Method and apparatus for generating performance curves

• US 20070247464A1
• Filed: 04/19/2006
• Published: 10/25/2007
• Est. Priority Date: 04/19/2006
• Status: Active Grant

• Current Assignee: Deloitte LLP (Deloitte & Touche)
• Original Assignee: Kneebone Inc.
• Inventors: Borys, Adrian, Malygin, Andrey, Wylie, John
• Granted Patent: US 7,908,165 B2
• Time in Patent Office: Days
• US Class Current: 345/442
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities
• CPC Class Codes: G06T 11/20 Drawing from basic elements, e.g. lines or circles

Abstract

A system and a method are set forth for generating performance curves for enabling user customized marketing communications proficiency and performance optimization, comprising collecting and storing historical vehicle data and response data relating for a marketing plan, normalizing the data to a standardized schema, performing non-linear polynomial regression analysis on the normalized data for generating a plurality of curves representing performance of the marketing plan, and displaying said plurality of curves to the user. Preferably, the plurality of curves are generated by a predictive analytics and forecasting engine using an optimized Chebyshev polynomial curve selected to a desired closeness of fit to the normalized data.

Claims

1. A method of generating performance curves for enabling user customized marketing communications proficiency and performance optimization, comprising:

collecting and storing historical vehicle data and response data relating to a marketing plan;

normalizing said data to a standardized schema;

performing non-linear polynomial regression analysis on the normalized data for generating a plurality of curves representing performance of each said vehicle; and

displaying said plurality of curves to said user.

2. The method of claim 1, further comprising creating a Vehicle Cast category for each distinct type of vehicle instance, creating a Default Classification category for each class of vehicle, and generating said plurality of curves for each said category resulting in Vehicle Cast curves and respective Default Classification curves.

3. The method of claim 2, wherein a first one of said plurality of curves comprises a decay curve representing the response to an instance in said marketing plan of one of either a vehicle in said Vehicle Cast category or class in Default Classification category, respectively.

4. The method of claim 3, wherein a second one of said plurality of curves comprises a weekly curve superimposed on said decay curve, said weekly curve representing a set of normalized ratios indicating performance of said one of either a vehicle in said Vehicle Cast category or class in Default Classification category, respectively, for each day of the week.

5. The method of claim 4, wherein a third one of said plurality of curves comprises an annual curve that modulates said decay curve and said weekly curve according to seasonal parameters indicating performance of said one of either a vehicle in said Vehicle Cast category or class in Default Classification category, respectively, when used at different times of the year.

6. The method of claim 3, further comprising comparing said plurality of Vehicle Cast curves to respective Default Classification curves and in the event respective ones of said curves differ from associated ones of said default curves by more than a specified amount then replacing said respective ones of said curves by said associated ones of said default curves.

7. The method of claim 6, wherein at least one of said Default Classification curves comprises a Master Default curve based upon one of either historical data for all classifications of said vehicles or representing industry-specific data.

8. The method of claim 5, wherein said annual curve is a Trusted Annual Curve with a deviation from its associated Default Classification curve capable of being overridden.

9. The method of claim 5, further comprising generation and display of a forecast curve based on said decay, weekly and annual curves.

10. The method of claim 1, wherein said non-linear polynomial regression analysis uses Chebyshev polynomials as basis functions for polynomial fitting of said plurality of curves to said data.

11. A system for generating performance curves for a user customized marketing communications proficiency and performance optimization, comprising:

a database for storing historical vehicle data and response data relating to a marketing plan;

a predictive analytics and forecasting engine for retrieving and normalizing said data to a standardized schema and performing non-linear polynomial regression analysis on the normalized data for generating a plurality of curves representing performance of said marketing plan; and

at least one server computer and one user computer in communication with said predictive analytics'"'"' and forecasting engine for user managing of said plurality of curves and displaying said plurality of curves to said user.

12. The method of claim 11, wherein said at least one user computer includes client software that operates on production and in-memory client datasets to perform one or more functions selected from the group comprising:

create and manage said marketing plan, model and perform forecasts based on said plurality of curves, create and forecast scenario plans, manage vehicles, advertisements, sources and modeling parameters, and provide integrated reports.

13. The system of claim 12, wherein said at least one user computer includes local data storage for saving one or more of scenarios, forecasts, and data presentation export files.

14. The system of claim 11, wherein said predictive analytics and forecasting engine and said at least one user computer communicate over a local area network and share two drive mapped file shares.

15. The system of claim 11, wherein said database further includes data warehousing processes for performing one or more of custom data imports, scheduled administrative functions and data mart processes.

16. The system of claim 11, further including a date generator for automating generation of dates conforming to a predetermined fiscal year for simplified display of said plurality of curves based on fiscal year, quarterly and weekly descriptors.

17. The system of claim 11, wherein said predictive analytics and forecasting engine uses Chebyshev polynomials as basis functions for polynomial fitting of said plurality of curves to said data.

18. A method of generating performance curves for enabling user customized marketing communications proficiency and performance optimization, comprising:

collecting and storing historical vehicle data and response data relating to a marketing plan;

normalizing said data to a standardized schema;

performing non-linear polynomial regression analysis on the normalized data for generating a plurality of curves representing performance of each said vehicle;

generating a forecast curve based on said plurality of curves;

building a profile illustrating lift effect over time of one said vehicle on another said vehicle; and

displaying said forecast curve and said profile to said user.

19. The method of claim 18, wherein building said profile comprises calculating said lift as Lift=b+s*exp [−

((t−

o)/2w)²], where b=Base Response of said another vehicle, w=Width (causal duration), o=Offset from Ran Date of said another vehicle, s=Scale that defines variation of lift about said Base Response, and t=Time from Ran Date.

20. The method of claim 18, wherein said profile is superimposed on said forecast curve for said one vehicle.

21. A method of generating performance curves for enabling user customized marketing communications proficiency and performance optimization, comprising:

collecting and storing historical vehicle data and response data relating to a marketing plan;

normalizing said data to a standardized schema;

performing non-linear polynomial regression analysis on the normalized data for generating a plurality of curves representing performance of each said vehicle;

generating a forecast curve based on said plurality of curves;

generating a geographical saturation model for scaling said forecast curve according to saturation ceiling for each of a plurality of geographical regions for said marketing plan; and

displaying said forecast curve as scaled by said geographical saturation model to said user.

22. The method of claim 21, wherein generating said geographical saturation model comprises aggregating individual forecasting results of said forecast curve into geography/week groupings, calculating a saturation ratio and scaling each individual forecast result in accordance with said saturation ratio.

23. The method of claim 22, wherein said saturation ratio is calculated as Y/X, where X=Sum of Forecast Leads for a specified region for a specified time duration and Y=said scaled individual forecast result based upon said saturation.

24. The method of claim 23, wherein said saturation ration Y/X=(A/x)*Tanh (x/A).