Method and system for forecasting a reach rate of a color cosmetic shade
A data-driven method using consumer attribute datasets and TURF analysis optimizes color cosmetic shade ranges, addressing inefficiencies in traditional forecasting methods by ensuring accurate and inclusive shade recommendations.
Patent Information
- Application Number
- PCT/CN2025/093344
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-05-08
- Publication Date
- 2026-11-12
AI Technical Summary
Traditional methods for developing color cosmetic shade ranges are time-consuming, costly, and ineffective in forecasting the reach rate of color cosmetic shades, failing to capture diverse consumer preferences.
A data-driven method utilizing consumer attribute datasets, image processing, and analytical techniques to calculate reach rates of color cosmetic shades, incorporating a unique color library and TURF analysis to optimize shade ranges.
Efficiently and accurately forecasts the reach rate of color cosmetic shades, ensuring inclusivity and reducing computational complexity by leveraging standardized lighting, CIELAB color space, and TURF analysis for optimized shade recommendations.
Smart Images

Figure CN2025093344_12112026_PF_FP_ABST
Abstract
Description
METHOD AND SYSTEM FOR FORECASTING A REACH RATE OF A COLOR COSMETIC SHADEFIELD OF THE INVENTION
[0001] The present disclosure relates generally to the field of cosmetics, and more specifically, to a method and system for forecasting a reach rate of a color cosmetic shade.BACKGROUND OF THE INVENTION
[0002] Color cosmetic, such as a lipstick or other product that may be applied to a keratin material (such as skin, nails, or hair) , plays a crucial role in self-expression and aesthetics. However, selecting a flattering shade remains a persistent challenge. A key aspect of a successful color cosmetic product line is the availability of a color cosmetic shade range that is considered "suitable" or appealing by a wide variety of consumers.
[0003] Traditionally, the development of a color cosmetic shade range has relied heavily on the expertise of colorists, makeup artists, and product developers. This process often involves extensive market research, trend analysis, and consumer testing, typically in the form of focus groups and in-person trials. While these methods provide valuable insights, they are inherently time-consuming, expensive, and often limited in scope. They may not fully capture the diverse preferences of a large and heterogeneous consumer population. Moreover, some traditional methods can be used to assess the suitability of individual shades, but such method cannot forecast the reach rate of a color cosmetic shade or a range of color cosmetic shades.
[0004] Therefore, a need exists for a method and system that can efficiently, cost-effectively and accurately forecast a reach rate of a color cosmetic shade or a range of color cosmetic shades.SUMMARY OF THE INVENTION
[0005] The summary is provided to introduce a selection of concepts in a simplified form that are further described below in detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.
[0006] Various aspects and features of the disclosure are described in further detail below.
[0007] According to one aspect of the disclosure, a method for forecasting a reach rate of a color cosmetic shade is provided. The method comprises: obtaining test application color data of the color cosmetic shade when applied to a keratin material; obtaining a set of records associated with a consumer attribute from a dataset, each record including application color data and a suitability metric indicating a perceived suitability of the color cosmetic shade; identifying, within the set of records, a subset of records with application color data that fall within a first predetermined color difference threshold of the test application color data; and calculating the reach rate as m / n, where n is a total number of records within the subset of records, and m is a number of records, within the subset of records, that have a suitability metric exceeding a predetermined suitability threshold.
[0008] Preferably, the consumer attribute is a skin tone, the method further comprising: calculating the reach rates of the color cosmetic shade for a plurality of different skin tones; and calculating a group reach rate of the color cosmetic shade for the target group based on the reach rates of the color cosmetic shade for the plurality of different skin tones.
[0009] Preferably, the consumer attribute is at least one of age of consumer or country of consumer.
[0010] Preferably, the method further comprises: calculating the reach rates of a plurality of color cosmetic shades in a color cosmetic shade range; and calculating a range reach rate of the color cosmetic shade range by aggregating the reach rates while accounting for overlapping consumers across the plurality of color cosmetic shades.
