Test runs versus production runs

The test run is identical to a production run with the only difference being the number of customers processed.

In a production run, the entire set of customers is processed. However, for a test run, the number of customers who are processed would be approximately equal to that specified by the test run sample percentage. In the test run, customer chunks are created similar to those created in a production run. By using the value that is specified for test run sample percentage, the number of customers in the test run are calculated. These customers are then divided into various chunks by using the CustomerSampleSize parameter. This is similar to what is done in a production run. If the number of customers for the test run works out to be lesser than those required to completely fill up a chunk, a ceiling is done for the number of customers so it completely fills the chunk. The number of customers in a test run are always a multiple of the chunk size.

The query run part of the test run is similar to that of the production run. The queries are run on all the tables that are associated with Contact Optimization session (that is, PCT, RC, CH, and DCH). For each of these tables, the queries are run on its entire data set. The time that is taken for the run of the queries is similar in both types of runs. In a test run, running the session stops only after the Contact Optimization server processes the number of customers that equals the test sample size percentage.

The rules are also processed similarly to the production run. The CC rule constraints are distributed across total number of chunks of PCT similar to a production run. The CC rule constraints are not distributed only across the test run chunks but across all the chunks that are created from the PCT. In other words, the calculation of the CC rule constraints distribution across the chunks is done based on the total number of chunks across PCT. It is not done based on the number of chunks that are to be processed specifically for the test run; therefore, the application of the rules is identical to the production run.

It is recommended to keep the same settings for the Contact Optimization tuning parameters (that is, Algorithm Tuning parameters, Database Tuning parameters, and so on) for both types of runs to correctly extrapolate results of a test run against a production run. Due to the complexity of rules and data, the results of a test run do not have to be the same as a production run.

Note: All the queries run on whole data, so the initial requirements of preparing the session run is the same as production run. The results can differ between a production run and a test run in terms of the time that is taken and the overall quality of the results because of the complexities of the associated data.