U.S. patent application number 13/906981 was filed with the patent office on 2014-09-11 for price scoring for vehicles.
The applicant listed for this patent is CarGurus, LLC. Invention is credited to Oliver I. Chrzan, Kyle Lomeli, Edward L. Steinert.
Application Number | 20140258044 13/906981 |
Document ID | / |
Family ID | 51488993 |
Filed Date | 2014-09-11 |
United States Patent
Application |
20140258044 |
Kind Code |
A1 |
Chrzan; Oliver I. ; et
al. |
September 11, 2014 |
PRICE SCORING FOR VEHICLES
Abstract
Vehicle pricing such as used vehicle pricing is improved by
supplementing statistical modeling techniques with additional
algorithms to accommodate factors such as geography and dealer
reputation that do not readily yield to regression analysis or
similar tools that might be used to characterize a population.
Inventors: |
Chrzan; Oliver I.;
(Somerville, MA) ; Lomeli; Kyle; (Belmont, MA)
; Steinert; Edward L.; (Concord, MA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
CarGurus, LLC |
Cambridge |
MA |
US |
|
|
Family ID: |
51488993 |
Appl. No.: |
13/906981 |
Filed: |
May 31, 2013 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
61776202 |
Mar 11, 2013 |
|
|
|
Current U.S.
Class: |
705/26.64 |
Current CPC
Class: |
G06Q 30/0205 20130101;
G06Q 30/0629 20130101; G06Q 30/0206 20130101 |
Class at
Publication: |
705/26.64 |
International
Class: |
G06Q 30/06 20060101
G06Q030/06 |
Claims
1. A method comprising: receiving a request for vehicle information
from a client; ranking a number of vehicles responsive to the
request based upon a relative value using a difference between a
fair market price and a listing price for each of the number of
vehicles, wherein the relative value is a dimensionless value
normalized according to a standard deviation of prices for the
number of vehicles, thereby providing a ranked list; adjusting a
position of one of the vehicles in the ranked list according to a
dealer reputation for a dealer offering the one of the vehicles for
resale, thereby providing an adjusted ranked list; and transmitting
one or more items in the adjusted ranked list to a client for
display.
2. The method of claim 1 wherein the fair market price for each of
the number of vehicles is determined using a regression model.
3. The method of claim 2 wherein the regression model uses a number
of regression parameters including one or more of a vehicle type, a
vehicle condition, a vehicle condition, and a vehicle trim.
4. The method of claim 2 wherein the regression model uses a number
of regression parameters including one or more of a vehicle fleet
history, a repair history, and a flood damage history.
5. The method of claim 2 further comprising retrieving vehicle
listings from a plurality of online sources and creating the
regression model using the vehicle listings.
6. The method of claim 2 further comprising retrieving vehicle data
for each one of the number of vehicles from one or more online
sources.
7. The method of claim 1 further comprising adjusting a position of
each one of the vehicles in the ranked list according to a
corresponding dealer reputation, thereby providing the adjusted
ranked list.
8. The method of claim 1 further comprising assigning a deal
quality score to a portion of the adjusted ranked list and
transmitting the deal quality score for one or more listings within
the adjusted ranked list to the client.
9. The method of claim 1 further comprising transmitting a number
of surveys to a number of purchasers of vehicles and processing
responses to the number of surveys to determine the dealer
reputation for the dealer.
10. The method of claim 1 wherein the request for vehicle
information specifies at least one of a type, a trim, a year, and a
mileage.
11. A computer program product comprising computer executable code
embodied in a non-transitory computer-readable medium that, when
executing on one or more computing devices, performs the steps of:
receiving a request for vehicle information from a client; ranking
a number of vehicles responsive to the request based upon a
relative value using a difference between a fair market price and a
listing price for each of the number of vehicles, wherein the
relative value is a dimensionless value normalized according to a
standard deviation of prices for the number of vehicles, thereby
providing a ranked list; adjusting a position of one of the
vehicles in the ranked list according to a dealer reputation for a
dealer offering the one of the vehicles for resale, thereby
providing an adjusted ranked list; and transmitting one or more
items in the adjusted ranked list to a client for display.
12. The computer program product of claim 11 wherein the fair
market price for each of the number of vehicles is determined using
a regression model.
13. The computer program product of claim 12 wherein the regression
model uses a number of regression parameters including one or more
of a vehicle type, a vehicle condition, a vehicle condition, and a
vehicle trim.
14. The computer program product of claim 12 wherein the regression
model uses a number of regression parameters including one or more
of a vehicle fleet history, a repair history, and a flood damage
history.
15. The computer program product of claim 12 further comprising
retrieving vehicle listings from a plurality of online sources and
creating the regression model using the vehicle listings.