[0011] Preferably, obtaining the test application color data of the color cosmetic shade when applied to a keratin material comprises: applying the color cosmetic shade to the keratin material; capturing image data of the color cosmetic shade as applied to the keratin material; and processing the image data to determine the test application color data of the color cosmetic shade.
[0012] Preferably, processing the image data comprises determining the test application color data as a median color value from a plurality of pixel color values within a detected application region in the image data.
[0013] Preferably, the test application color data is represented in CIELAB color space, and / or wherein the first predetermined color difference threshold is based on the CIEDE2000 color difference formula.
[0014] Preferably, the method further comprises creating a unique color library, wherein creating the unique color library comprises selecting a representative color from each of a plurality of groups of application color data to be included in the unique color library.
[0015] Preferably, the plurality of groups of application color data are formed by grouping application color data from the dataset based on a second predetermined color difference threshold.
[0016] Preferably, selecting the representative color further comprises filtering the representative colors to ensure that color difference between any two representative colors remaining in the unique color library exceeds a third predetermined color difference threshold.
[0017] Preferably, the method further comprises: selecting, from the unique color library, a plurality of representative colors, to maximize a range reach rate of a corresponding color cosmetic shade range, based on a constraint, preferably, wherein the constraint specifies at least one of: a desired number of color cosmetic shades in the color cosmetic shade range, a desired proportion of color cosmetic shades from different color families, and one or more existing color cosmetic shades that must be included in the color cosmetic shade range.
[0018] Preferably, the selecting is performed using a Total Unduplicated Reach and Frequency (TURF) analysis.
[0019] Preferably, the method further comprises providing the plurality of color cosmetic shades corresponding to the plurality of representative colors as a recommendation.
[0020] According to another aspect of the disclosure, a system for forecasting a reach rate of a color cosmetic shade is provided, the system comprising: at least one memory; and at least one processor coupled to the at least one memory, wherein the at least one processor is configured to perform operations of the methods disclosed herein.
[0021] According to yet another aspect of the disclosure, a non-transitory computer-readable medium is provided. The non-transitory computer-readable medium stores instructions that, when executed by a processor, cause the processor to perform operations of the methods disclosed herein.
[0022] The present disclosure offers at least one of the following advantages over prior art methods. By utilizing a data-driven approach and incorporating key features such as test application color data and records associated with consumer attributes such as skin tones, the disclosure achieves efficient and cost-effective forecasting of a reach rate of a color cosmetic shade. By calculating a group reach rate, the disclosure is capable of forecasting a reach rate of a color cosmetic shade for a target group. By calculating a range reach rate, the disclosure is capable of forecasting a reach rate of a range of color cosmetic shades. The use of a pseudo-spectral camera system and CIELAB color space ensures accurate and consistent color data. The creation and utilization of a unique color library, with its grouping and selection process based on color difference and reach rate, further enhances the efficiency and effectiveness of shade range recommendations by ensuring perceptual distinctiveness and reducing computational complexity. The use of analytical methods, such as TURF analysis, allows for the data-driven optimization of shade ranges, taking into account various constraints and maximizing the overall reach for a target consumer group. Moreover, the inclusivity of the recommended shades can be quantified by the reach rate and maximized by the use of the analytical methods.BRIEF DESCRIPTION OF THE DRAWINGS
[0023] The above and other aspects, features, and benefits of various embodiments of the disclosure will become more fully apparent, by way of example, from the following detailed description with reference to the accompanying drawings, in which like reference numerals or letters are used to designate like or equivalent elements. The drawings are illustrated for facilitating better understanding of the embodiments of the disclosure and not necessarily drawn to scale, in which:
[0024] Fig. 1 is a flowchart illustrating an example method for forecasting a reach rate of a color cosmetic shade, according to one or more embodiments of the present disclosure.
