predictive and consistent. As a result, lead scoring remains an enormously valuable, but underutilized, tool for school marketers until now.

Datamark’s partnership with eBureau provides clients with a customized, high-end lead scoring solution, which provides unprecedented performance, speed and ease that enables these schools to fully reap the benefit without all the costs and pain borne by attempting to develop a similar system in-house.

Predictive Modeling precisely integrates data, analytics and processing components within a cohesive, specialized scoring architecture. This modeling system assimilates past results with eBureau’s immense repository of identity, behavioral and purchase data, with a multiplicity of complex statistical models (e.g., regression, OLS, CHAID, neural networks). Additionally, 50,000 pre-built modeling variables are utilized to generate custom models around individual, household or neighborhood (9-digit zip code) characteristics. This system will then begin producing customized predictive scores for a school in a day, rather than weeks or months.



IV. The Inefficiency of Working Poor Online Leads

Cost per Lead (CPL) online advertising is an important source for many school marketers in attaining monthly and quarterly enrollment goals. Consequently, online lead generation has become the fastest-growing segment of online advertising expenditures, soaring by a whopping 74% to $1.3 billion in 2006, according to the Interactive Advertising Bureau (IAB). Unfortunately, the reality is most of the CPL costs are being used to chase down poor-quality leads to their inevitable dead ends, which, in the end, exceeds the price paid for the leads.

According to a 2006 Eduventures Higher Education Survey, 46% of for-profit institutions use vendors to generate up to half of their monthly volume of leads and 86% of institutions use lead-to-enrollment rates to evaluate lead quality. Poor lead-to-enrollment rates were cited by schools as the greatest concern of using online lead generation.

Despite the enormous amounts of money being invested in online lead aggregators, schools have weathered dramatic fluctuations in the quality of these leads; ranging from a handful of very good leads with high conversion rates to a substantial block of very poor leads with abysmal conversion rates.

For a school, the most accurate and direct measure of a lead’s quality is by certain traits, such as: lead exclusivity, completeness of data, data verification, incentive promotions, lead origin and consumer motivation et al, and, most importantly, whether it converts into an actual start.

To prevent a school marketing budget free-fall, Predictive Modeling provides a simple three-digit score that can be used in real time to assess each new incoming lead and to automate decisions. By scoring every lead received from the lead source instantly, marketers no longer need to fly blind with their lead management efforts.

Marketers will have a reliable measure of a lead’s probability of actual conversion enabling them to objectively assess the quality of every online lead. A high-scoring lead equates to a high probability of converting into an actual customer, while a low-scoring lead equates to a low probability of conversion. Admissions staff will be able to reduce unnecessary expenditures on poor leads by focusing on leads that have a significant probability to convert.

Predictive Modeling will make the lead aggregation business more accountable for bad affiliates that they have historically been able to hide in their lead mix because they were judged on the average. Scoring can also be used to evaluate which online lead sources are performing and which need to be cut. It totally changes the game.



V. What Does Datamark and eBureau Offer

Datamark’s exclusive partnership with eBureau brings a powerful combination to the table. eBureau has built extensive data assets, extremely fast and accurate modeling capabilities and the real time scoring engine. This trove of data resources combined with Datamark’s seamless integration of LeadBin, our lead management system, creates a full suite of media management tools to leverage the scores to improve our clients cost per starts in truly meaningful ways. We are talking about cost per start reductions in the 20-30% ranges right out of the gate.

The Predictive Modeling system produces highly accurate results that schools rapidly exploit to realize the complete benefits of highly predictive lead scores. Unlike past predictive scoring solutions, this solution requires:


• No upfront, expensive and time-consuming professional services projects
• No specialized software or hardware purchases
• No additional hiring of specialized staff
• No purchases of overlay consumer data
• No lengthy integration exercises

Predictive scoring models are only as good as the data upon which it is based. Most schools rely on internally generated customer data for scoring because third-party databases are difficult and costly to integrate into their existing warehouses. As a result, they forecast consumer behavior with incomplete information, limiting the accuracy of their scores.

Data-rich content is where eBureau’s models set themselves apart. This unique system is endowed with access to eBureau’s data warehouse containing more than 125 billion unique records on over 210 million U.S. adult consumers. This data includes; consumer credit data, real property records, household demographic and identity files, unique catalog and direct marketing purchase histories, payment histories, and various other public records, including bankruptcies and deceased files. More than 200 Terabytes (trillion bytes) of continually updated consumer data is routinely refreshed to ensure all information remains current and approximately 3 billion additional records are added monthly.



VI. ROI on Marketing Money Spent

Real-time predictive scoring of online leads has major implications for school marketers who rely heavily on aggregator-fed CPL advertising. First, it provides schools with protocols dictating which leads to accept or outright reject. Also, because many CPL agreements include “scrub rate” provisions, scoring presents a simple way to isolate leads for rejection.

From a school marketer’s perspective, the true value of a lead can be determined with a single, objective measurement: lead Return on Investment (ROI). Most CPL-centric advertisers discovered this conclusion the hard way by sinking a lot of time and advertising capital into inflated CPL costs and fruitless affiliate campaigns.

Predictive Modeling substantially benefits education marketing by changing how leads will be managed going forward. In the past, leads were managed on averages. When vendor’s average cost per start got too high they got pulled from the buy. Now with Predictive Modeling leads are managed at the lead level not the vendor level which enables Datamark to make improvements in tighter increments which drive results for clients.

The true brilliance of Predictive Modeling is that it can be used to set a rational, efficient, dynamic, tiered pricing structure for leads generated from specific third-party suppliers. Schools can opt to negotiate a lower cost for lower scoring leads or choose not to purchase them at all. Since they’re saving money by reducing or eliminating questionable leads, they enhance their ROI by paying more for premium, higher-scoring leads.

The graph below illustrates that when Predictive Modeling is employed 75% of all leads accounted for 97% of all applications. This means that a school not using Predictive Modeling and that paid $40 per lead could have reallocated or saved $501,240 in media costs.



VII Conclusion

Ultimately, the costs of CPL advertising need to be weighed against the returns, which can be measured objectively. In dealing with third parties that provide large numbers of low-cost leads, marketers need to be smarter than ever in managing both advertising costs and operating costs in the lead-to-conversion process.

Just as banks use credit scores to make instant decisions about whether a consumer qualifies for a loan and at what interest rate, our Predictive Modeling can help school marketers to effectively assess CPL buys, and in the process, maximize ROI from their CPL advertising budgets.

Although maximizing the performance of online generated leads using external consumer data and Predictive Modeling may sound complex and daunting, Predictive Modeling is available as an easy, automated and cost-effective outsourced scoring service, and represents a new best practice in CPL advertising.



####



Hosted by www.Geocities.ws