16. The computer program product of claim 12 further comprising
retrieving vehicle data for each one of the number of vehicles from
one or more online sources.
17. The computer program product of claim 11 further comprising
code that performs the step of adjusting a position of each one of
the vehicles in the ranked list according to a corresponding dealer
reputation, thereby providing the adjusted ranked list.
18. A system comprising: a database storing a regression model that
characterizes a fair market value of a vehicle according to a
number of regression parameters; a server configured to receive a
request from a client for vehicle information and to transmit to
the client an adjusted ranked list responsive to the request; and a
processor configured to rank a number of vehicles responsive to the
request based upon a relative value using a difference between a
fair market price for each of the number of vehicles determined
using the regression model and a listing price for each of the
number of vehicles, wherein the relative value is a dimensionless
value normalized according to a standard deviation of prices for
the number of vehicles, thereby providing a ranked list, the
processor further configure to adjust a position of one of the
vehicles in the ranked list according to a dealer reputation for a
dealer offering the one of the vehicles for resale, thereby
providing the adjusted ranked list.
19. The system of claim 18 further comprising a dealer evaluation
module executable to transmit a survey to a purchaser of a vehicle
and to process a survey response to determine the dealer reputation
for the dealer.
20. The system of claim 18 wherein the database stores a plurality
of regression models for different vehicles, and wherein the
processor is configured to select a best one of the plurality of
regression models for a type of vehicle specified in the
request.
21-40. (canceled)
Description
RELATED APPLICATIONS
[0001] This application claims the benefit of U.S. App. No.
61/776,202 filed on Mar. 11, 2013, the entire content of which is
hereby incorporated by reference.
BACKGROUND
[0002] While statistical techniques such as regression permit
characterization of a sample of data such as a population of used
cars for sale, such models do not readily accommodate certain
relevant factors. For example, fair market value for a vehicle may
depend on geography. However, the use of geography to restrict a
data set for price estimation may reduce available data (e.g., cars
list for sale) so much that reliable statistical inferences about
fair market value become difficult or impossible. Similarly,
factors such as the reputation of a dealer who is offering a
listing may be highly relevant to a purchaser when evaluating the
desirability of a particular listing, but may not yield a
quantitative price adjustment that can be used for comparison to
other, similar vehicles.
[0003] There remains a need for improved scoring models to assist
consumers when comparing listings of used vehicles.
SUMMARY
[0004] Vehicle pricing such as used vehicle pricing is improved by
supplementing statistical modeling techniques with additional
algorithms to accommodate factors such as geography and dealer
reputation that do not readily yield to regression analysis or
similar tools that might be used to characterize a population.
BRIEF DESCRIPTION OF THE FIGURES
[0005] The invention and the following detailed description of
certain embodiments thereof may be understood by reference to the
following figures:
[0006] FIG. 1 shows entities participating in a scoring system.
[0007] FIG. 2 is a flow chart of a method for ranking vehicle
listings.
[0008] FIG. 3 shows a web page that contains ranked vehicle
listings.
[0009] FIG. 4 is a flow chart of a method for making geographic
adjustments to vehicle prices.
DETAILED DESCRIPTION
[0010] All documents mentioned herein are hereby incorporated in
their entirety by reference. References to items in the singular
should be understood to include items in the plural, and vice
versa, unless explicitly stated or otherwise clear from the
context. Grammatical conjunctions are intended to express any and
all disjunctive and conjunctive combinations of conjoined clauses,
sentences, words, and the like, unless otherwise stated or clear
from the context. Thus the term "or" should generally be understood
to mean "and/or" and so forth.
[0011] The following description emphasizes pricing and scoring
techniques for used automobiles. However, it should be understood
that the methods and systems described herein may be applied to
other vehicles such as motorcycles, sport utility vehicles, light
trucks, trucks, and the like, and that the methods and systems may
also or instead be readily adapted to new vehicle pricing where
factors such as geography and dealer reputation may be relevant to
a purchasing decision. More generally the methods and systems
disclosed herein may be usefully employed in any context where
price comparisons are made between similar items offered for
sale.
[0012] It will also be noted that terms such as price, score and
price scoring are frequently used in the following description.
Such terms are intended to encompass calculations of price, such as
a fair market price for a vehicle, as well as representations of
price or value that are relative in nature, such as a price divided
by a standard deviation of a regression model for determining
price. In general, such various representations of value, or value
relative to fair value, may be interchangeably used or calculated
with simple, linear scaling or other straightforward adjustments
and/or transformations. Any such representation of price or value
may be used for ranking or other comparison of items offered for
sale, and terms such as "price," "score," and "price score" are
intended to include all such variations unless a more specific
meaning is explicitly provided or otherwise clear from the
context.
