A Consultant's Casebook
about me: Andres Inn, Ph.D.
A Dynamic Simulation Model of the Recruiting Process
I developed a computer simulation of the recruiting process to analyze and explain the apparent difficulties in achieving the goals of a changed recruiting mission. The model represented the basic process that recruiters are taught at Fort Benjamin Harrison. The model also reflects the field observations of our staff. Further, the recruiting model parallels the RAPS (Recruiter Applicant Processing System) in moving through the stages of recruiting: (1) prospecting for applicants, (2) selling the Army, and (3) contracting the applicant. Finally, the simulation model reflects the existing data that summarize recruiter productivity.
The computer simulation model illustrates the dynamics of the recruiting process and demonstrates that recruiters and their commanders have few leverage points with which to affect the system. Headquarters staff emphasized only one measurable variable: the number of recruits that are enlisted into the Army. General Wheeler and his staff had a much fuzzier conception of how many steps are necessary to transform high-school graduates to recruits, and how much time each step can take. In effect, Headquarters concentrated on managing the single outcome variable – recruits, and ignored the intervening variables that determined the number of recruits.
In many ways the system we describe is similar to popular illustration in Peter Senge's book, The Fifth Discipline. Consider a person in a shower who turns on the water and is concerned with only a single outcome variable - warm, even, and comfortable flow. He is unconcerned that his warm even flow of water depends upon a number of other factors, all of which are associated with some time delay. Senge described how the warm flow of water depends upon a gas line, a pilot light, a temperature sensitive control that turns on the gas under the hot water heater, a supply of cold water, the time that it takes for the water to heat, etc. The warm even flow of water can be adjusted only after all of these factors are operating. Even then, there will be some delay before the warm water makes its way through the pipes to the shower.
In this analogy, feedback plays an important part. In many situations there is a considerable delay between the time an adjustment is made, (say turn the water to hot) and the time in which we notice the effect (the water comes out hot). Unrecognized delays lead to instability and breakdown, especially when they are long. As Senge notes, adjusting the shower temperature is far more difficult when there is a ten-second delay before the water temperature adjusts, than when the delay takes only a second or two.
ten seconds after you turn up the heat, the water remains cold.
You receive no response to your action; so you perceive
that your act has had no effect. You
respond by continuing to turn up the heat.
When the hot water finally arrives, a 190-degree water gusher erupts from
the faucet. You jump out and turn
it back; and, after another delay, it’s frigid again.
On and on you go, through the balancing loop process. Each cycle of adjustments compensates somewhat for the cycle
The important factors in the recruiting model are also associated with considerable time delay. The model of recruiting presents the fundamentals of the dynamic simulation model we developed to study the recruiting process. As in the shower analogy, there is only one effect of consequence; the number of recruits. The rest of the model corresponds to the workings bethind the shower wall – the workings that are essential to assuring an even flow of qualified warm bodies under contract to the Army.
It is rather easy to walk through the model. The left side of the model represents the five states through which recruiters process prospective recruits. The right side of the model represents the processes with which prospective recruits are transformed to successive states. Leads, and Referrals flow into a pool of Leads. Recruiters prospect from this pool to identify interested Prospects. Prospects are chosen for a Sales Process that takes approximately one day. Prospects sold on military service become Applicantsl. Applicants are scheduled for the Military Enlistment Processing Station (MEPS) where applicants are qualified both physically and mentally for processing. The MEPS generally schedules no more than 2 applicants from each recruiter per day and is associated with a time delay. The delay can be as short as one day or as long as six months. Qualified Applicants who endure the enlistment processing can either drop out of the system, or begin training as new recruits. Of course, potential recruits can drop out of the process at any point during the entire process.
While the picture of a model may be important to concetpual understanding, it won’t necessarily convince anyone. It is important is to demonstrate that the collections of Leads, Prospects, Applicants, Qualified Applicants, and Recruits approximate the numbers found in real life. To do that, we try to find values for the pictured faucets that lead to consistent, representative results over hundreds of trials. Some of these values are estimated from data. Others are estimated from experience with the process. For example, experienced recruiters readily and accurately estimate how long enlistment processing takes, and how long it takes to develop a prospect into an applicant. And, some estimates are simply best guesses. Possible estimates are used to run a trial simulation for a given time period. When the model yields data inconsistent with experience, the estimated parameters are tweaked, and we try again until the model appears stable. Only after the model demonstrates consistently accurate results is it possible to experiment with the model and ask the questions that the General might have asked.
