Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
DETAILED ACTION
Claims 1-5, 7-11, 14-26, 30, 32-39 have been examined.
Election/Restrictions
Applicant’s election without traverse of Group I, claims 1-31, in the reply filed on 9/5/25 is acknowledged.
Response to Arguments
Applicant's arguments with respect to the claims have been considered but are moot in view of the new ground(s) of rejection. On 4/30/26, Applicant amended the independent claims. Applicant’s remarks address this added amended feature. See the new 103 with the addition of Lewis (previously listed on Notice of References cited) that address these new features.
Also, the 4/30/26 amendments and remarks are considered substantive to pass 101.
Also, the prediction server in claim 1, 30 is interpreted as a physical server so claim 1 is Not interpreted as software per se (see Applicant Spec at “[8]… the prediction server comprising a server configured to receive an advertisement request from a client demand-side platform (DSP)”).
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-5, 7-11, 14-26, 30, 32-39 are rejected under 35 U.S.C. 103 as being unpatentable over Skudlark (20100057560) in view of Lewis (10387921) in view of Shor (20200184510).
Claim 30, 1 (independent) and also dependent claims 2, 3, 5, 6, 8-15, 20-24, 26, 27: Note that independent claim 30 includes the features of independent claim 1 and also some of claim 1’s dependents. So, Claim 30 is the first claim listed as rejected. Skudlark discloses a performance optimization system (POS) comprising:
a POS data platform configured to store data usable to determine the POS score (Fig. 1);
a machine learning platform configured to use machine learning (Fig. 1 and see neural networks and machine learning at [78]) and also determine the POS score (see score at [49]) and also uses real time data for the POS score (see real time and score at [94]). Skudlark does not explicitly disclose the machine learning determining the score or the machine learning platform operably connected to the POS data platform or to determine the POS score in real time. However, Skudlark discloses a plurality of models and scoring and predicting and data [49, 53, 62, 75, 94, 98]. And, Skudlark discloses the models connected to the scoring data (Fig. 1). And, Skudlark discloses using machine learning for predicting and models [78]. Therefore, it would have been obvious to one having ordinary skill in the art at the time the invention was made to add Skudlark’s models and machine learning to Skudlarks’s models and score so that Skudlark can use machine learning models to score. One would have been motivated to do this in order to better score using available computational techniques.
And, in further regards to determine the POS score in real time, Skudlark does not explicitly disclose to determine the POS score in real time. However, Skudlard discloses and also uses real time data for the POS score (see real time and score at [94]). And, Skudlark further discloses that customer real time data is used to create a customer behavior predictor used to estimate customer likely ad response [0014]. And, this predictor And, this predictor presents a score for the ad appeal [49, 55]. Hence, real time data are used to create a customer behavior predictor that generates a score for customer likely response to advertisements. In another embodiment, Skudlark discloses that customer data is used to generate a score for the ad [75]. And, customer data includes real time customer data [14]. In another embodiment, Skudlark discloses live/realtime/current customer information like current weather and events is used to asses responsiveness in a particular environment/current weather/current event [113, 1]. And, responsiveness to ads can be scored [98]. Hence, current responsiveness to ad can be scored based on current environment/weather in real time. And, Skudlark discloses that real time customer info can be used to refine the predictor and model [100, 92] and this predictor and model produces a score [49, 55]. And, Skudlark discloses that there are millions of sites and customers for a particular ad and that these particular ads need scored for the particular sites and customer combos [75]. And, at [94], Skudlark discloses that real time data are used to refine the model and that customer response may be correlated with indicia to generate indication of customer interest and that these indications may be expressed as a score (“[94]… the real time data are used to create and refine the model 154…. Customer response and access to the various services may be correlated with the indicia surrounding the services to generate indications of customer interest. These indications may be expressed in the form of interest scores, for example.”). Therefore, it would have been obvious to one having ordinary skill in the art at the time the invention was made to add Skudlark’s realtime data used to refine the model to produce a score to Skudlark’s site and customer combo that needs a score now for the customer and site combo that needs an ad so that a score can now, or in real time, be provided based on current or real time data and conditions. One would have been motivated to do this in order to better present an ad now based on the current site and customer and the current customer data and conditions/environment.