[0013] FIG. 1 shows entities participating in a scoring system. The
system 100 may include a data network 102 such as the Internet that
interconnects any number of clients 104, data sources 106, and a
server 108 (which may include a database 110). In general, the
server 108 may secure data from the various data sources 106 such
as dealer listings and other third party data sources, and
construct a price model for determining a fair price for vehicles.
This price model can then be deployed to determine relative value
for vehicles offered for sale, such as by comparing each listing
price to a fair market price determined using the price model. In
this manner, the server 108 can respond to inquiries from clients
104 with ranked lists of vehicles offered for sale, where the list
is ranked according to a relative value for each listing. Elements
of the system 100 are described in greater detail below.
[0014] The data network 102 may include any network or combination
of networks suitable for interconnecting other entities as
contemplated herein. This may, for example, include the Public
Switched Telephone Network, global data networks such as the
Internet and World Wide Web, cellular networks that support data
communications (such as 3G, 4G and LTE networks), local area
networks, corporate or metropolitan area networks, wide area
wireless networks and so forth, as well as any combination of the
foregoing and any other networks suitable for data communications
between the clients 104, the data sources 106 and the server
108.
[0015] The clients 104 may include any device(s) operable by end
users to interact with the server 108 through the data network 102.
This may, for example, include a desktop computer, a laptop
computer, a tablet, a cellular phone, a smart phone, and any other
device or combination of devices similarly offering a processor and
communications interface collectively operable as a client device
within the data network 102. In general, a client 104 may interact
with the server 108 and locally render a user interface such as a
web page or the like supporting interaction by the end user with
services provided by the server 108.
[0016] The data sources 106 may include any sources of data useful
for pricing/scoring as contemplated herein. In one aspect, this may
include dealer listings, which may be provided as a data feed,
database, or the like available through the data network 102 using
a suitable programming interface. In another aspect, dealer
listings may be obtained from a website using scraping, bots, or
other automated techniques. Dealer listings may include information
useful for price modeling or relevant to determination of a fair
price for a particular vehicle including, without limitation, a
vehicle type (e.g., make or model), a vehicle mileage, a vehicle
year (of manufacture), a vehicle trim (e.g., option packages,
features, etc.), a vehicle transmission, a vehicle condition, a
vehicle interior/exterior color, a vehicle history (accident/repair
history, rental fleet status, etc.) and so forth. Dealer listings
may include other information useful to consumers for decision
making but not directly quantitatively applicable to a model for
pricing. For example, a listing may include photographs of a
vehicle, or a narrative description of the automobile prepared by
the dealer. Such information may also be retrieved from the dealer
website for use when presenting aggregated listings from the server
108 to a user at a client 104.
[0017] In another aspect, data sources 106 may include third party
data providers. For example, a variety of commercial services are
available that provide vehicle history such as a repair history, a
fleet history (use in a rental fleet or commercial fleet of
vehicles), a flood damage history, and so forth. Where data such as
a vehicle identification number is available in dealer listings,
such data may be directly matched to various listings. Other
techniques can be used to correlate such third party data to
vehicle listings or otherwise infer vehicle condition or history.
Other data such as data provided by government agencies may, where
available, provide useful information relating to vehicle title,
vehicle inspection history, vehicle mileage, vehicle accident
history, and so forth.
[0018] The server 108 may in general be configured as described
above to create one or more price models using data obtained from
the data sources 106, and to respond to user inquiries from the
clients 104 with ranked lists and other data. In embodiments, the
server 108 may employ multilinear regression analysis to derive a
pricing model that relates vehicle price to various vehicle
attributes. The resulting model may take the general form:
y.sub.i=.beta..sub.1x.sub.i1+.beta..sub.2x.sub.i2+ . . .
+.beta..sub.px.sub.ip+.epsilon..sub.i [Eq. 1]
where x.sub.ij is the i.sup.th observation on the j.sup.th
independent variable (where the first independent variable takes
the value 1 for all i). A model may be created, for example, for
each vehicle type, and the regression parameters, {circumflex over
(.beta.)}, for each such model may be calculated for independent
variables such as the condition, the mileage, the year, and so
forth from the data sources 106. It will be readily appreciated
that, while the residual error may be minimized for any given data
set, the goodness of fit for a model and the statistical
significance of the estimated parameters may be subject to review,
and the model may be revised, e.g., by the addition or removal of
parameters or the removal of outlier observations, until an
adequate model is obtained. Such a process may be manual,
automated, or some combination of these, and may be informed by
subjective or objective characterizations of the quality of the
resulting model. Suitable objective criteria for various models
include a standard error, an R-squared analysis of residuals, an
F-test of overall fit, and a t-test for individual regression
parameters.