[0025] Fig. 2 is a schematic diagram illustrating a workflow for forecasting a reach rate of a color cosmetic shade, according to one or more embodiments of the present disclosure.
[0026] Figs. 3A and 3B are diagrams illustrating the data collection process for building an application color database, according to one or more embodiments of the present disclosure.
[0027] Fig. 4 is a schematic diagram illustrating a workflow for generating a unique color library, according to one or more embodiments of the present disclosure.
[0028] Fig. 5 is a schematic diagram illustrating a workflow for utilizing a unique color library with analytical methods for applications.
[0029] Fig. 6 is a block diagram of an example system for forecasting a reach rate of a color cosmetic, according to one or more embodiments of the present disclosure.
[0030] Fig. 7 is a block diagram of an example device for forecasting a reach rate of a color cosmetic shade, according to one or more embodiments of the present disclosure.DETAILED DESCRIPTION
[0031] Embodiments herein will be described in detail hereinafter with reference to the accompanying drawings, in which embodiments are shown. These embodiments herein may, however, be embodied in many different forms and should not be construed as being limited to the embodiments set forth herein. The elements of the drawings are not necessarily to scale relative to each other. Like numbers refer to like elements throughout. It is to be understood that any depiction of human faces or facial features in the drawings is for illustrative and exemplary purposes only and represents computer-generated imagery or stylized representations, and does not depict any actual individual.
[0032] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting. As used herein, the singular forms "a" , "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" , "comprising" , "includes" and / or "including" when used herein, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.
[0033] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meanings as commonly understood. It will be further understood that a term used herein should be interpreted as having a meaning consistent with its meaning in the context of this specification and the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
[0034] The present disclosure is described below with reference to block diagrams and / or flowchart illustrations of methods, apparatus (systems) and / or computer program products according to the present embodiments. It is understood that blocks of the block diagrams and / or flowchart illustrations, and combinations of blocks in the block diagrams and / or flowchart illustrations, may be implemented by computer program instructions. These computer program instructions may be provided to a processor, controller or controlling unit of a general-purpose computer, special purpose computer, and / or other programmable data processing apparatus, such that the instructions, which execute via the processor of the computer and / or other programmable data processing apparatus, create means for implementing the functions / acts specified in the block diagrams and / or flowchart block or blocks.
[0035] Accordingly, the present technology may be embodied in hardware and / or in software (including firmware, resident software, micro-code, etc. ) . Furthermore, the present technology may take the form of a computer program product on a computer-usable or computer-readable storage medium having computer-usable or computer-readable program code embodied in the medium for use by or in connection with an instruction execution system. In the context of this document, a computer-usable or computer-readable medium may be any medium that may contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.
[0036] As mentioned above, traditional approaches for development of color cosmetic shades have difficulties in forecasting a reach rate of a color cosmetic shade, and are also lengthy and costly.
[0037] The present disclosure provides a method and system for forecasting a reach rate of a color cosmetic shade or a range of color cosmetic shades. This is achieved through a data-driven calculation of the percentage of consumers who consider the color cosmetic shade or the range of color cosmetic shades suitable.
[0038] As used herein, "color cosmetic" refers to any cosmetic product whose purpose is to impart color when applied to a keratin material. Examples include, but are not limited to, lip products (lipsticks, glosses, stains) , face products (foundations, concealers, blushes, eyeshadows) , hair products (dyes, temporary colorants) , and nail products (polishes) . As used herein, "keratin material" refers to biological materials primarily composed of keratin. Examples of "keratin material" relevant to cosmetic application include, but are not limited to, skin (including facial skin and lips) , hair, and nails. As used herein, "shade" , "cosmetic shade" or "color cosmetic shade" refers to a specific color variant of a color cosmetic product. As used herein, ′application color data′refers to data representing the color appearance of a color cosmetic shade after it has been applied, physically or by virtually, to a keratin material, such as a skin model or a consumer′sskin, hair, or nails. This is distinct from the inherent color data of the color cosmetic itself or the color cosmetic shade.
