Data integration and analysis
View Patent ↗Uncorrelated data from a variety of sources, such as consumer panels or retailer points of sale, are combined with maximal coverage of a universal data set for a population in a manner that permits useful inferences about behaviorial propensities for the population at an individual or household level.
1. A computer program product embodied in a non-transitory computer readable medium that, when executing on one or more computing devices, performs the steps of:
identifying a data set including:
characteristics of a population of consumers within a geographic area partitioned into sectors;
ancillary data from at least two sources describing, on a sector-by-sector basis, consumer behaviors within the geographic area, wherein one of the sources is a panel data source;
identifying, using a first data source, a first propensity of consumers in a first sector to engage in a specified behavior by computing a ratio of a number consumers in the first sector who engage in the specified behavior to a total number of consumers in the first sector;
identifying, using a second data source, a modeled second propensity of consumers in the first sector to engage in the specified behavior, wherein the data set does not include data sufficient to compute an actual second propensity corresponding to the modeled second propensity; and
using the first propensity and the modeled second propensity, identifying a third and fourth modeled propensities of consumers in a second sector to engage in the specified behavior with, respectively, the first and second data sources, wherein the data set does not include data sufficient to compute an actual third and forth propensities corresponding to the third and fourth modeled propensities.
2. The computer program product of claim 1 , wherein identifying the modeled second propensity includes setting the second propensity equal to the first propensity.
3. The computer program product of claim 1 , wherein identifying the modeled second propensity includes employing a shrinkage estimator.
4. The computer program product of claim 1 , wherein identifying the modeled second propensity includes using an elasticity model describing propensity variation across data sources.
5. The computer program product of claim 1 , wherein identifying the modeled second propensity includes using regression, imputation, projection, or similarity-based techniques.
6. The computer program product of claim 1 , wherein the specified behavior includes buying a particular brand of goods or services.
7. The computer program product of claim 1 , wherein the specified behavior includes consuming media from a particular media outlet.
8. The computer program product of claim 1 , wherein the specified behavior includes purchasing a specified collection of goods or services.
9. The computer program product of claim 1 , wherein the specified behavior includes purchasing at least a specified dollar amount worth of goods or services.
10. The computer program product of claim 1 , wherein the specified behavior includes shopping at a particular store.
11. The computer program product of claim 1 , wherein the specified behavior includes buying goods or services at a discount.
12. The computer program product of claim 1 , wherein identifying the third and fourth modeled propensities includes regression, imputation, projection, or similarity-based techniques.
13. The computer program product of claim 12 , wherein identifying the third and fourth modeled propensities includes verifying that a score for the third and fourth modeled propensities totals to a known value.