Skudlark further discloses a prediction server operably connected to the machine learning platform, the prediction server comprising a server configured to receive an advertisement request from a client demand-side platform (DSP) (see Fig. 1 and predictor server at item 152 which responds to the ad request from the end user devices in Fig. 1), the prediction server further configured to create a prediction request from the advertisement request by selecting relevant prediction request data from the advertisement request (see Fig. 1 and predictor server at item 152 which responds to the ad request from the end user devices in Fig. 1 and uses the model 154 and the ad data at 130; see [97, 54, 56]), the prediction request data comprising end user data (see Fig. 1 and predictor server at item 152 uses the model 154 and the customer profile data at 162, see Fig. 2 with predictor 152 connected to user data at 132 and 146), the prediction server then copying the relevant prediction request data to the prediction request, the prediction server sending the prediction request to the machine learning platform (see neural networks and machine learning at [78] and see machine generated filter at Fig. 2 item 215, see predict and advertisements and responsiveness and influence at [81]),
the prediction server further configured to score the prediction request to determine a likelihood to influence an end user by exposing the end user to the brand advertisement (see predict and advertisements and responsiveness and influence at [81], see score and predict and ad at [49, 53, 55, 98], see score at [53]).
Skudlark does not explicitly disclose wherein the prediction server randomly labels a predetermined fraction of prediction requests as control prediction requests belonging to a control group. However, Lewis discloses machine learning for advertising effect (41:60-42:15) and further discloses wherein the prediction server randomly labels a predetermined fraction of prediction requests as control prediction requests belonging to a control group (see control group and test group where test group in Lewis reads claimed treatment group, 7:57-8:50). Therefore, it would have been obvious to one having ordinary skill in the art at the time the invention was made to add Lewis ML and advertising and control and test/treatment group to Skudlark’s ML and advertising. One would have been motivated to do this in order to better use ML to asses advertising effect.
Skudlark further discloses a prediction request log operably connected to the prediction server (see report, predict, score at [49], where report reads on log).
Skudlark does not explicitly disclose wherein the prediction server logs in the prediction request log a control prediction request belonging to the control group, wherein the prediction server logs in the prediction request log a remaining prediction request as a treatment prediction request belonging to a treatment group. However, Lewis discloses machine learning for advertising effect (41:60-42:15) and further discloses wherein the prediction server logs in the prediction request log a control prediction request belonging to the control group, wherein the prediction server logs in the prediction request log a remaining prediction request as a treatment prediction request belonging to a treatment group (see control group and test group where test group in Lewis reads claimed treatment group and note that of the 100 users some are in control group while rest are in treatment/test group, 7:57-8:50). Therefore, it would have been obvious to one having ordinary skill in the art at the time the invention was made to add Lewis ML and advertising and control and test/treatment group to Skudlark’s ML and advertising. One would have been motivated to do this in order to better use ML to asses advertising effect.
In further regards to claim 1, the prediction request log feature is found in the following features of claim 30. In further regards to dependent claims 2, 3 5, the copy feature, end user data feature, and send the prediction request features are found in the prediction server clause above.
Skudlark further discloses wherein the machine learning platform comprises a model builder configured to build a predictive model usable to determine the POS score (see model and score and predict at [49, 53, 62]; note neural networks and machine learning at [78] ), the machine learning platform further comprising a model scorer operably connected to the model builder, the model scorer configured to receive the predictive model from the model builder, the model scorer further configured to use the predictive model to determine the POS score, wherein the POS score comprises an end user engagement metric predicting end user engagement with an advertisement (Figs. 1, 2, [49, 53, 55, 62, 81, 98]).
Skudlark further discloses an estimate of whether an end user receiving an advertisement is likely to be influenced by exposure to the brand advertisement (“[0014] Customer static data and customer real time data are used to create a customer behavior predictor used to estimate a customer's likely response to advertisements and other content,”; also note predict and advertisements and responsiveness and influence at [81]).
Skudlark does not explicitly disclose wherein the POS score further comprises, a purchase intent score, and another metric configured to estimate awareness of an end user of an advertised brand. However, Skudlark discloses scores related to ads and ad appeal [49] and also interest scores [94] and scores for response advertiser is trying to elicit [98] and tracking interest in particular brands or products that may reflect interest in a purchase and how this correlates to info/ad interest (see Honda and automobile and purchase and interest at [71]). Also, Skudlark further discloses scores and a likelihood to influence an end user by exposing the end user to the brand advertisement (see score and predict and ad at [49, 53, 55, 98], see score at [53]) and also predict level of interest and scores [53, 55] and likelihood and scores [75]. Therefore, it would have been obvious to one having ordinary skill in the art at the time the invention was made to add Skudlarks’s brand and product and purchase interest Skudlark’s variety of scores related to ad influence or response so that brand awareness scores, purchase intent scores, brand consideration scores can be scored. One would have been motivated to do this in order to better present a useful score.
