|
|  |
Given the difficulty of explicitly canvassing employees’ intentions in this highly sensitive area (also because of the problem of socially desirable answering behaviour), ZebraZone has developed a prediction model that reflects the extent to which an organization is at risk of losing employees. We refer to this model as “retention analysis”.
The prediction model uses a retention predictor developed from the results of ZebraZone Benchmarks. In these Benchmarks we very explicitly ask the question of retention (“Are you planning to change employer?”). Here the problem of socially desirable answers features much less given the company-independent context in which the questionnaire is answered. This survey is also totally anonymous.
The prediction itself takes place in two phases:
- In the first phase it is important to know which variables play a role in predicting retention behaviour, and to what extent. ZebraZone uses for this a model based on discrimination analysis. This analysis technique allows us to divide up various individuals into one of two or more pre-defined categories. ZebraZone uses two categories, category 1 “planning to leave their company in the short term” and category 2 “not planning to leave the company in the short term”. In discrimination analysis we then go looking for those variables that significantly influence employees’ retention behaviour. People scoring high here will, we assume, tend not to quit their company in the short term, whilst low scorers may well be planning to quit in the short term. Once the variables have been selected, the appropriate coefficients are sought for each variable that indicate the extent to which these variables contribute to the observed retention behaviour. Based on these coefficients a number is calculated for each individual expressing the likelihood that he or she belongs to this category. Based on this number the individual is classified in one of the two categories.
- In a second phase these coefficients are used in a new discrimination analysis model in order to predict retention behaviour based on the variables that have been found.
For reasons of anonymity and confidentiality this prediction is possible only with groups of 30 respondents and upwards. This means that this predictor can be calculated at both total company and segment level, providing that this condition is taken into account. In addition the analysis can be used at segment level only if just one filter is ticked. It will not be possible to cross-segment it, here again in order to preclude the tracing of individuals.
But bear in mind that no prediction is ever exact. We must always factor a certain margin of error into this prediction. This is why ZebraZone speaks of an indication of the number of employees planning to leave the company in the short term. This indication is always given in the form of a bar diagram (see also Reporting).
In this particular bar diagram, the retention percentage of the organisation is compared with the retention percentage of the benchmark. In the diagram given here we see for example that, for the “Maxishop” company, the estimated percentage of people intending to leave the company in the short term (“yes” scorers) is equal to the actual (non-estimated) percentage of the Benchmark dataset.