[0019] It will be understood that a variety of other statistical
techniques such as nonlinear regression, curve-fitting, and so
forth may be appropriate in various data modeling contexts. More
generally, a wide range of modeling techniques are known in the art
for predictive analysis including without limitation neural
networks, fuzzy logic models, case-based reasoning, rule-based
systems, regression trees, and so forth, any of which may be
employed to computationally derive suitable predictive algorithms
for fair market value. Furthermore, numerous computational
techniques are known for estimating parameters for a regression
model including without limitation percentage regression, least
absolute deviations, nonparametric regression, distance metric
learning, and so forth, any of which may be suitably employed for
various types of populations or data sets. Still more generally,
these techniques are provided by way of non-limiting examples, and
any such techniques or other techniques, as well as combinations of
the foregoing, may be usefully adapted to obtain predictive models
for vehicle price that can be implemented by the server 108. All
such variations are intended to fall within the scope of the term
"model" as used herein unless a different meaning is explicitly
provided or otherwise clear from the context.
[0020] However derived, a price model may be stored in the database
110 along with underlying data for vehicle listings. The server 108
may be configured to calculate fair market value according to the
price model, and to provide this information to clients 104, such
as in the form of a ranked list of vehicles for sale. The list may
be ranked according to a price score that provides a dimensionless,
numerical representation of relative value. In one embodiment, a
price score, S, for a vehicle may be calculated as:
S = P fm - P l .sigma. [ Eq . 2 ] ##EQU00001##
where P.sub.fm is the fair market value of the vehicle (as
calculated using the price model), P.sub.1 is the list price at
which the vehicle is offered for sale (according to the vehicle
listing), and .sigma. is the standard deviation for the price
model. A list of results ranked according to the price score may be
transmitted from the server 108 to one of the clients 104, along
with related data for each vehicle (photos, narrative description,
attributes, etc.) so that a user of the client 104 can browse
listings and compare vehicles listed for sale.
[0021] It will be understood that while a single server 108 is
depicted in FIG. 1, any number of logical servers or physical
servers may be used as the server 108 according to, e.g., server
traffic, desired level of service, and so forth. Similarly, server
functionality may be divided among different platforms in a number
of ways. For example, one server or group of servers may be used to
obtain data from the data sources 106 and create price models for
various vehicle types. Another server or group of servers may be
configured to provide a web interface for receiving and responding
to client requests for vehicle price information using the price
model(s) created by the first group of servers. Any such
configuration suitable for responding to clients 104 based upon
user-provided parameters and data obtained from the data sources
106 may be employed as the "server" described herein.
[0022] Having described a general system for providing ranked
listings of vehicles for sale in response to client requests, this
description now turns to modifications and adaptations to such a
system that address certain characteristics of vehicle listings
that are not amenable to direct price modeling.
[0023] FIG. 2 is a flow chart of a method for ranking vehicle
listings.
[0024] As shown in step 202, the method 200 may include creating a
price model. This may include using any of the data sources and
modeling techniques described above to create a predictive model
relating various vehicle attributes to a fair market price. By way
of non-limiting example, this may include creating a regression
model as described above so that a fair market price for vehicles
can be determined using the regression model. The regression model
may use any of a number of regression parameters such as a vehicle
condition, a vehicle trim, a vehicle fleet history, a repair
history and/or a flood damage history. A vehicle type may also be
used as a regression parameter, or a different regression model may
be constructed for each vehicle type, or some combination of these
according to, e.g., the variability in configurations of different
vehicles of a particular "type" or the quantity and quality of data
for a type. As described above, creating a price model may include
retrieving vehicle listings from a plurality of online sources and
creating the regression model using the vehicle listings. This may
also or instead include retrieving vehicle data for each vehicle
from a number of different data sources, such as third party data
sources that provide specific types of vehicle data.
[0025] As shown in step 204, the method may include obtaining
dealer reputation data. This may include a variety of data
gathering techniques which may be used alone or in combination with
one another. In one aspect, this may include transmitting a number
of surveys to a number of purchasers of vehicles and processing
responses to the surveys to determine the dealer reputation for the
corresponding dealers. Such data may be conveniently gathered for
purchasers who shop for and purchase vehicles using the server
described above through the use of automated electronic surveys or
the like, and such survey information may be gathered during an
online interaction related to the purchase, or in a subsequent
communication such as an electronic mail or the like sent to
purchasers after completing transactions that were initiated
through the server. In such a survey, a dealer may be evaluated
against one or more criteria using an objective scale (e.g., one to
five), and the results may be aggregated in any suitable manner for
each dealer. Although depicted sequentially in FIG. 2, there is no
particular reason for dealer data to be preferentially gathered
before or after creation of a pricing model. These two steps may
occur concurrently, sequentially or asynchronously. For example,
dealer reputation data may be accumulated over long periods of
time, and may remain relevant for extended periods. Thus this data
may be gathered and updated incrementally as new survey data
becomes available, or on some scheduled or other periodic basis
such as once per hour, once per day, once per week, or on any other
suitable schedule. By contrast, fair market value may be preferably
modeled as an instantaneous, current value, and the various price
models might be updated at the greatest possible or practical rate
to provide current data. Thus price models may be updated once per
hour, once per day, or on any other schedule according to, e.g.,
processing resources available to create new models and the rate of
change in data sources used to create the price models.
