Ex Parte Dao et al

Board of Patent Appeals and Interferences

Jul 10, 2012

11975533 (B.P.A.I. Jul. 10, 2012)

UNITED STATES PATENT AND TRADEMARK OFFICE
UNITED STATES DEPARTMENT OF COMMERCE
United States Patent and Trademark Office
Address: COMMISSIONER FOR PATENTS
P.O. Box 1450
Alexandria, Virginia 22313-1450
www.uspto.gov
APPLICATION NO. FILING DATE FIRST NAMED INVENTOR ATTORNEY DOCKET NO. CONFIRMATION NO.
11/975,533 10/19/2007 Fu-Tak Dao 94621 8618
22242 7590 07/10/2012
FITCH EVEN TABIN & FLANNERY, LLP
120 SOUTH LASALLE STREET
SUITE 1600
CHICAGO, IL 60603-3406
EXAMINER
HAIDER, FAWAAD
ART UNIT PAPER NUMBER
3627
MAIL DATE DELIVERY MODE
07/10/2012 PAPER
Please find below and/or attached an Office communication concerning this application or proceeding.
The time period for reply, if any, is set in the attached communication.
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE 1
___________ 2
3
BEFORE THE BOARD OF PATENT APPEALS 4
AND INTERFERENCES 5
___________ 6
7
Ex parte FU-TAK DAO, 8
RICARDO MARTIJA, 9
THOMAS SPACEK, 10
and SAMARADASA WEERAHANDI 11
___________ 12
13
Appeal 2011-002371 14
Application 11/975,533 15
Technology Center 3600 16
___________ 17
18
19
Before ANTON W. FETTING, JOSEPH A. FISCHETTI, and 20
MEREDITH C. PETRAVICK, Administrative Patent Judges. 21
FETTING, Administrative Patent Judge. 22
DECISION ON APPEAL 23
Appeal 2011-002371
Application 11/975,533

2
STATEMENT OF THE CASE1 1

1 Our decision will make reference to the Appellants’ Appeal Brief (“App.
Br.,” filed June 23, 2010) and Reply Brief (“Reply Br.,” filed October 21,
2010), and the Examiner’s Answer (“Ans.,” mailed September 15, 2010).
Fu-Tak Dao, Ricardo Martija, Thomas Spacek, and, Samaradasa 2
Weerahandi (Appellants) seek review under 35 U.S.C. § 134 (2002) of a 3
final rejection of claims 14-30, the only claims pending in the application on 4
appeal. We have jurisdiction over the appeal pursuant to 35 U.S.C. § 6(b) 5
(2002). 6
Although Appellants state at Appeal Brief 4, 7
[t]here are no related appeals or interferences known to 8
appellant, the appellant's legal representative, or assignee that 9
will directly affect, or be directly affected by or have a bearing 10
on the Board's decision in the pending appeal[,] 11
we find that Appeal 2011-002013 is for parent Application 10/139,503 from 12
this application as parent, and both appeals had their Notices of Appeals 13
filed on April 14, 2010, their Appeal Briefs filed on June 23, 2010, and their 14
Reply Briefs filed on October 21, 2010. Both applications were examined 15
by the same Examiner and rejected over the same art in Examiner’s Answers 16
entered in the record on October 13, 2010. Thus, we treat these appeals as 17
being related. 18
The Appellants invented a way of analytically determining the revenue 19
of certain types of Internet companies using Web based and equipment 20
based metrics related to Internet companies for analytically determining the 21
Appeal 2011-002371
Application 11/975,533

3
current revenue and statistically projecting the future revenue of these 1
companies (Specification 1:12-15). 2
An understanding of the invention can be derived from a reading of 3
exemplary claim 14, which is reproduced below [bracketed matter and some 4
paragraphing added]. 5
14. A method for estimating current revenue of a company, the 6
method comprising: 7
[1] a computing device 8
obtaining 9
from network resources 10
data points associated with the company 11
for each of a plurality of Internet metrics, 12
other than sales data, 13
over a plurality of past time periods; 14
[2] the computing device generating a plurality of revenue 15
models 16
for the company 17
by using the data points 18
for each of the plurality of Internet metrics; 19
[3] the computing device choosing 20
from the plurality of revenue models 21
a revenue model 22
for estimating current revenue; 23
[4] the computing device obtaining 24
from the network resources 25
current data points 26
associated with the company 27
for each of the plurality of Internet metrics; 28
Appeal 2011-002371
Application 11/975,533

