A-Score Computation

The overall A-Score on a customer is composed by combining the normalized scores of all the composite scales of the overall A-Score scale. There are actually two scores for a scale: the raw score and the normalized score.

Raw Score

Initially, raw scores are generated by summing up the points from any Customer Actions that qualify for scoring weights and applying scoring decay, if any.

Normalized Score

Next, all the raw scores within a scale are compared and normalized scores are generated based on those raw scores. The normalized scores are generated based on percentile ranks. The normalized score will always be between 1 and 100, while raw scores can be any decimal number. Whoever has the highest raw score, whatever that number may be, will wind up with a 100 normalized score, and so on. If your normalized score is a 50, then that means your raw score was higher than half of all the other raw scores.

Percentile rank

For scales with the "percentile rank" computation method, the computation of the normalized score is done in the stored procedure co_ascore_history_set_normalized_score.

This SP can be called in one of two ways explained below.

All Customers

When this SP is called for all customers, the SP will recompute the normalized scores for all customers (for a given scale and month date range) using a specialized percentile rank methodology. The logic is complex. Refer to the SP code for the logic.

Generally, netFORUM will run this operation only in maintenance operations that are run on a daily interval (at most) as it is not necessary for these scores to be "trued up" continuously.

Single Customer

When computed for a single customer, the logic of the SP is simpler than it is when the normalized scores are calculated for all customers. The SP will, in general, find the closest existing normalized score of other customers, based on the particular customer's raw score for the scale and month. This operation will not recompute all the other scores out there based on the change in this particular customer. Therefore, it should be noted that until the normalized scores are all trued up (reset), that over the course of each day, the normalized scores will veer ever so slightly from what they theoretically should be if those scores were all continuously trued up all day long. netFORUM does not try to do this all day long because it would add unnecessary overhead to the application and would not make an appreciable difference in scores, since in any given day only a small percentage of customers will experience scoring events.

Primary Scale

When scores from composite scales are combined into the A-Score scale, the normalized scores (not raw scores) are used from the composite scales.

First, netFORUM will look at all of a customer's scales that are a Composite Scale of the A-Score Scale.

Next, the normalized scores of each of those scales are multiplied by the percentage weight of that composite scale. Taking a look at the nearby image, we see that the Shopping scale has a 25% percentage. So, if a customer had a normalized score of 65 for Shopping, the calculation would take 25% of that score, which is 16.25. This will be done for each scale.

The percentage-weighted numbers for each scale are then summed up and that score will be the customer's raw A-Score.

Last, normalized scores for the A-Score will be computed based on those raw scores, using ranked percentiles. For example, if a customer's raw A-Score is 76, and that number is in the 83rd percentile of all customers, then that person's A-Score will be 83.

The reason that the normalized scores of composite scales are used to compute the A-Score instead of raw scores is because we don't want any single scale with very large raw scores to skew or dominate the overall A-Score.

Case Study

Consider the following scenario for a single customer who has the scores in several scales:

This customer has points in four different Scales. The Raw scores in each Scale are shown in Column B. These raw scores evaluate to the Normalized scores in Column C. For example, a 42 raw score for Events, when you look at all the other customers who have scores in Events, gets you to the 33rd percentile overall. Column D shows the percentage that the composite scale has been weighed for the overall A-Score. In this example, the Membership scale is over-weighted relative to Events and Community, while the Shopping scale is under-weighted.

Next, let's look at the Calculation in Column E. As shown below, the column is computed by a formula:

For each Scale, the calculation takes the Composite Percentage in Column D, divides it by 100, then multiplies it by the Normalized Score (Column C).

Finally, the overall Raw Score for the A-Score is calculated by summing up the calculations in Column E.

For this particular customer, that raw score is 19.05.

Finally, the raw score is evaluated across the range of all other raw A-Scores, and netFORUM computes the percentile of that raw score. So, if 19.05 were to fall in the 55th percentile of all other raw A-Scores, then that customer's A-Score would be 55.