predictive edge's smart recruitment predictor provides real solutions to the difficult questions faced in enrollment management:

• analysis of the institutional data to determine the critical variables.
• analysis of geodemographic data to determine elements relevant to the institution's recruiting effort.
• mapping of the density of the distribution of inquiries, applicants, and matriculants by zip code.
• evaluation of counselors and admissions programs based on the strength of the inquiry or applicant pool.
• the probability of enrollment of each inquiry. continuous scoring updates are available throughout the admissions cycle.
• ongoing assessment of the each admissions cycle based on the strength of the inquiry pool.

• inquiry source
• eps region
• high school
• probable major
• sat scores
• campus visits

• median income
• median home value or median rent
• percentage of population under 17

visit code	inquiry	dn apply	applied	app/inq	paid	paid/app	paid/inq
summer program	1	1	0	0.000	0	0.000	0.000
art tour	7	2	5	0.714	1	0.200	0.143
math competition day	127	114	13	0.102	6	0.462	0.047
summer institute	67	49	18	0.269	8	0.444	0.119
off-campus interview	185	56	129	0.697	28	0.217	0.151
campus tour	151	85	66	0.437	32	0.485	0.212
junior open house	191	100	91	0.476	39	0.429	0.204
accepted students' open house	126	0	126	1.000	65	0.156	0.516
fall open house	299	91	208	0.696	86	0.413	0.288
on-campus interview	969	319	650	0.671	274	0.422	0.283
no entry in visit code 1	80086	77092	2994	0.037	729	0.243	0.009

this data allows us to assess the effectiveness of all the programs. obviously some of the programs are effective and others are not. in this abbreviated table we can see that if we can get students on-campus for a visit the chances of the student applying go from 3.7% to around 70%, but the math competition day is not a particularly effective recruiting tool. this component analysis is carried out on all significant variables. there are similar reports by high school, major, etc. these reports, of course, run for pages. there is also much to be learned from examining a variable as straightforward as the sex of the inquiry. in the case shown below, and in most cases, women inquire far more frequently than men. but their rate of application (apps/inquiry) is only about 65% of that of the men. it is also clear that if the institution didn't know the sex of the inquiry, they were far less likely to convert that inquiry into an applicant:

sex	inquiries	dn apply	applied	app/inq	paid	paid/app	paid/inq
unknown	2322	2285	37	0.016	9	0.243	0.004
female	45420	43621	1799	0.040	494	0.275	0.011
male	31134	29105	2029	0.065	567	0.279	0.018

the analysis of data by zip code by its nature includes the geodemographic data found to be significant. this data is critical not only in modeling the admissions process, but in analyzing new territories as well.

moreover, there is no substitute for being able to see the geographic distribution of your inquiries, applicants and deposits. predictive edge provides these maps as part of the smart recruitment predictor. with these maps, it is possible to examine the regions at various levels of detail to determine the efficiency of the recruiting effort.

predictive indices

when we analyze your college's data, we construct predictive indices in two distinct forms: static and dynamic. a static 'fit index' depends on parameters assembled early in the decision process--parameters that typically do not change during the process. a static index might include origin of inquiry, date of inquiry, home address, high school, geodemographic data, and any other relevant inquiry data. by contrast, a dynamic, or 'hot index,' is updated throughout the process as new information is logged, such as a campus visit, phone call, or e-mail to the admissions office, sat or act scores received, and additional contacts.

this construction allows for differing uses of the predictor throughout the admissions cycle. the static fit index can be utilized to make strategic decisions early into the admissions cycle, such as which high schools should be visited, which inquiries should receive the most, or least, attention, and which new territories should be investigated. the fit index ensures that the admissions staff is directing its recruiting efforts to students well-suited for the institution. the dynamic hot index provides a continually updated measure of the probability of enrollment as the admissions cycle progresses. the arrival of sat scores, multiple contacts, a campus visit--these are clear indicators of a student's interest in the institution. coach or faculty contacts are also often promising. tracking these indicators allows the admissions staff to focus their efforts on the most interested students within the larger pool defined by the fit index.

analyzing the data: understanding the total index, fit Index, and hot index tables and charts

the data for each index are displayed on a spreadsheet. these spreadsheets have:

inquiries:

applicants:

matriculants:

the distribution of inquiries, applicants, and paid deposits, as a function of the index, is plotted, as is the probability of an inquiry's applying or matriculating. the cumulative probability of applying or matriculating is plotted as a function of the % of prospects.

