DMI - Case Studies

Prospect Model Analysis Case Study

The Problem:

A Client with an affinity group of 192,392 needed to target their appeal more effectively in order to reduce mailing costs, increase response rates and identify likely donors.

Five years of direct mail campaigns had provided Data Marketing, Incorporated (DMI) with solid baseline response data from the client’s other campaigns. The goal was to create an affinity prospect selection system that could provide ongoing improvement and targeted data selection for the affinity group in question.

The Solution:

The Affinity Prospect Scoring Model (APSM) was developed as a predictive tool that can be used for prospect selection from a preexisting set of data. Using logistic regression, the PSM analyzes which variables among the approximately applied 150 are significant predictors of response. Variables deemed insignificant are dropped from the model, while significant variables are used to create a model that predicted the probability of response.

For this client’s affinity group, the variables deemed most significant to response and value by the PSM were as follows:

Recency – Individuals who had visited more recently increased the model score.
Number of attempts – the modeled score decreases as number of attempts increase
Member Type – Individuals who had spent time on-site received a higher model score
Owner Type – Homeowners received a higher model score
Ethnicity – Asian surname members showed a higher model score; Indian and Hispanic surnames a lower score
Income – The model score increased as income increased

Geography was not included directly in the model score, but was treated as an independent factor in the affinity prospect selection process.

The Outcome:

Affinity Group Response by Job Number

Model Score	Response				Donation
Model Score	Response	Mailed	Response Rate	Response Index	Median	Average	Sum	$/Mailed Piece
Apr 2005	167	25,104	0.665%	117	$87.74	$50.00	$14,652.47	$0.59
Oct 2004	303	46,000	0.659%	115	$31.00	$116.28	$35,233.26	$0.77
Mar 2004	112	25,000	0.448%	79	$30.09	$65.44	$7,329.30	$0.29
Oct 2003	165	52,942	0.312%	55	$35.00	$99.97	$16,495.70	$0.31

The APSM was first applied in Fall 2004, and recalibrated for application to the Spring 2005 campaign. Letter copy and package were held constant across all four mailings. The APSM was quite successful in identifying prospects for both the Fall and Spring campaigns. Spring appeal campaigns generally produce 20%-30% fewer responses and lower donation averages than Fall/Holiday campaigns. Thus the fact that the April 2005 campaign was able to hold response rates equivalent to the October 2004 was extremely positive.

In addition, use of the APSM yielded an increase of 38 donors in Spring 2005 over Spring 2004, even though the number of outbound mailers remained constant. Using the APSM in Fall 2004 yielded 60 more donors, while mailing out nearly 7,000 fewer pieces than the previous year. In both cases, return per piece increased markedly.

The client was able to reduce overall mailing costs by contacting the target group with this special solicitation.

Affinity Group Response by APSM Score

Model Score	Response				Donation
Model Score	Response	Mailed	Response Rate	Response Index	Median	Average	Sum	$/Mailed Piece
.50	18	4,120	0.437%	65	$30.25	$161.47	$2,906.50	$0.71
.55	32	4,895	0.654%	98	$50.00	$49.61	$1,587.39	$0.32
.60	17	3,710	0.458%	68	$30.25	$59.52	$1,011.76	$0.27
.65	19	2,619	0.725%	108	$50.00	$84.20	$1,599.89	$0.61
.70	18	2,038	0.883%	132	$50.00	$184.96	$3,329.32	$1.63
HIGH	58	6,830	0.849%	127	$50.00	$64.91	$3,764.60	$0.55
Total	162	24,212	0.669%	100	$50.00	$87.65	$14,199.46	$0.59

When broken down into a response table by model score, it is clear that the PSM was able to accurately determine which prospects would respond at a higher rate, strongly reinforcing the value of the PSM.

Response by Model Score

The Conclusion:

The APSM proved valuable in targeting mailing efforts and maximizing ROI for both the Fall 2004 and Spring 2005 campaigns. Therefore, the client is considering applying the modeling tool next to its top tier donor data to identify:

Potential top donors to move out of the regular mail stream and into a more personal effort
Most and least likely to respond donors for differentiated handling
Small statistically valid cells for creative and copy testing

Don't know what binary logistic regression is? It's ok, we do!

DMI— transforming data for sustainable and measurable results.

DMI Industries

Our proven data solutions and expert fulfillment services ensure your project will be managed seamlessly from beginning to end.

DMI Industries

Our proven data solutions and expert fulfillment services ensure your project will be managed seamlessly from beginning to end.