[0026] As shown in step 206, the method 200 may include receiving a
request for vehicle information from a client. This may, for
example, include a request posted to a web page from a client
device that includes a vehicle make, model, trim, mileage, year and
other attributes to narrow or define a search. Attributes may be
specified in a variety of ways such as with a range of possible
values (e.g., for mileage, year or list price) or as a filter to
include or exclude certain attributes such as a vehicles having a
certain trim, feature, option package or the like. Where a server
provides data for both new and used vehicles, these categories may
be modeled differently, and a web site or other interface for
configuring the user inquiry may request this information first.
More generally, techniques for gathering such information
interactively from a user of a client device are well known in the
art, and such techniques may be used in any suitable manner to
parameterize a user request for vehicle information.
[0027] As shown in step 208, the method 200 may include ranking a
number of vehicles responsive to the request (e.g., meeting the
various parameters of the request). The ranking may be based upon a
relative value using a difference between a fair market price and a
listing price for each of the number of vehicles. The relative
value may be a dimensionless value normalized according to a
standard deviation of prices for the number of vehicles, such as by
using the price score shown above in Eq. 2. In this manner, a
ranked list of the vehicles may be provided.
[0028] While a ranking based upon relative dollar value provides
useful information to a consumer who is considering various similar
vehicles, other information may be relevant to a purchasing
decision even though the other information does not directly affect
the fair market value of a vehicle. For example, dealer reputation
may be relevant to the desirability of a vehicle, or to the
expected purchasing experience for the vehicle, even where the
reputation of a seller does not directly translate to a change in
the fair market value of the vehicle. That is, one vehicle having
certain attributes may be more or less desirable than another
vehicle with the same attributes because of the differences in the
dealers offering each vehicle for sale, even though the vehicles
are objectively identical (and therefore of equal value). In order
to address such noneconomic factors, rankings may be adjusted to
account for additional information. Or stated slightly differently,
vehicles may be ranked using a scoring system that accounts for
such factors in addition to a price model that is based upon
objective vehicle attributes.
[0029] As shown in step 210, the method 200 may include adjusting a
position of one of the vehicles in the ranked list according to a
dealer reputation for a dealer offering the vehicle for resale,
thereby providing an adjusted ranked list. More generally, one,
some or all of the vehicles may receive an adjusted ranking
according to a dealer reputation for each corresponding listing. In
general, a scoring system may be devised to account for dealer
reputation that complements the scheme used for determining fair
market value, i.e., that permits reputation-based scoring to
supplement rather than substitute for scoring based on objective
vehicle attributes. In one aspect, a Bayesian average or similar
metric for dealer reputation may be calculated. A Bayesian average
advantageously provides a weighting for each dealer's reputation in
proportion to the amount of reputation data available. So for
example, where each dealer is evaluated on a scale of 1-5 by n
customers, this data may be combined in a weighted manner with
other dealer reputation data using an averaging formula of the
general form:
x _ = Cm + i = 1 n x i C + n [ Eq . 3 ] ##EQU00002##
where m is a population mean and C is an assigned value. The
constant, C, may for example represent the average number of
reputations scores for each dealer, or any other number according
to a desired weighting of the reputation of a particular dealer
toward the population mean.
[0030] As a result of such an adjustment, when vehicle results are
displayed to a client, certain listings that appear to offer a
better economic deal may under certain circumstances be ranked
lower than other listings of lesser or equal relative value. In
this manner, cars may be ranked according to the manner in which
dealer reputation influences desirability without changing
estimated fair value based on objective attributes. It should be
appreciated that this approach contemplates a separation of
reputation affects from attribute-based value, so that two
identical automobiles will have the same fair market price
regardless of the respective reputations of dealers offering such
vehicles for sale. In an alternative scoring system, the dealer
reputation may also be used to adjust perceived economic value. For
example, a scaling factor based upon the dealer reputation may be
used to adjust a calculated fair price for each vehicle. This may,
e.g., take a form such as:
.DELTA.P=k( x-m) [Eq. 4]
where m and x are the population mean and the individual dealer
mean respectively (see Eq. 3 above), and where k is an empirically
selected scaling factor. In one practical application, a scaling
factor of about 0.2 standard deviations (for the price model) has
been found to yield satisfactory results for a price adjustment,
.DELTA.P. Thus it will be appreciated that separately calculating
affects of dealer reputation does not prevent corresponding
adjustments to price, and the impact of reputation may be expressed
as a non-dollar denominated "score" that influences ranking, or the
impact may be directly incorporated into a price calculation after
attribute-based calculations are completed.
