How Huff Model Calibration By Survey Data works

Available with Business Analyst license.

For this method of calibration, you must make an appropriate division of your study area into small subareas. The main difference from the Huff Model Calibration By Real Data method is that you are calculating the probabilities here based on survey data showing the proportions of trips made from each subarea to the store areas. Therefore, you don't need to have a point customer layer, because the survey is operating directly on the subareas.

The survey data must be collected in a survey table using one of the following formats:

• Several lines per one interviewed customer

  List of customers' answers to the question, How often did you visit stores 1, 2, 3, or 4 during the last few weeks? The answer could be, Store 4—two times; store 3—four times. It should be tabulated as follows:

  Several lines per one interviewed customer

  Blockgroup ID	Store ID	Count
  1232234234	4	2
  1232234234	3	4
  1232234234	4	1
  1232234234	3	2

  The Blockgroup ID should be the ID of the block group (BG) in which the customer lives. In this example, the last two lines came from interviewing another customer living in the same BG.

• One line per geography unit

  It is a summarized version of the first variant. It can be calculated by summarizing the overall patronage count for a particular store from a particular block group, then calculating the percentage of people from this block group who visit each store. The same percentage can represent the percentage of money spent in this store.

  One line per geography unit

  Blockgroup ID	Store ID	Percentage
  1232234234	4	12
  1232234234	3	45

• One line per interviewed customer

  This variant may be more convenient for the people who do the survey.

  One line per interviewed customer

  Blockgroup ID	Store 1 Count	Store 2 Count	Store 3 Count	Store 4 Count
  1232234234	4	1	3	0
  1232234234	0	0	0	1

  For this variant, you will have a separate count (or dollar amount) field for every store. The table contains as many lines as the number of customers surveyed. Each customer indicates the number of times he or she visited each store. With this survey data, you can estimate the probabilities for each subarea that customers from that subarea will make trips to a specific store area.