[0039] While the methods and systems described herein are applicable to a wide range of color cosmetics and keratin materials, the following detailed description will focus on a preferred embodiment involving the prediction of lipstick shade suitability when applied to lips. This focus is for illustrative purposes and is not intended to limit the scope of the invention.
[0040] Fig. 1 is a flowchart illustrating an example method 100 for forecasting a reach rate of a color cosmetic shade, according to one or more embodiments of the present disclosure. The method 100 is a computer-implemented method, performed by a processing unit. The processing unit may be a general-purpose computer, a specialized processor, or any other suitable computing device capable of executing instructions stored in a memory. Reference may be made to Fig. 2, which is a schematic diagram illustrating a workflow for forecasting a reach rate of a color cosmetic shade, according to one or more embodiments of the present disclosure.
[0041] As shown in Fig. 1, the method 100 may comprise, at operation 102, obtaining test application color data of the color cosmetic shade when applied to a keratin material. The test application color data is data representing the color appearance of the color cosmetic shade under test after it has been applied.
[0042] In an example, the test color cosmetic shade is applied to a keratin material, for example, a keratin material model, which can be a physical model or a computer-simulated model) .
[0043] Although a specific test color cosmetic shade is described in operation 102, the approach described above is applicable to a range of different test color cosmetic shades (as illustrated at block 202 in Fig. 2) , and a range reach rate of the color cosmetic shade range may be calculated, as described later in reference to the range reach rate.
[0044] Image data of the color cosmetic shade (such as the lipstick shade) , as applied to the keratin material (such as the lip) , is then captured. As shown at block 204 in Fig. 2, preferably, a pseudo-spectral camera system is be used for capturing image data of the applied lipstick shade. This system utilizes two lights to simulate D65 standard illuminant lighting conditions and a DSLR camera (such as a Canon DSLR camera) for capturing images. D65 is a standard illuminant defined by the International Commission on Illumination (CIE) . The use of D65 lighting is preferred to ensure consistency with industry standards for color evaluation; however, other standardized lighting conditions could also be employed. Leveraging standardized lighting and color calibrated cameras, this system can use mathematical models to reconstruct spectrum based on the raw RGB values of camera sensors. This spectrum can provide more dimensional information of color, ensuring the fidelity and consistency of captured color data. While this specific camera system is preferred for its accuracy and color fidelity, other suitable imaging devices capable of capturing color information could also be used.
[0045] The captured image data is then processed to determine the test application color data for the specific skin tone. Preferably, this processing involves detecting the test application region (such as the lip region in the case of lipstick) using an appropriate detection algorithm (such as a lip detection algorithm) . Several lip regions with different lipstick shades applied are illustrated at block 206 in Fig. 2. For example, an image segmentation algorithm, such as those available in the OpenCV library (Open Source Computer Vision Library) , may be used. The pixel color values within the detected application region are then analyzed to determine test application color data. Preferably, the test application color data is the median color value of the pixels within the detected application region. This approach was selected based on its effectiveness in representing the applied application color. The median color value, typically obtained as RGB values, is then converted to the CIELAB color space (or, optionally, the CIE LCh color space) for further processing. The CIELAB color space is a device-independent color space that is designed to be perceptually uniform, meaning the numerical color differences generally correspond to perceived color differences. The use of the CIELAB color space (using a color space transformation algorithm such as rgb2lab function) facilitates accurate color comparisons. Although CIELAB color space is preferred, other color spaces, such as CIELUV, XYZ, YUV, CIE CAM02-UCS, CAM16-UCS, or original RGB, may be used.
[0046] After the processing, test application color data for the color cosmetic shade as applied to the keratin material is obtained, as illustrated at application color 210 in Fig. 2. This test application color data is preferably represented in the CIELAB color space (e.g., as CIELAB L*a*b*values or L*C*h values) .