[0031] As shown in step 212, the method 200 may include assigning a
deal quality score to a portion of the adjusted ranked list and
transmitting the deal quality score for one or more listings within
the adjusted ranked list to the client. The deal quality score may
be a figure of merit calculated for each listing, or the deal
quality score may be a categorical score, which may be based on
percentiles or other ranking ranges, or some combination of these.
Thus, for example, the bottom twenty percent (or any other suitable
percentile range) may be given a deal quality score of "bad deal"
or "poor deal", and the top twenty percent (or any other suitable
percentile range) may be given a deal quality score of "great deal"
or "good deal". Groups of vehicles within various percentile ranges
may be given corresponding, intermediate rankings, which may be
determined with any suitable or convenient degree of granularity.
For a quantitative figure, the deal quality score may, for example,
be a numerical representation of relative value such as that
provided by Eq. 2, or any other suitable representative number. It
should also be noted that the relative value may be based on the
fair market value either before or after accounting for dealer
reputation as discussed above, and may be dimensionless, or may be
expressed in dollars or any other suitable units.
[0032] As shown in step 214, the method 200 may include
transmitting one or more items in the adjusted ranked list to a
client for display. This may include associated data such as a deal
quality score, a fair market value, a relative value, a numerical
ranking and any other calculated data for each listing, along with
metadata such as photographs, narrative description, and contact
information or a location where the vehicle is offered for sale
(and/or available for inspection).
[0033] It will be appreciated that the methods disclosed with
reference to FIG. 2 may be deployed in the system disclosed with
reference to FIG. 1 to provide a vehicle price evaluation system
that includes a database and a server configured to receive a
request from a client for vehicle information and to transmit to
the client an adjusted ranked list responsive to the request. The
database may, for example, store one or more regression models that
characterize a fair market value of a vehicle according to a number
of regression parameters. The database may store a plurality of
regression models for different vehicles along with individual
vehicle listings. The processor may be configured to select a best
one of the plurality of regression models for a type of vehicle
specified in the request, and to provide ranked lists and other
data to clients as generally discussed above.
[0034] For example, the server (or other location and/or computing
hardware) may include a processor configured to rank a number of
vehicles responsive to a request based upon a relative value using
a difference between a fair market price for each of the number of
vehicles determined using the regression model and a listing price
for each of the number of vehicles. The relative value may be a
dimensionless value normalized according to a standard deviation of
prices for the number of vehicles, thereby providing a ranked list.
The processor may be further configure to adjust a position of one
of the vehicles in the ranked list according to a dealer reputation
for a dealer offering the one of the vehicles for resale, thereby
providing the adjusted ranked list.
[0035] Such a system may include a dealer evaluation module, e.g.,
in software executing on the server or some other server, to
transmit a survey to a purchaser of a vehicle and to process a
survey response to determine the dealer reputation for the dealer.
In this manner, dealer reputation data may be gathered for improved
vehicle rankings as described above.
[0036] FIG. 3 shows a web page that contains ranked vehicle
listings. The web page 300 may be transmitted from a server such as
any of the servers described above to a client. The web page 300
may include a number of listings 302 ranked according to relative
value, adjusted for dealer reputation as described above.
[0037] Each listing 302 may include additional data such as a
dealer rating 304, a list price 306, a deal quality score 308 and
any other information characterizing a particular listing or
information about the listed vehicle. The deal quality score 308
may include various representations of deal quality such as text
(e.g., "Great Deal," "Fair Deal," etc.), a graphic (e.g., an up
arrow, down arrow or sideways arrow), a quantitative statement of
value (e.g. "$1,134 BELOW fair market value", "Top Ten!", "top ten
percent", etc.) or any other representation or combination of
representations of the quality of each listing.
[0038] The web page 300 may also include a variety of tools to
provide or revise search parameters including, for example, sliders
to specify ranges, drop down lists to select from among a number of
options, text boxes to enter search terms and check boxes to
specify use of various filters. More generally, any controls that
can be used to parameterize user input within a web page or other
interface may be used to gather user input specifying a vehicle
search.
[0039] FIG. 4 is a flow chart of a method for making geographic
adjustments to vehicle prices. In general, a price model developed
for a national market may provide accurate inferences concerning
the affect of various vehicle attributes on fair market value. At
the same time, metropolitan markets may exhibit a skew in average
vehicle prices due to local differences in tastes, incomes, vehicle
availability, and so forth. Using the following techniques, a
pricing model can realize accurate attribute-based pricing based
upon a large population of vehicles while reducing
nationally-influenced excursions from local pricing norms.
