Evaluation and Optimization of a Selection Process for Insurance Sales Agents
Client: a German insurance company operating internationally
Starting point: The client wanted to install a new selection process for field sales agents. The new process was going to be composed of three parts, each comprising a number of evaluative instruments:
Our task was to evaluate the new process and, if necessary, improve it.
In order to evaluate the predictive quality of the new selection process, it was necessary to measure subsequent performance both for candidates who passed the new process and for candidates who did not perform well or failed the process. Otherwise, variance would have been too limited and we would not have been able to use variance-based analytic methods such as regression analyses.
This meant that participants for the study had to be chosen from among the sales agents already working at the insurance company. In order to avoid possible negative consequences for participants, all data was gathered and evaluated anonymously. 80 sales agents were chosen to go through the new selection process and receive a score of their aptitude. Subsequently, we evaluated the sales performance of each candidate during a period of 18 months, using the number of sold non-life insurance policies and the value of sold life insurance policies as quantitative measures of success.
In order to evaluate the predictive quality of the selection process, participants’ aptitude scores were correlated to their performance data by means of a multiple regression analysis. Additionally, a factor analysis was performed in order to assess the internal consistency of the new selection process.
In the first stage of the factor analysis, the three parts of the new selection process were evaluated separately in order to determine whether their instruments (e.g., the interview questions) showed correlations among themselves, that is to say, whether they showed a tendency to produce similar values for any given candidate independently of the characteristics they were supposed to measure.
In the second stage of the factor analysis, we tested for correlations between the three parts of the selection process in order to determine whether they showed a tendency to produce similar values for any given characteristic and candidate.
The regression analysis showed no correlation between a good score in the selection process and future success on the job. Among the three parts of the process, only the intelligence test showed some predictive power, although correlations were only in a medium range.
The only characteristic that showed a significant predictive value for future success was previous knowledge of the insurance sector at the time of hiring.
The factor analysis revealed significant correlations between instruments when looking at each of the three parts separately, but no correlations across the three parts of the process.
These results were the opposite of what would be expected of a valid selection process (namely, low correlations between instruments but high correlations between parts). This was taken as evidence that the selection process was dominating over the characteristics to be measured, or in other words, that the process was measuring artefacts.
An anonymous survey was carried out among the 80 participants from Step 1 as well as their line managers and heads of agency. The personality traits were measured by means of a self-assessment by the 80 participants as well as assessments by their superiors.
The quantitative results of the survey were correlated with the performance data gathered in Step 1 by means of a multiple regression analysis.
A number of factors were determined to have a significant or even strong correlation with success as a sales agent:
If the selection process is modified to include a quantitative assessment of the individual predictors of success identified in Step 2, then what is the predictive quality of the modified selection process regarding the future success (sales performance) of candidates?
First, the new selection process was modified in the light of the insights gained from Step 2. The role play exercises were discarded and instead the interview was amended with questions designed to measure the personality traits that had been identified as predictors of success. Subsequently, the 80 participants went through the modified selection process and received new aptitude scores.
The results of the modified selection process were correlated with the existing performance data from Step 1 by means of a regression analysis.
The modified selection process showed good predictive power for success as a sales agent. On average, the sales performance of agents who passed the modified selection process was significantly higher than that of agents who did not pass the modified process.
Within the subgroup of agents who had no prior knowledge of the insurance industry at the time of hiring, the difference in sales performance was strongest, with “apt” sales agents outperforming “unapt” agents by almost 100%.
Client: a major German insurance company
Starting point: The client was faced with high unwanted attrition of sales agents within the exclusive sales organization.
We were hired to determine the causes of attrition and suggest measures to reduce it.
A survey was conducted among 35 dropped-out and 34 remaining sales agents. For an additional perspective, the former line managers of the dropped-out agents were questioned as well.
The other most frequently given reasons were: conflicts with superiors, lack of reaction on the part of superiors to voiced dissatisfaction, dissatisfaction with earnings, and a better offer from a competing company.
Regarding their former superiors, dropped-out sales agents criticized a lack of sales support, insufficient encouragement to develop professionally, low reliability in keeping promises, as well as a lack of human attention, appreciation, trust and sympathy.
Furthermore, superiors confirmed that in many cases the organization had responded inadequately to the expressed dissatisfaction of sales agents who then ended up leaving the company. Structural factors were given as the principal reason for this inadequate response. On the one hand, high administrative workloads and manager-to-staff ratios stand in the way of intensive personal supervision of sales agents, with the consequence that superiors often do not find out about an agent’s intention of leaving the company until it is too late.
On the other hand, retaining good sales agents requires more flexible structures such as shorter chains of command and individually adaptable remuneration and employment models.