Walk-Ins, Referrals, Leads, and Prospects
The model begins with Leads, and Referrals. Recruiters prospect for Leads by visiting football and basketball games, and other community events. Referrals are persons referred to the recruiter by applicants, parents, school guidance counselors, etc. A faucet for which we developed a mini-model that described how and when it was opened represents each of these processes.
Walk-ins are persons who walk into a recruiting office to ask for information. Walk-ins are very promising prospects. They demonstrate a genuine interest in an Army career by visiting a recruiting office. A Poisson process with a mean of .012 people per day was used to generate the Walk-ins for the simulation model. These Walk-ins flow directly into the Prospect Pool and are thereafter considered as prospects.
Referrals are also considered as prospects. For each applicant, we assumed that the recruiter could generate a corresponding referral. As each applicant is processed, a new referral flows into the system as yet another prospect.
Leads, on the other hand, were not considered good prospects. Leads are generated in many ways; including class lists obtained from schools, the "bingo" cards in magazines, and more casual referrals. Generally, each recruiter develops such leads, and the majority of these leads are unlikely to join the Army. School lists are considered the most likely source of casual leads. Our model assumed these become available twice during the year; 300 in the Fall with the beginning of the school year (day 20 in our model) and another 300 in the Spring when the next cohort of Seniors (day 160 in our model) graduates. Leads thus generated are dumped into the reservoir of Prospects and are available to the recruiter for prospecting. We assume that each recruiter telephones approximately 15 prospects each day. Only a small portion of these prospects will be interested in enlistment, and hence another Poisson process is used to introduce them to the Sales Process that follows.
All of the Prospects are available to the recruiter for making Sales Contacts. The only constraint in making Sales Contacts is that our model assumes that a recruiter can make only one sales presentation at a time and to only one prospect at a time. Further, we assume that the Sales Time is exponentially distributed with a mean of 2 days.
Prospects that have gone through the Sales Process successfully become applicants. The model assumes that 15% of applicants drop out at this stage and do not take part in the enlistment process at the Military Enlistment Processing Station, MEPS. The remainder participates in the Enlistment Proces. Again, the Enlistment Process is associated with a delay. The delay is due to a number of factors among which is the assumption that a MEPS cannot take more than two applicants per day from each recruiter. Enlistment Processing is again assumed to be exponentially distributed taking an average of 10 working days. Some applicants drop out of the process during Enlistment Processing. The successful candidates become Qualified Applicants. After MEPS Processing, the Qualified Applicants. From observation, we estimated that the QNA rate is 40% and by extension, slightly over half of the MEPS applicants sign up as new Army Recruits.
RAPS is a computer expert system that we built for the United States Army Recruiting Command under a study contract with the Planning, Analysis, and Evaluation Section.
Peter M. Senge, The Fifth Discipline, Doubleday Currency, New York, 1990.
Peter M. Senge, The Fifth Discipline, Doubleday Currency, New York, 1990, p 91.
Major General Jack C. Wheeler, of the Army Recruiting Command, illustrates this point with his comment, "It takes six months to turn the force." In recruiting, as in the shower model, there is an evident time delay between the command and the perceived effects of the command.
The model was constructed with the Macintosh software product, iThinkª version 2.2.1 by High Performance Systems, Inc., 45 Lyme Road, Hanover NH 03755. Tel: (603) 643-9636, Fax: (603) 643-9502.
The Poisson process used to introduce Prospects to the Sales Process has a mean equal to the proportion of good leads times the number of leads prospected at that moment in time. Usually, the recruiter prospects about 15 leads per day, but when he is low on Leads, this figure will drop, and the Poisson process will deliver fewer Prospects into the Sales Process.
The exponential distribution is frequently used to model the time necessary to perform some service. Only one parameter lambda (l), I s necessary to estimate the probability density function: ¦(x)=le-lx, where x is the moment in time, and ¦(x) is the probability that the service will be completed at that time.
All material is copyrighted! Not to be used without the written permission of Andres Inn. ©2000.
Send mail to Andres.Inn@mail.ee with questions or comments about this web site.