4
and 1
[5] the computing device estimating 2
the company's current revenue 3
by applying the current data points 4
to the chosen revenue model. 5
The Examiner relies upon the following prior art: 6
Singh US 2002-0169657 A1 Nov. 14, 2002
Rotman US 2003/0018550 A1 Jan. 23, 2003
Claims 14-30 stand rejected under 35 U.S.C. § 103(a) as unpatentable 7
over Rotman and Singh. 8
ISSUES 9
The issue of obviousness turns primarily on whether the references 10
may be combined, which in turn depends on whether Singh’s forecasting 11
techniques are applicable to Rotman’s revenue forecast. 12
FACTS PERTINENT TO THE ISSUES 13
The following enumerated Findings of Fact (FF) are believed to be 14
supported by a preponderance of the evidence. 15
Facts Related to the Prior Art 16
Rotman 17
01. Rotman is directed to provide near real-time market information 18
predictions based on money flow maps derived from payment 19
transaction information. Payment system operators, such as credit 20
card issuers, have transactional data that is representative at a 21
Appeal 2011-002371
Application 11/975,533

5
statistically significant level of general market trends of individual 1
companies and broader econometric data, and payment system 2
operators have access to transactional data on a real-time basis. 3
The data can be manipulated to best fit corporate revenue trends, 4
giving equity analysts more information on how a company is 5
doing. More specifically, the embodiment can process the 6
transactional data to produce revenue predictions for a company 7
over the next month or quarter. Rotman ¶ 0024. 8
02. Rotman uses transaction data from credit cardholders to project 9
corporate revenues for companies traded on any of the major 10
exchanges (e.g., NYSE, NASDAQ, AMEX, etc.). Rotman ¶ 11
0053. 12
03. For each company of interest, the dollar value of all 13
transactions can be accumulated over a predetermined period of 14
time, for example from the first day of current business quarter 15
until the current day of business quarter. Rotman ¶ 0054. 16
04. The accumulated transaction data is then processed by an issuer 17
system. For example, the issuer system may normalize the data. 18
After normalizing the data, the issuer system scales the data. The 19
normalized data may be scaled using a number of different 20
techniques such as using linear regression or pattern recognition 21
via a neural network. Rotman ¶ 0055. 22
05. Depending on the particular parameter, different scaling 23
methods may be required to obtain a good prediction. For 24
example, to predict corporate revenue trends, some information 25
Appeal 2011-002371
Application 11/975,533

6
about a company's accounting practices may be necessary. An 1
airline may, for example, obtain payment for a July flight in 2
January, when a traveler plans her trip. However, the airline may 3
not actually book the revenue until the ticket is used. Therefore, 4
there may be some delay between the time when a payment 5
transaction occurs and when the transaction is reflected in 6
revenue. Rotman ¶ 0056. 7
06. Finally, the issuer system applies the data to provide financial 8
information to end users. This may involve either providing the 9
scaled and normalized sales information in its raw form or 10
applying the scaled and normalized information to known or 11
newly created models for predicting financial metrics, such as 12
stock price, interest rates, or commodity supplies. One example is 13
the application of sales information to predict a company stock 14
performance. In another example, near real-time predictions may 15
be made about stored supplies of a commodity (such as gasoline). 16
Periodically, government agencies report information about 17
aggregate stored supply levels of particular commodities. Thus, 18
the last published report of commodity storage levels may be used 19
in conjunction with normalized and scaled information about 20
recent sales of the commodity to make near real-time predictions 21
of the actual storage levels, before the next report becomes 22
available. Rotman ¶ 0057. 23
07. FIG. 3E is an exemplary flow chart of a process for applying 24
normalized data to calculate predicted performance as a revenue 25
model. First, a period for which to predict revenue for a company 26
Appeal 2011-002371
Application 11/975,533