[0047] As shown in Fig. 1, the method 100 may further comprise, at operation 104, obtaining a set of records associated with a consumer attribute from a dataset.
[0048] As used herein, “a consumer attribute” refers to an attribute of the consumer, particularly those relevant to the suitability of a color cosmetic shade. Skin tone is a primary example of a "consumer attribute" relevant to the suitability of many color cosmetic shades. The term "skin tone, " as used herein, refers to a categorization of human skin color, typically based on visual assessment or on objective color measurements. Examples of skin tone categories include, but are not limited to, "light, " "medium, " and "deep" , as illustrated at blocks 212 and 216 in Fig. 2. It should be noted, however, that the present disclosure is not limited to these specific categories, and any suitable skin tone classification system may be used. Other relevant consumer attributes may include age of consumer, country of consumer, hair base tone (particularly for hair colorant products) , eye color, or other characteristics relevant to cosmetic selection.
[0049] Although a single, specific consumer attribute (e.g., skin tone) is primarily discussed in relation to obtaining records in operation 104, and a single keratin material (e.g., skin) is primarily discussed for applying the test shade in operation 102, the approach described above is applicable to a plurality of different consumer attributes and keratin materials. The reach rate may be calculated for each consumer attribute individually, and optionally, aggregated across different consumer attributes representing a target group, as described later in reference to the group reach rate.
[0050] Each record may include at least application color data and a suitability metric indicating a perceived suitability of the color cosmetic shade. Preferably, each record in the dataset includes at least one of consumer attributes, such as skin tone of consumer, age of consumer, or country of consumer, allowing selecting records based on the consumer attribute, to determine a reach rate of a color cosmetic for consumers associated with a certain consumer attribute.
[0051] As can be seen at block 208 in Fig. 2, a massive dataset for consumers of a certain country is illustrated. The dataset may be categorized into a plurality of sets of records based on skin tone. As can be seen at block 208 in Fig. 2, three sets of records, corresponding to light, medium, and deep skin tones respectively, are illustrated. This categorization may be performed using various methods, including but not limited to: a computer-implemented categorization or clustering method; manual categorization based on visual assessment; objective color measurements (such as those obtained using a spectrophotometer or a calibrated camera system) ; or categorization during the data collection process itself (e.g., by separately collecting records for different skin tone groups) .
[0052] For example, the dataset may be derived from a large-scale, multi-countries project that collected data on consumer preferences for lipstick shades. It is understood that other datasets containing information on lipstick shade suitability could also be used within the scope of the present disclosure. As shown in Fig. 3A, the project involves 6 skin tone clusters from 4 countries. This project amassed over 200,000 consumer ratings of over 400 lipstick shades. This project utilized a virtual try-on system, as described in further detail with reference to Fig. 3A (block 402 in Fig. 4) and Fig. 3B (block 404 in Fig. 4) , to simulate the appearance of various lipstick shades on consumers of diverse skin tones and demographics. The project involved participants from a plurality of countries. Each participant evaluated a plurality of lipstick shades, which were virtually applied to their image. As shown in Figs. 3A and 3B, consumers viewed the virtually applied lipstick shades on a display (as shown at a display 302 in Fig. 3A and a display 304 in Fig. 3B) and provided suitability ratings using an on-screen interface. Preferably, the display is a color calibrated display to improve the accuracy of the displayed color. As shown at suitability ratings 306 in Fig. 3B, consumers provided suitability ratings for the virtually applied lipstick shades, typically on a numerical scale of 1 to 5, with 1 representing "Not Suitable" and 5 representing "Very Suitable. " As shown at a screenshot 308 in Fig. 3A, the participants provided feedback in a color self-evaluation format, indicating their agreement with statements such as "This lipstick color suits me" and "I would wear this color, " which were then quantified into the numerical suitability ratings. Specifically, the ratings were provided on a five-point scale with the following options: "Completely Disagree" (corresponding to a rating of 1) , "Somewhat Disagree" (2) , "Neither" (3) , "Somewhat Agree" (4) , and "Completely Agree" (5) . Beyond suitability, the project also collected consumer ratings on other attributes, such as wearability. The result of the project is a large database (or dataset) 310, with dataset for consumers from several countries, as shown in Fig. 3B. Preferably, the skin colors of the participating consumers are captured using a pseudo-spectral camera system.