[0040] As shown in step 402, the method 400 may include creating a
pricing model for fair market value of a vehicle type based upon a
first data set of vehicles obtained from a first geographic region.
The pricing model may be any of the models described above. For
example, the pricing model may be a regression model based on a
plurality of regression parameters such as a model, a year, a
mileage, and a trim. Other regression parameters such as a rental
fleet history, a repair history and a flood damage history may also
or instead be used. The first geographic region may be a national
region, with the first pricing model being a corresponding national
model that includes vehicle data for the entire country, or the
first geographic region may be some other large geographic area
that consistently provides a large data set for price
estimation.
[0041] As shown in step 404, the method 400 may include identifying
a second data set, such as a local data set, of vehicles from a
second geographic region within the first geographic region. For
example, where the first geographic region is a national region,
the second geographic region may be a metropolitan region within
the national region. Any practical technique may be employed to
define the second geographic region and to select vehicles for
inclusion in the second data set. For example, the second data set
may be based on zip codes or other geographically identifiable
regions within a predetermined radius of a metropolitan center. The
second data set may also non-exclusively include vehicle data for a
number of adjacent metropolitan regions. That is, a listing on an
outlying perimeter of one metropolitan center may also be included
in similar data sets for adjacent metropolitan centers. Where a
large radius (e.g., seventy five miles) is used about each
metropolitan area and a large number of metropolitan regions (e.g.,
the one hundred largest metropolitan centers) are to be analyzed,
overlapping geographies will frequently occur. By non-exclusively
using perimeter data for each such overlapping metropolitan region,
differences in pricing bias between adjacent metropolitan areas can
be normalized somewhat to reduce significant pricing
discontinuities between geographically proximate listings
associated with different metropolitan centers.
[0042] A rural region may also be created for each state or other
recognizable geographic expanse to account for listings that are
not within the predetermined radius about any of the metropolitan
regions used for calculating median offsets from the national
model.
[0043] As shown in step 406, the method 400 may include calculating
a median offset between local and national data. For example, a
median offset may be calculated between a first median for the
first data set scored according to the pricing model (e.g.,
national scores) and a second median for the second data set scored
according to the pricing model (e.g., local scores). More
generally, any representation of a bias between local and national
pricing may be usefully employed as a median offset when scoring a
local data set using a national model.
[0044] As shown in step 408, the method 400 may include adjusting
the pricing model according to the median offset to provide an
adjusted pricing model for the second geographic region. For
example, price or score calculations for vehicles in a local or
regional market may be adjusted by using the difference between the
two medians as an offset to adjust scores calculated using a
national model. In effect, this uses all of the pricing information
available from national data to construct a pricing model, while
preserving aggregate biases peculiar to local markets.
[0045] As shown in step 410, the method 400 may include receiving a
request from a client for local listings. The request may specify a
geographic region such as a metropolitan region or city within the
first geographic region.
[0046] As shown in step 412, the method 400 may include scoring a
number of vehicles within the second geographic region with the
adjusted pricing model, thereby providing scored vehicles. Where
the second data set non-exclusively includes data from adjacent
metropolitan regions, the number of vehicles scored for
presentation to a client is preferably formed from an exclusive
subset of the listings in the second data set. Thus the scored
vehicles may be selected from a smaller region within the second
geographic region and closer to the corresponding metropolitan
center. This approach generally returns vehicle listings more
responsive to a client request that specifies a particular
metropolitan region. This also has the practical advantage of
ensuring that a particular vehicle is not listed twice (e.g., for
two adjacent metropolitan regions) with different estimated fair
market values.
[0047] As generally described above, the pricing model may
calculate an estimated fair market value, or the pricing model may
calculate a relative or absolute metric representative of deal
value. Thus, the numeric quantity calculated in the scoring process
may be a dollar amount, or a numeric score relative to a standard
deviation for the pricing model, or a score expressed or normalized
in any other manner suitable for ranking vehicle listings. It will
be appreciated that scoring of vehicles for a metropolitan region
may be performed dynamically (i.e., in response to a specific
client request), or scoring may be precomputed for use in response
to multiple client requests.
