Predicting performance of creative content

Patent metadata

Field	Value
Publication number	US-12469046-B2
Application number	US-202217937342-A
Country	US
Kind code	B2
Filing date	Sep 30, 2022
Priority date	Sep 30, 2022
Publication date	Nov 11, 2025
Grant date	Nov 11, 2025

Abstract

Methods and systems for predicting performance of creative content are disclosed. Exemplary implementations may: receive a collection of images; provide a context to a user; serially cause display of pairs of images on a computer interface; receive user responses indicating which image of each pair is preferred given the context; determine a resonance value for each image based on a number of times the user responses indicate each image is preferred when displayed in a pair of images; determine a confidence score for each image; generate one or more models for predicting image performance based on one or more of the resonance value and the confidence score for each image; receive a plurality of candidate images; determine, using at least one model, a first metric set for each candidate; and cause display of a listing of the candidate images, the listing including the first metric set for each candidate image.

First claim

What is claimed is:

1 . A computer-implemented method for predicting performance of creative content, the method comprising:
receiving, by one or more modules of a predictive content performance system, a collection of images from one or more image sources; generating, based on the collection of images, a network graph, wherein each node of a plurality of nodes of the network graph represents an image of the collection of images, and wherein each link of a plurality of links of the network graph represents a context associating at least two nodes; providing, by the one or more modules of the predictive content performance system, the context to at least one user; serially causing display, by the one or more modules of the predictive content performance system, of pairs of images from the collection of images on a first computer interface, wherein the pairs of images are associated by the context; receiving, by the one or more modules of the predictive content performance system, from the at least one user, multiple user responses indicative of which image of each pair of images displayed on the first computer interface is preferred in view of the context; determining, using a ranking algorithm, a resonance value for each image of the collection of images based, at least in part, on a number of times the user responses indicate each image of the collection of images is preferred when displayed in a pair of images; generating a first similarity vector for each image of the collection of images by comparing at least a first attribute of each image with at least a second attribute of a subset of images of the collection of images, the first attribute including pixel data; generating, by the one or more modules of the predictive content performance system, a training dataset comprising at least the collection of images, the user responses, the resonance value for each image of the collection of images, and the first similarity vector for each image of the collection of images; generating, by the one or more modules of the predictive content performance system, based on the training dataset, one or more models for predicting image performance by training a machine learning algorithm to label each image of a plurality of images with a respective resonance value; providing, by the one or more modules of the predictive content performance system, to the one or more generated models, a plurality of candidate images for a creative campaign; determining, by the one or more generated models, a set of resonance values for the plurality of candidate images for the creative campaign; and causing display, by the one or more modules of the predictive content performance system, of a listing of the candidate images on a second computer interface, the listing comprising a resonance value of the set of resonance values for each candidate image of the plurality of candidate images.

2 . The computer-implemented method of claim 1 , wherein:
the context is indicative of a segment selected from one of an interest segment, a business segment, or an audience segment; the network graph includes two or more nodes of the plurality of nodes connected by one or more of the plurality of links, each node of the two or more nodes associated with an image related to the segment; and serially causing display of the pairs of images comprises selecting the pairs of images from a first node and a second node in the network graph that are separated by one link.

3 . The computer-implemented method of claim 1 , wherein serially causing display of the pairs of images comprises:
selecting, from a database, two or more images associated with the context, the two or more images represented as two or more nodes of the network graph, each node having a same number of links joining the node with another node; and selecting a pair of images from a first node and a second node joined by a link.

4 . The computer-implemented method of claim 1 , further comprising:
aggregating user response data based on receiving the user responses for the pairs of images from the collection of images.

5 . The computer-implemented method of claim 1 , further comprising selecting the context from one or more categories selected from a group consisting of fashion, fitness, food, business, creative, real estate, beauty, medical, consumer goods, travel, outdoors, and home services.

6 . The computer-implemented method of claim 1 , wherein the one or more models comprise one or more of a general audience model, a social media model, a stock images model, an interests model, an audience segment or audience cohort model, a context model, and a segment model.

7 . The computer-implemented method of claim 1 , wherein generating the one or more models for predicting image performance comprises:
identifying one or more attributes for each image from the collection of images, wherein the one or more attributes include one or more of stylistic information, an image vector, keyword metadata or tags, and labels; grouping the collection of images into one or more groups based on the resonance value from the user responses; and determining an average of attributes for each group of the one or more groups based at least in part on the one or more attributes for each image in each group.

8 . The computer-implemented method of claim 1 , further comprising:
determining one or more attributes for one or more of the plurality of candidate images; determining a second similarity vector for each candidate image of the plurality of candidate images by comparing the one or more attributes for each candidate image of the plurality of candidate images with an attribute average for a subset of images of the plurality of candidate images; and assigning a plurality of labels to the candidate images, based on the second similarity vector.

9 . The computer-implemented method of claim 1 , further comprising:
receiving advertisement performance data for a plurality of advertisements, wherein each advertisement of the plurality of advertisements displays a candidate image from the plurality of candidate images; determining, for each image displayed in an advertisement, an aggregate score based, at least in part, on the resonance value of the set of resonance values for each candidate image; and validating the one or more generated models for predicting image performance based, at least in part, on comparing the respective aggregate score and the respective advertisement performance data.

10 . A system configured for predicting performance of creative content, the system comprising:
one or more hardware processors configured by multiple machine-readable instructions to:
generate, by one or more modules of a predictive content performance system, based on a collection of images, a network graph, wherein each node of a plurality of nodes of the network graph represents an image of the collection of images, and wherein each link of a plurality of links of the network graph represents a context associating at least two nodes;
access, by one or more modules of a predictive content performance system, the network graph;
provide, by one or more modules of a predictive content performance system, the context to at least one user;
serially cause display, by the one or more modules of the predictive content performance system, of pairs of images from the network graph, on a first computer interface, wherein the pairs of images are associated by the context;
receive, by the one or more modules of the predictive content performance system, from the at least one user, multiple user responses indicative of which image of each pair of images displayed on the first computer interface is

Assignees

• Shutterstock Inc

Inventors

• Teoh Emily
• Finnerty Diarmaid
• Burnham-King Lauren Sarah
• Darling Veronica
• Sala Alessandra

Classifications

• G06F3/0482

  Interaction with lists of selectable items, e.g. menus · CPC title

• G06V2201/10

  Recognition assisted with metadata · CPC title

• G06V20/30

  in albums, collections or shared content, e.g. social network photos or video · CPC title

• G06N5/022

  Knowledge engineering; Knowledge acquisition · CPC title

• G06Q30/0277

  Online advertisement · CPC title