When a decision offer is executed, the rules configured therein are used to determine which of its test offers is the winning offer.
It is important to note that it is not the raw number of records matching a given fulfillment state that determines the winning test offer. Rather, the number of records in a given state as a proportion of the records targeted by the test offer is used to decide which is the winner. This allows you to target test offers at data sets of differing sizes and use their relative efficacy as the medium through which to determine the winner.
Having determined the winning test offer, a decision offer then fulfills that offer to its own input data set (which must differ from the data sets at which its test offers were targeted).
If it was not possible to determine a winning test offer (e.g. in the event of a tie), the default test offer, as defined at the decision offer’s configuration panel, is fulfilled by the decision offer. If a default offer was not specified, a test offer is picked at random.