7
of interest is selected and the period is labeled "t". Next, a 1
corresponding past period with known revenue is selected and 2
labeled "t-1". Next, normalized transaction data is calculated, 3
labeling the normalized data TRXt and TRXt-1. Next, the gross 4
growth rate is calculated by dividing TRXt by TRX t-1. Once a 5
gross growth rate has been determined, the actual revenue for 6
period t-1 is obtained and multiplied by the gross growth rate to 7
provide a revenue prediction. Rotman ¶ 0070. 8
08. A scale factor is useful for scaling normalized transaction data 9
(N) to produce a prediction of monthly or quarterly corporate 10
revenue. For example, to scale the data, a historical ratio can be 11
used that compares the normalized data to the actual reported 12
revenue of a company. Rotman ¶ 0071. 13
09. For example, to scale the data, the value N is calculated which 14
corresponds to the result of normalizing the data. Next, the 15
number of past periods X that will be considered is determined. 16
Then, the ratio of normalized results to actual revenue for each of 17
the past X periods is calculated. Thereafter, the ratios for the past 18
X periods are averaged to determine the average ratio of 19
normalized transaction amounts to actual corporate revenue. 20
Finally, N is divided by the average ratio to produce a revenue 21
prediction. Rotman ¶ 0072. 22
10. As an example using a regression analysis to predict actual 23
revenue, a number X of past periods to consider in a regression 24
analysis is determined. Next, the normalized data corresponding to 25
Appeal 2011-002371
Application 11/975,533

8
the X periods is calculated. Then the actual revenue corresponding 1
to the relevant past periods is obtained. Next, based on the 2
normalized data and the actual revenue figures, a regression 3
equation is formulated using, for example, a "least squares" 4
method. Next, a normalized transaction figure is calculated 5
corresponding to a period for which revenue is to be predicted. 6
Finally, predicted revenue is calculated based on the formulated 7
regression equation. Rotman ¶ 0073. 8
11. FIG. 4C is an exemplary flow chart of a process for scaling data 9
using a neural network. First, the value N is calculated which 10
corresponds to the result of normalizing the data. Next, the 11
number of past periods X to be considered is determined and data 12
from the past X periods is provided as input to a known neural 13
network capable of matching patterns corresponding to metrics, 14
such as, for example, accurate revenue prediction. Thereafter, a 15
best fit model is derived from the neural network. A best-fit model 16
may be selected on the basis of the model's ability to accurately 17
predict past performance. Finally, the best fit model is applied to 18
normalization data N to calculate a prediction. Rotman ¶ 0074. 19
12. FIG. 5D is an exemplary flow chart of a process for applying 20
data using industry trends. Tracking indices for specialized 21
industries can be created using the predicted revenue data. 22
Examples of industries include retail sales, transportation, airlines, 23
mail order, Internet, etc. Tracking indices could be faxed or 24
emailed to "licensed customers" or they could log onto the issuer's 25
system to examine the indices. Industry selections are received 26
Appeal 2011-002371
Application 11/975,533

9
from a licensed end user. Based on the selections made by the end 1
user, tracking indices are created for each of the selected 2
industries and are computed and sent to licensed end users for the 3
selected industries. Rotman ¶ 0086. 4
Singh 5
13. Singh is directed to accurately forecasting future demand for 6
many products and product types in many markets. Singh enables 7
organizations to produce and compare alternative models of 8
forecasted demand in order to constantly improve demand-9
forecasting capabilities, producing and identifying optimal 10
demand forecasts that take into account independent causal 11
factors, such as new competitive products and price promotions, 12
that will impact upon future demand, so users can easily account 13
for current demand trends without having to produce an entirely 14
new forecast. Singh ¶’s 0014-0018. 15
14. In the manufacturing and distributing industries, supplying 16
products in response to the current level of customer demand with 17
a minimum of overstocking reduces stocking costs and 18
distribution expenses and thus leads to a reduction of the sales unit 19
price of products. Typically, this also leads to an enhancement of 20
profit margins. It is therefore necessary for sellers to forecast 21
product demand precisely such that they then can determine a 22
sales plan, production plan, and distribution plan according to an 23
accurate forecast of the demand trend of customers. Singh ¶ 0004. 24
Appeal 2011-002371
Application 11/975,533