[0053] Each record in the dataset comprises application color data and a suitability metric. Preferably, the record further comprises the skin tone of the consumer, and other parameters such as age of the consumer and country of the consumer. Each suitability metric is linked to the corresponding application color data and the skin tone, and optionally, other parameters such as age and country. As mentioned earlier, the application color data may be represented using the CIELAB color space, providing a device-independent and perceptually uniform representation of color. The suitability metric may be a suitability rating or any other suitability indication, such as a categorical label. Preferably, the suitability rating is a numerical value, as described above, indicating the suitability of the lipstick shade for the consumer’s skin color, perceived by the consumer or someone else, such as a color expert. In a non-limiting example, each record may include one or more of the following fields: consumer age and skin tone cluster, the application color data in CIELAB values (Lip_L, Lip_a, Lip_b) , the lipstick family, a model ID, a lipstick ID, and the suitability rating.
[0054] Preferably, consumers / models were classified into six skin tone clusters (C1-C6) based on extracted skin color: C1 (very light) , C2 (light cool) , C3 (light warm) , C4 (medium) , C5 (deep) , and C6 (very deep) . These clusters reflect the diverse skin tones and makeup needs of global consumers. A chi-square test of consumer ratings revealed that consumers from country A, categorized into three major skin tone groups -light skin tone (C1-C3) , medium skin tone (C4) , and deep skin tone (C5-C6) (as illustrated at block 410 in Fig. 4) -demonstrated significant differences in lipstick shade selection. The chi-square test is a statistical hypothesis test used to determine if there′sa significant association between two categorical variables. It assesses whether the observed distribution of data differs significantly from what would be expected if the variables were independent. Subsequent reach rate forecasting and shade recommendations for consumers from country A were therefore based on these three groups. In contrast, consumers from country B in the skin tone clusters C1-C3 range exhibited more homogeneous rating behavior and were therefore treated as a single light skin tone group.
[0055] To obtain the skin tone, for each consumer, images of the consumers′faces were captured using the pseudo-spectral camera system (such as the one illustrated at block 204 in Fig. 2) . Facial landmark detection, for example using the Dlib library, was then employed to identify facial features. These features were used to create a mask, isolating the skin pixels from other parts of the image. A clustering algorithm, such as k-means clustering, was then applied to the skin pixels to group them into two clusters, representing lighter and shadowed areas of the face. The centroid of the lighter cluster was selected as the representative skin color for that consumer. The representative skin color may be further categorized into any one of the predefined skin tones. This approach was selected after a visual workshop with color experts, which explored various color representation methods, including dominant color and mean color, and determined that the centroid provided the most effective representation of the skin color.
[0056] As shown in Fig. 1, the method 100 may further comprise, at operation 106, identifying, within the set of records obtained in operation 104, a subset of records with application color data that fall within a first predetermined color difference threshold of the test application color data. This operation filters the records to include only those that are similar to the specific test application color data of the test lipstick shade.
[0057] The "first predetermined color difference threshold, " as used herein, defines the maximum allowable difference between the test application color data of the test lipstick shade and the application color data in the records for those records to be considered "similar enough" for the reach rate calculation. This threshold ensures that the reach rate is calculated based on consumer preferences for colors that are perceptually close to the test shade.
[0058] Preferably, the color difference is calculated using the CIEDE2000 color difference formula. The CIEDE2000 formula is a color difference metric that is designed to be perceptually uniform, meaning that a given numerical difference in CIEDE2000 values corresponds approximately to the same perceived color difference, regardless of the specific colors being compared. This makes it well-suited for determining color similarity in this context.