Skudlark further discloses wherein the prediction server is further configured to add the POS score to the prediction request, creating a scored prediction request, wherein the prediction server is further configured to send the scored prediction request to the client DSP (see predict and score at [49, 53, 55, 98]; see network connections in Figs. 1, 2 to client), wherein the prediction server is further configured to log the scored prediction request in the prediction request log, wherein the prediction request log is configured to log the scored prediction request (see report, predict, score at [49], where report reads on log), wherein the POS data platform comprises a profile store comprising a plurality of profiles of end users, the POS data platform further comprising a model store configured to store predictive models that the model builder builds (see profile and predictor at Figs. 1, 2; see score and customer information at [53, 62]), wherein the performance optimization system is operably connected to a client demand-side platform (DSP), wherein the client DSP comprises an entity configured to do one or more of run an advertising campaign directly as an advertiser and run the advertising campaign on behalf of an advertiser (see Fig. 1 and advertisement manager 166 and/or advertisement service 170), wherein the model scorer scores the prediction request without using the end user data (see anonymize at [50, 57]; see [66] where customer information is removed and anonymous information is used, so end user data, such as particular user information, is not used in the prediction or score), wherein the model scorer scores the prediction request without using personally identifiable information (PII) regarding the end user (see anonymized at Abstract and [37, 46, 50]), wherein the prediction request comprises a request for a POS score from a client DSP to the prediction server, wherein the model builder builds the model using customer engagement data (see customer and predict at Figs. 1, 2; see customer and score at [49, 53, 55, 98]).
Skudlark further discloses wherein the customer engagement data comprises survey responses (see questions and questionnaires and preferences which reads on survey at [36, 38], and questions and customer interests at [87]) and also advertising and campaign details [26]. Skudlark does not explicitly disclose the survey responses comprising an end user’s response to a customer engagement campaign or wherein a relatively small number of survey responses are needed with the consent of the end user to build a predictive model that can then be used to target all end users, whether those end users provide their consent or not. However, Examiner notes that Applicant Spec at [100] was found relevant for this feature. And, Skudlark discloses questions for customer preferences and interests [36, 38, 87] and that these go into the customer profile [36, 87]. And, Skudlark discloses that the customer profile tracks behavior and interest in a product [79] and that the profile is uses to develop a model for response to advertising to better presents ads aligning with interests [12, 13] and that the profile has customer interactions and behaviors [13] and profiles and measuring response to ads [28] and also that ads are in a campaign and targeting profiles for ad campaigns [26] and also using machine learning [78] and targeting [26, 80]. And, Shor further discloses wherein the customer engagement data comprises survey responses and also that the survey responses comprising an end user’s response to a customer engagement campaign (see survey and ad at [14, 19]). And, Shor discloses that the survey is related to ads and multiple ads [7, 19, 24]. And, Shor also further discloses wherein a relatively small number of survey responses are needed with the consent of the end user to build a predictive model that can then be used to target all end users, whether those end users provide their consent or not (Shor shows surveying uses on an ad response [14, 19] and then taking that survey sample to build a predictive model [18-20] where that small survey group is used to find a larger unsurveyed group to target [26, 28-29] and also Figs. 1, 3; also, Examiner interprets that the surveyed users are providing consent via their answers to direct questions, meanwhile, the unsurveyed, target users do Not provide any direct consent since they are not surveyed). Shor also discloses using machine learning [20] and targeting (Abstract, [7]). Also, Shor discloses taking a sample of surveyed users to extrapolate to a larger unsurveyed group. The sampled users provide their consent via the survey and direct answers to questions on personal items (Shor discloses that the survey takers provide detailed user info like name and birthdate [21] and PII [26]). So, this group in Shor provides consent via their answers to PII and personal info questions. Then, in Shor there is the extrapolation to the unsurveyed group who has Not provided PII type info. This also makes the combination of Skudlark and Shor possible since both protect PII and anonymous users. Therefore, it would have been obvious to one having ordinary skill in the art at the time the invention was made to add Shor’s surveys and models and machine learning for ad(s) targeting to Skudlarks’s profiles and questions on preferences/interests and Skudlarks’s ad responses and behavior and models and ad campaigns and machine learning for targeting. One would have been motivated to do this in order to better track ad response and interests in order to better use models and machine learning to target.
In regards to dependent claim 6, 8, 9, 10, 11, the POS score feature and DSP feature and log feature and profile store feature and profile/model store features are found in the clause preceding. In further regards to dependent claim 12, 13, 14, 15, 20-24, 26, 27, the features are found in the clause preceding.
Claim 4. Skudlark further discloses the performance optimization system of claim 3, wherein the end user data comprises one or more of end user personal data, end user device data, contextual data, advertisement spot data, website data, mobile app data, network data, and privacy data (see profile at Figs. 1, 2, see interactions and behavior at [13], see demographic, geographic, behavior, context data at [34]).
Claim 7. Skudlark does not explicitly disclose the performance optimization system of claim 6, the machine learning platform comprising both the model builder and the model scorer. However, Skudlark discloses using models for building and scoring (see model building, creating at [12, 14] and model and score at [49]) and also using machine learning techniques ([78]). Therefore, it would have been obvious to one having ordinary skill in the art at the time the invention was made to add Skudlark’s machine learning to Skudlark’s model building and model scoring. One would have been motivated to do this in order to better score using available machine learning and computational techniques.