10
15. Conventional methods of forecasting demand by analyzing the 1
trend of past sales results are performed with the goal of the 2
forecaster being to apply the most accurate statistical analysis 3
techniques and econometric modeling to provide the most 4
accurate forecast possible. In these conventional methods, time 5
series forecasting is performed which develops and uses various 6
forecasting algorithms that attempt to describe the knowledge of 7
the business and fluctuation trend of sales results as evidenced by 8
past history in the form of a rule. Singh ¶ 0005. 9
16. Unfortunately it is common for product demand trends to 10
change in a short lifecycle. When this is the case, the data used in 11
providing a forecast rapidly becomes old and the precision of 12
forecasting lowers. Therefore, in order to keep a high precision in 13
forecasting, algorithms (and historic data points used to generate 14
those algorithms) must be maintained on an ongoing basis as well 15
as be able to adjust their forecasts relatively easily. Singh ¶ 0006. 16
17. Singh provides a multiple model framework that allows 17
multiple alternative forecasting algorithms, including well known 18
statistical algorithms such as Fourier and multiple linear 19
regression ("MLR") algorithms and proprietary algorithms useful 20
for modeling the causal factors relating to various products and/or 21
industries, to be associated with various data streams of demand 22
history to product advanced forecasting models. The present 23
invention helps users to determine the forecast algorithm that best 24
suits a given problem/industry/business from the available history 25
streams of demand data by enabling the production of various 26
Appeal 2011-002371
Application 11/975,533

11
forecasts for comparison with one another and, eventually, with 1
incoming data relating to actual demand. Singh ¶ 0020. 2
18. To generate a particular forecast, Singh calculates a variety of 3
forecasts according to alternative models and then selects the 4
results of a particular forecast model that one feels is most 5
accurate and would like to submit (or "publish") for subsequent 6
supply planning. Algorithms as utilized in the multiple model 7
framework (MMF) are composed of a series of mathematical 8
calculations that are executed based upon history stream data in 9
order to create a statistical forecast. Preferred embodiments enable 10
users to achieve a balance between strictly statistical forecasts and 11
judgment forecasts by controlling the use of history data. 12
Statistical forecasts assume that products generate stable demand. 13
As such, statistical forecasts are typically more useful to reflect 14
the characteristics of mature products that are produced and 15
maintain long life cycles with many customers. These forecasts 16
rely upon a base history (i.e., history not containing such 17
anomalous or unusual data points) for predictable and repeatable 18
demand. Singh ¶ 0050. 19
19. Demand history data often includes data points (such as spikes 20
or valleys in demand) that can cause problems if they are included 21
within calculations to predict future demand. Examples of such 22
problematic data include variation in demand caused by unusual 23
market conditions, decreases in demand due to obsolete or 24
superseded products, general data errors, and the inability of the 25
user to distinguish between base and non-base data. In order to 26
Appeal 2011-002371
Application 11/975,533

12
account for such problematic data in base or non-base history, the 1
user is provided the opportunity to adjust history to overcome 2
such problems. Mechanisms by which a user can adjust history 3
data include manually editing extremely high or low values, 4
controlling the time period of the relative history used to ignore 5
periods of sporadic and/or otherwise unreliable history, and 6
masking specific periods of unreliable and/or sporadic history. 7
Singh ¶ 0081. 8
ANALYSIS 9
Independent claims 14, 21, and 28 are method, computer media, and 10
system variants of forecasting revenue by applying some past and present 11
Internet metric data to a revenue model. We initially note that although 12
limitation [1] obtains data other than sales data, this does not mean that 13
limitations [2] and [5] cannot use sales data. Instead, the calculations in 14
limitations [2] and [5] somehow use the data from limitation [1], which can 15
be in addition to sales data. Thus, the claims do not require forecasting sales 16
data without relying on past sales data. 17
The Examiner found that Sing generally describes forecasting revenue 18
by applying some demand data to a revenue model in those instances where 19
past revenue streams alone might prove insufficient for accurate forecasting. 20
This left the choice of using some Internet metric data as for such demand 21
data as the sole issue. In resolving this, the Examiner found that Singh 22
described Internet commerce demand data among other industry options. 23
Before even taking up the Examiner’s findings regarding Singh, we 24
initially find that the only calculations claimed are described by Rotman. 25
Appeal 2011-002371
Application 11/975,533