[0059] The value of the first predetermined color difference threshold may be chosen based on various factors, including the desired level of precision and the characteristics of the dataset. Preferably, as supported by empirical data and expert assessments within the project discussed herein, the first predetermined color difference threshold is set to a CIEDE2000 value of no more than 1.5, as illustrated at color difference 214 in Fig. 2. It has been found that consumers generally have difficulty distinguishing between colors with a CIEDE2000 difference of less than 1.5, and colors within this threshold tend to receive similar suitability ratings. However, it is important to note that the present disclosure is not limited to this specific value, and other suitable thresholds could be used.
[0060] The operation 106 involves calculating the CIEDE2000 color difference between the test application color data (obtained in operation 102) and the application color data value of each record within the set of records obtained in operation 104 (i.e., the records associated with the specific consumer attribute, such as a specific skin tone) . If the calculated CIEDE2000 value is less than or equal to the first predetermined color difference threshold, the record is included in the subset. If the CIEDE2000 value is greater than the threshold, the record is excluded.
[0061] The result of operation 106 is a subset of records. This subset contains only those records from the original set (associated with the specific consumer attribute) that have application color considered "similar enough" to the test application color of the test lipstick shade, based on the first predetermined color difference threshold.
[0062] As shown in Fig. 1, the method 100 may further comprise, at operation 108, calculating the reach rate based on the subset of records identified in operation 106. The reach rate, as used herein, represents an estimate of the ratio or percentage of consumers (having the specific consumer attribute, or in another case, within a target consumer group) who would find the test lipstick shade suitable.
[0063] The reach rate is calculated as a ratio, m / n, where m is the number of records within the subset of records (identified in operation 106) that have a suitability metric exceeding a predetermined suitability threshold; and n is the total number of records within the subset of records.
[0064] The predetermined suitability threshold defines what is considered a "suitable" metric. Preferably, where the suitability metric is a numerical rating on a scale of 1 to 5 (as described above) , the predetermined suitability threshold is set to a value of 4. This means that records with ratings of 4 ( "Somewhat Agree" ) or 5 ( "Completely Agree" ) are counted in the numerator (m) , while records with ratings of 1, 2, or 3 are not. This threshold reflects the goal of identifying shades that are likely to be well-received by consumers, not just those that are considered "neutral" or "tolerable. " However, the present disclosure is not limited to this specific threshold, and other suitable thresholds could be used, depending on the specific application, the desired level of stringency, etc.
[0065] The denominator (n) represents the total number of records in the subset. This is the total number of consumer ratings for application colors that are considered "similar enough" to the test shade′sapplication color (based on the first predetermined color difference threshold, as described in operation 106) . For robust forecasting, the number of records ′n′in the subset is preferably above a certain threshold (e.g., 50 or 60) to ensure statistical confidence.
[0066] The numerator (m) represents the number of records within that subset that also meet the suitability criterion (i.e., have a suitability metric exceeding the predetermined suitability threshold) .
[0067] The reach rate (m / n) is therefore a percentage, representing the proportion of ratings for similar colors that are considered "suitable" . This percentage is used as an estimate of the proportion of consumers within the target group who would find the test lipstick shade suitable.
[0068] As shown at block 212 in Fig. 2, the reach rate of the shade for each skin tone may be obtained. In the examples shown in Fig. 2, the reach rate of Shade 1 for light skin is 82.6%, for medium skin is 88.1%, and for deep skin is 84.5%.
[0069] In many cases, it is desirable to determine a reach rate for a test lipstick shade across a plurality of different skin tones, which for example, represents a target group. A target group, as used herein, refers to a group formed by one or more consumers. For example, a target group may refer to a group of consumers that have a certain skin tone (or other consumer attribute) , a group of consumers that have several skin tones, a group of consumers from a certain market or country, a group of consumers of a certain age range, or an individual consumer, etc. Such a reach rate is called group reach rate in the present disclosure.