components of a predictor

a predictor uses data from the recent past to anticipate the near future. the choices made in establishing the database are critical. if the model is constructed using only one year's data, there is the danger that a particular year was not typical. if the model is based on multiple years' data, there is the danger that conditions have altered significantly over the time period. a critical assumption for any predictive model is that the boundary conditions are unchanged: an improved economy might positively impact college attendance; a major campus crime might have negative impact. weighted averages and restricted time spans help avoid these dangers.

a single variable predictor is the simplest to calculate. the high schools attended by prospective students, for instance, represent a single variable; a certain number of inquiries, applicants, and matriculants are frequently associated with a high school year after year.

multiple variable models are richer and become increasingly predictive as parameters are added. students inquiring by letter typically apply in higher numbers than those filling out college fair reply cards. add sex, intended major, date of inquiry, test scores, high school rank, and we can refine our analysis of the likelihood of that student's applying and matriculating. an index is then a mathematical combination of a number of probabilities, based on past data that predicts the likelihood of an event.

as noted above, the fit index uses early data to identify inquiries with a relatively high probability of applying. although test scores and class rank are important predictors, they are typically not available early in the application cycle. the dynamic portion of the index has only transactional components triggered by student actions, and is, therefore, a powerful portion of the predictor. these actions, however, tend to occur late in the cycle. the total index is calculated by adding the later hot index to the earlier fit Index. this combined index provides a linear correlation between the student's index and the probability that a student will apply.

outcomes

the modeling algorithms developed by predictive edge have been applied to data from a number of different colleges. these institutions differed in location, mission, and size. in all three cases unique algorithms suited to the institution have been developed with excellent results, and the index is in use in the admissions process.

by way of example, college a's inquiry, applicant, and matriculant pools were analyzed. over 80,000 inquiries, 4,000 applicants, and 1,200 matriculants were scored. the table below shows a sample of the student by student output for the model. for each inquiry there is a score that translates into a probability that the student will apply. this table is provided electronically and will have the score for every inquiry or prospect in the pool.

as would be expected, as data is added, the total index ultimately becomes the better predictor of outcomes. this system allows the best available data to be used initially and an ever more powerful model (totalindex) to be used through the cycle.

data acquisition

predictive edge prefers to analyze the entire data file rather than specific identified fields. often the data collected by your institution will provide unique insights into the behavior of the students and will result in a superior model. but experience indicates that we need at minimum for our fit index: address, high school, sex, date of inquiry, origin of inquiry, major of interest, and/or college of interest. the hot index uses the record of all transactions between the student and the institution initiated by the student. this would include: campus visits, sending sat or act scores, returning a questionnaire, etc.

inquiry, applicant, accepted and matriculant data must be in a format compatible with microsoft access (text file, excel, lotus 123, dbase III, dbase IV, dbase 5, fox pro, fox pro 3, paradox, or odbc). the data should be sent electronically or on a zip disk or cd rom. the character used to delimit the data should be identified. the first entry in each column should be the field name (i.e. last name, first name, etc.). each student record must have a unique identifier--presumably their inquiry number or student number. do not send the student's name, social security number, or phone number. in this way the privacy of the individuals isn't compromised. zip codes should be provided in two fields (the first five digits in one field; the 'plus four' in a separate field). please provide the key for any local codes, majors, sports, activities, or counselors' names, for instance.

saving money with the predictor

there are many advantages to using smart recruitment predictor. it will help increase your applicant pool and increase the yield of applicants. how?

it will:

although listed last, smart recruitment predictor's capacity to identify inquiries with virtually no prospects of applying is one of its most powerful features. most colleges mail substantial packages to inquiries without concern for the likelihood that they will apply. the result is that they continue to commit the same error over and over again: expending funds equally between high probability prospects and low probability prospects. year after year the institution works with inquiries with the same profile that, year after year, don't apply. the waste is frequently $50,000 to $200,000 per year in publication and mailing costs alone, depending on the size of the inquiry pool.in addition, counselors and students calling low probability inquiries get discouraged, while at the same time high probability inquiries are neglected.

pricing for smart recruitment predictor

smart recruitment predictor includes:

for institutions using smart recruitment predictor for more than one year, a multi-year comparison of all relevant variables is carried out.

the price for smart recruitment predictor is based on the breadth and scope of the work. Contact us today to discuss your needs and receive a quote.

smart recruitment predictor will pay for itself many times over. because of the robust nature of the model, it normally retains its effectiveness for two or three years. for a modest fee, predictive edge will check the efficacy of the model each year, provide year-to-year comparisons, and inform you when a new algorithm needs to be developed.

unique advantages of smart recruitment predictor