Claim 16. Skudlark further discloses the performance optimization system of claim 15, wherein the client DSP comprises an entity doing one or more of running an advertising campaign directly as an advertiser and running the advertising campaign on behalf of an advertiser (see advertiser at Fig. 1).
Claim 17. Skudlark further discloses the performance optimization system of claim 1, wherein the performance optimization system is operably connected to an analytics end user, the analytics end user comprising an end user of data collected by the performance optimization system (see Fig. 1, 2 with end user data, and see report at [49).
Claim 18. Skudlark further discloses the performance optimization of claim 17, wherein the performance optimization system provides analytics data to the analytics end user using one or more of the scored prediction request and the end user profiles stored in the profile store (see Fig. 1, 2 with predictor and profile and see report at [49]).
Claim 19. Skudlark further discloses the performance optimization system of claim 17, wherein the client DSP comprises the analytics end user (see Fig 1 and see report at [49]).
Claim 25. Skudlark further discloses the performance optimization system of claim 1, wherein the prediction server receives the advertisement request directly from an SSP (see advertiser at Fig. 1).
Claim 32, 33. Skudlark does not explicitly disclose the performance optimization system of claim 1, wherein the performance optimization system effectively targets advertisements to end users when a number of end users providing consent is as low as ten percent (10%). However, the prior art combination renders obvious the survey and predictive model features above. And, Shor further discloses using a sample of surveyed users to target a bigger, unsurveyed group (Figs. 1, 3; [18-20, 26-29]). And, Shor further discloses that the size of the survey sample group compared to the unsurveyed target group can vary in order to more precisely target or in order to have a broader reach [5]. And, the MPEP states that a change in size/proportion (MPEP 2144.04.IV.A) and that ranges, amounts and proportions are obvious (2144.05). Therefore, it would have been obvious to one having ordinary skill in the art at the time the invention was made to add Shor’s range of sample size such that is can be as low as 10% to Skudlarks’s ad responses and behavior and models and machine learning and targeting. One would have been motivated to do this in order to better use models and machine learning to target and better find a balance between targeting accuracy and reach (as Shor discloses at [5]).
Claim 34, 37. Skudlark does not expliclity disclose the performance optimization system of claim 1, wherein the prediction server randomly labels 1% of the prediction requests as the control prediction requests. However, Skudlark in view of Lewis renders obvious the control and treatment/test group features above. And, Lewis further discloses machine learning for advertising effect (41:60-42:15) and where the ratios can vary of control to treatment/test and some of the 100 are in control while the remaining are in treatment/test (see control group and test group where test group in Lewis reads claimed treatment group and note that of the 100 users some are in control group while rest are in treatment/test group, 7:57-8:50; and 14:55-60, “The threshold can be set to produce a desired test/control group ratio for the particular experiment; for example, 50/50, 90/10, 10/90, etc.”). And, the MPEP states that changes in size/proportion (MPEP 2144.04.IV.A) or changes in range/proportion (MPEP 2144.05.A) are obvious. Therefore, it would have been obvious to one having ordinary skill in the art at the time the invention was made to that Lewis ML and advertising and control and test/treatment group with varying percentages of control to treatment/test can be 1% control to Skudlark’s ML and advertising. One would have been motivated to do this in order to better use ML to asses advertising effect and better keep advertiser costs down as a smaller control group means less wasted money for the advertiser (as Lewis states at 7:60-8:5, “That is, if a test group contains…exposure rate for its cost.”).
Claim 35, 38. The prior art further renders obvious the performance optimization system of claim 1, wherein the prediction server logs in the prediction request log each of the control prediction requests (Lewis further discloses the control and treatment/test group and logging each of the control prediction requests, 27:60-28:15 “…The method 800 can log the auction result and the results of all simulations (ACT 875).”, and the motivation is the same as already provided above for better use ML to asses ad effect).
Claim 36, 39. The prior art further renders obvious the performance optimization system of claim 1, wherein the prediction server logs in the prediction request log each of the remaining prediction requests as the treatment prediction requests belonging to the treatment group (Lewis already shows that requests remaining are in treatment/test group, see above, and Lewis further discloses the control and treatment/test group and logging each of the treatment/test prediction requests, 27:60-28:15 “…The method 800 can log the auction result and the results of all simulations (ACT 875).”, and the motivation is the same as already provided above for better use ML to asses ad effect).
Conclusion
The following prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Shor and the other cited art disclose relevant features.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any extension fee pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ARTHUR DURAN whose telephone number is (571)272-6718. The examiner can normally be reached Mon-Thurs, 7-5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Ilana Spar can be reached at (571) 270-7537. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/ARTHUR DURAN/Primary Examiner, Art Unit 3621 5/21/2026
Prosecution Insights