AHP (Analytic Hierarchy Process) Scripts

Unlike Decision Tree and Rule-Based scripts--which search for a single, best solution--AHP scripts provide your Web site visitors with a range of possible solutions in ranking order. By asking questions and obtaining preferences, DecisionScript can determine which of your products best meets your customers' needs. The products are then listed in ranking order, allowing the customers to make the final choice. This script is particularly useful for companies that offer a wide range of similar products, such as computers, stereo equipment, appliances, etc. It can also be used to rank places to live, vacation destinations, etc. This type of script is easy to maintain and update once it is in place.

The exact method behind Analytic Hierarchy Process is beyond the scope of this document. However, AHP can be somewhat accurately described as a fancy form of weighted average ranking. By asking the user a series of pair-wise comparisons between each of the ranking criteria, an AHP script will calculate the relative importance the user places on each criteria. This information is combined with an expert's rating of each product or alternative in each of the categories and a weighted average score is computed.

To illustrate the steps used to build an AHP script, Jupiter Cellular will once again serve as the company. However, for this script customers will be trying to determine which of the company’s cellular phones is the best model to buy. Customers rank the importance of the features: price, size, weight, and battery life. This information is combined with the performance of each phone in the feature categories to find the best phone for the particular customer's needs.

The data used to create the script is provided below. Open the file named AHP found in the DecisionScript Examples directory to see a completed script.

Creating an AHP script involves two basic steps: (1) creating a data table that lists the products and characteristics that will be compared, and (2) building the tree that will run the comparison by asking your Web site visitor a series of questions. If you look at the completed example, you will note that the script contains two sheets--Tree1 and Table. Tree1 is the actual script. Table is a sheet that contains the information presented in the chart above.

A key difference between the chart and the table is that the data in the table has been altered so that comparisons can be made across categories. For example, Standby Time has been converted from hours to minutes so that a value for Battery Life (based on Talk Time and Standby Time) can be determined. In addition, Size is a product of the length multiplied by the width multiplied by the height.

In order for the script to run correctly, the rating given for each product in a particular category must be based on a linear scale and include zero as the base. This allows the script to make the assumption, for example, that a product costing $7.00 is twice as desirable as a product costing $14.00.

Another key element of the table is the top three rows in the first column--Objective, Report, and Direction--which define how the data in each column is used. Objective, which is defined as either a 0 or 1, tells the script that data in a column should not be weighted (0) or should be weighted (1). For example, the images that are loaded when the script runs do not need to be weighted. However, Price, Size, etc., should be weighted.

Report, which is defined as either a 0 or 1, tells the script that the data in a column should not be shown to the Web site visitor (0) or should be shown (1). For example, the actual Dimensions of each product should be presented, but the value used to determine Size should not.

Finally, Direction, which is defined as either a 1 or -1, tells the script that the data in a column should be ranked so that increasing values are better (1) or increasing values are worse (-1). For example, the higher the Price of an item, the worse its perceived value. Conversely, the longer the Battery Life, the better its perceived value.

To build the Tree portion of the script, follow the steps outlined on the following pages.

See Also

Building an AHP Script