[0048] As shown in step 414, the method 400 may include
transmitting the scored vehicles to a client for display. This may
include transmitting a web page such as the web page describe
above, along with any vehicle information, deal quality evaluation,
or other data useful to a consumer in reviewing and comparing
vehicle listings. As noted above, the number of vehicles scored and
transmitted to clients may be a subset of the second data set used
to determine the median offset, and may more particularly be a
subset of the second data set exclusive to one of the number of
metropolitan regions. For example, where a request from a client
specifies listings within a city, the number of listings returned
may be a subset of the second data set including listings
exclusively associated with that city. Thus a first, non-exclusive
data set may be used when determining a median offset for a
metropolitan region, while a second, exclusive data set selected
from within the non-exclusive data set may be used when calculating
market value and returning listings to a client for the
metropolitan region.
[0049] A server may be configured to perform regional pricing and
respond to user requests for listings as generally described above.
Thus in one aspect there is disclosed herein a system including a
database storing a national regression model that characterizes a
fair market value of a vehicle according to a number of regression
parameters based upon a national market; a server configured to
receive a request from a client for vehicle listings in a
metropolitan market within the national market and to transmit to
the client an adjusted ranked list responsive to the request; and a
processor configured to select a number of vehicles responsive to
the request, to calculate a score for ranking the number of
vehicles using the national regression model, and to adjust the
score for each of the number of vehicles according to a difference
between a first median score for the number of vehicles within the
metropolitan market using the national regression model and a
second median score for vehicle listings within the national
market, thereby providing the adjusted ranked list.
[0050] The processor may be a processor within the server, and the
database may be any suitable memory device. The score returned to a
client for a particular vehicle may be a price, or the score may be
a relative value based upon a difference between a fair market
price and a listing price for one of the number of vehicles
relative to a standard deviation for the metropolitan regression
model.
[0051] A number of variations are possible for the above pricing
techniques. In one aspect, dealer reputation and regional
adjustments may be combined to provide scoring that reflects local
market conditions and the reputations of dealers offering certain
vehicles for sale. In another aspect, further refinements may be
made to regional price calculations. For example, certain vehicle
attributes may be subject to different local pricing, such as where
a particular feature is valuable in one region but not in another.
Where a particular feature of a vehicle is uncorrelated or
negatively correlated to price between two different regions, the
corresponding vehicle attribute may be excluded from a national
model and/or independently priced within each geographic region.
Thus in one aspect the methods disclosed herein may include
identifying a vehicle attribute that has a geography-dependent
influence on price and modeling that vehicle attribute
independently from a regression model used for other vehicle
attributes.
[0052] The methods or processes described above, and steps thereof,
may be realized in hardware, software, or any combination of these
suitable for a particular application. The hardware may include a
general-purpose computer and/or dedicated computing device. The
processes may be realized in one or more microprocessors,
microcontrollers, embedded microcontrollers, programmable digital
signal processors, or other programmable device, along with
internal and/or external memory. The processes may also, or
instead, be embodied in an application specific integrated circuit,
a programmable gate array, programmable array logic, or any other
device or combination of devices that may be configured to process
electronic signals. It will further be appreciated that one or more
of the processes may be realized as computer executable code
created using a structured programming language such as C, an
object oriented programming language such as C++, or any other
high-level or low-level programming language (including assembly
languages, hardware description languages, and database programming
languages and technologies) that may be stored, compiled or
interpreted to run on one of the above devices, as well as
heterogeneous combinations of processors, processor architectures,
or combinations of different hardware and software.
[0053] Thus, in one aspect, each method described above and
combinations thereof may be embodied in computer executable code
that, when executing on one or more computing devices, performs the
steps thereof. In another aspect, the methods may be embodied in
systems that perform the steps thereof, and may be distributed
across devices in a number of ways, or all of the functionality may
be integrated into a dedicated, standalone device or other
hardware. In another aspect, means for performing the steps
associated with the processes described above may include any of
the hardware and/or software described above. All such permutations
and combinations are intended to fall within the scope of the
present disclosure.
[0054] It should further be appreciated that the methods above are
provided by way of example. Absent an explicit indication to the
contrary, the disclosed steps may be modified, supplemented,
omitted, and/or re-ordered without departing from the scope of this
disclosure.
[0055] The method steps of the invention(s) described herein are
intended to include any suitable method of causing such method
steps to be performed, consistent with the patentability of the
following claims, unless a different meaning is expressly provided
or otherwise clear from the context. So for example performing the
step of X includes any suitable method for causing another party
such as a remote user, a remote processing resource (e.g., a server
or cloud computer) or a machine to perform the step of X.
Similarly, performing steps X, Y and Z may include any method of
directing or controlling any combination of such other individuals
or resources to perform steps X, Y and Z to obtain the benefit of
such steps.
[0056] While particular embodiments of the present invention have
been shown and described, it will be apparent to those skilled in
the art that various changes and modifications in form and details
may be made therein without departing from the spirit and scope of
this disclosure and are intended to form a part of the invention as
defined by the following claims, which are to be interpreted in the
broadest sense allowable by law.
* * * * *