13
The nature of the particular factors used to normalize or scale are perceptible 1
only to the human mind. 2
In a recent non-precedential decision, our reviewing court reminded us 3
of the applicability of the precedential In re Gulack, 703 F.2d 1381 (Fed. 4
Cir. 1983), In re Bernhart, 417 F.2d 1395 (CCPA 1969) and In re Lowry, 5
32 F.3d 1579 (Fed. Cir. 1994) decisions. 6
We have held that patent applicants cannot rely on printed 7
matter to distinguish a claim unless “there exists [a] new and 8
unobvious functional relationship between the printed matter 9
and the substrate.” In re Gulack, 703 F.2d 1381, 1386 (Fed. 10
Cir. 1983). [] [T]he Board did not create a new “mental 11
distinctions” rule in denying patentable weight []. On the 12
contrary, the Board simply expressed the above-described 13
functional relationship standard in an alternative formulation—14
consistent with our precedents—when it concluded that any 15
given position label’s function [] is a distinction “discernable 16
only to the human mind.” []; see In re Lowry, 32 F.3d 1579, 17
1583 (Fed. Cir. 1994) (describing printed matter as “useful and 18
intelligible only to the human mind”) (quoting In re Bernhart, 19
417 F.2d 1395, 1399 (CCPA 1969)). 20
In re Xiao, 2011 WL 4821929 (Fed Cir 2011)(Non-precedential). Thus 21
non-functional descriptive material, being useful and intelligible only to the 22
human mind, is given no patentable weight. See also In re Ngai, 367 F.3d 23
1336, 1339 (Fed. Cir. 2004). That is, we give weight to using some factor to 24
adjust prior revenue for modeling a forecast, but not to the particular factor 25
actually used, which is no more than arbitrary binary data as actually 26
employed. 27
The Examiner found that even giving such weight, however, the 28
independent claims were at least predictable. The Examiner found that 29
Singh adjusted revenue by essentially normalizing with respect to demand 30
Appeal 2011-002371
Application 11/975,533

14
changes. In any distribution model which relied on e-commerce, such 1
demand statistics would naturally be an Internet metric, being a metric of 2
demand communicated over the Internet. As Rotman describes the bulk 3
of the independent claims and Singh shows the predictability of using an 4
Internet metric for the one limitation not in Rotman, the independent claims 5
are predictable and therefore obvious. 6
We are not persuaded by the Appellants’ argument that Rotman teaches 7
away from using revenue data. Appeal Br. 11-12. Appellants cite Rotman ¶ 8
0018, which describes background art including providing merchant 9
rankings instead of sales data as having limited utility. Appellants take this 10
to be discouraging the use of non-sales data. First, Rotman does admit that 11
such rankings have some utility. Second, rankings are but one form of 12
metric and that paragraph does not discourage using other non-sales metrics. 13
Indeed Rotman describes using transaction data generally, and not just sales 14
data. More to the point, the claims also use sales data – they just also use 15
something else as well – which both Rotman and Singh do too. 16
We are not persuaded by the Appellants’ argument that Rotman fails to 17
disclose Internet metrics other than sales data. Appeal Br. 12. Rotman’s 18
Internet tracking indices, although themselves being sales data, at least 19
shows that using Internet commerce data was predictable. Singh shows 20
that data other than sales data, such as demand data would be applied to 21
Rotman’s model. Thus, the combination shows it was predictable to apply 22
Internet demand data to Rotman’s model. 23
We are not persuaded by the Appellants’ argument that Singh’s history 24
streams do not contain Internet metric data. Appeal Br. 14. Singh’s history 25
Appeal 2011-002371
Application 11/975,533

15
data would represent the nature of the transaction history. One of the 1
exemplary contexts for such history is in Internet commerce. FF12. 2
We are not persuaded by the Appellants’ argument that one cannot 3
combine the references. Appeal Br. 14-16. Both references describe 4
forecasting techniques. Such forecasting techniques are mathematical in 5
nature, and thus, the practicality of combining techniques is not bounded 6
by the specific examples described by the references. Singh describes 7
adjustments that might be applied to Rotman’s revenue forecasts using 8
demand data to improve forecasts. As revenue is clearly dependent on 9
demand, the applicability of Singh to Rotman is all too apparent. Whether 10
one relies on customer specific data and the other on anonymous data as 11
argued does not diminish the teaching that one of ordinary skill would take 12
from these references. 13
CONCLUSIONS OF LAW 14
The rejection of claims 14-30 under 35 U.S.C. § 103(a) as unpatentable 15
over Rotman and Singh is proper. 16
DECISION 17
The rejection of claims 14-30 is affirmed. 18
No time period for taking any subsequent action in connection with this 19
appeal may be extended under 37 C.F.R. § 1.136(a). See 37 C.F.R. 20
§ 1.136(a)(1)(iv) (2011). 21
22
Appeal 2011-002371
Application 11/975,533

16
AFFIRMED 1
2
3
JRG 4