[0070] In the case that the consumer attribute is a skin tone, and the target group is represented by a plurality of different skin tones, preferably, the method further comprises calculating a reach rate for a single color cosmetic shade across a plurality of skin tones. Preferably, the plurality of skin tones represents a target group. In this instance, the reach rate may be called a group reach rate. One approach of calculating the reach rate (e.g., the group reach rate) is by weighted averaging. In this approach, the individual reach rate of the color cosmetic shade for each skin tone of the plurality of skin tones is first determined, following the procedures described in operations 102, 104, 106, and 108 described above. Then, the individual reach rates for the plurality of skin tones are multiplied by weighting factors that represent the proportion of each skin tone within the target consumer group. These weighted reach rates are then summed to produce a group reach rate. Other approaches of calculating the group reach rate based on the individual reach rates are also possible.
[0071] Often, it is also desirable to determine a reach rate for a range of color cosmetic shades, which may be called a range reach rate in the present disclosure. As can be seen at block 202 in Fig. 2, three color cosmetic shades are illustrated to represent a range of color cosmetic shades. This may be useful when developing or evaluating a new product line.
[0072] Preferably, the method may further comprise calculating a range reach rate of a color cosmetic shade range. One approach of calculating the range reach rate is by aggregating. In this approach, the reach rate of each shade is first determined, as described in operations 102, 104, 106, and 108. Then, the reach rates of a plurality of color cosmetic shades in the color cosmetic shade range are aggregated, e.g., summed, to produce a range reach rate.
[0073] In practice, some consumers may be counted more than once, but it is required that each consumer is counted only once. Therefore, more preferably, an aggregating and removal approach is employed, wherein the reach rates are aggregated while accounting for overlapping consumers across the plurality of color cosmetic shades. In this approach, the reach rates of a plurality of color cosmetic shades in the color cosmetic shade range are first determined. Then, the consumers predicted to find each individual shade suitable (or to find the shade suitable on each skin tone) are identified. For example, each record may be associated with (e.g. includes) a consumer identifier. These sets of consumers are combined, and crucially, any duplicate consumers are removed. Removing records with duplicated consumer identifiers ensures that each consumer is counted only once, regardless of how many shades they find suitable or how many skin tone representations they find a particular shade suitable on. For instance, as illustrated at block 216 in Fig. 2, while Shade 1 might reach 78%of light skin tone consumers and Shade 2 reaches 17%, the combined reach of both shades is 81%, not 95%, because 14%of consumers may be reached by both shades. Similarly, as illustrated at block 216 in Fig. 2, the range reach rate of Shade 1, Shade 2, and Shade 3 for the light skin tone is 90%, rather than 154%. Other approaches of calculating the range reach rate based on the individual reach rates are also possible. In the range reach rate calculation stated above, if there are records with duplicated consumer identifiers, similar removal operation may be performed such that each consumer is counted only once.
[0074] More preferably, the method further comprises calculating a total reach rate for a plurality of color cosmetic shades across a plurality of different skin tones, representing a comprehensive assessment for an entire shade range and a diverse target consumer group. This total reach rate considers both the multiple-shade aspect (like the range reach rate) and the multiple-skin-tone aspect (like the overall reach rate for a single shade) . The calculation follows the same core principle: reach rate is determined for each shade on each skin tone; sets of consumer identifiers are created for each shade / skin-tone combination; these sets are aggregated; duplicate consumer identifiers are removed; and the total reach rate is calculated by dividing the number of unique consumers by the total number of consumers in the target group.
[0075] Preferably, a distinction is made between an initial reach rate assessment and a final, more comprehensive reach rate calculation. The initial reach rate, sometimes referred to as an "inter-reach, " is typically determined based on a subset of consumers: those who have provided ratings for application colors deemed "similar" to the test shade (or shades) based on the predetermined color difference threshold. This provides a useful initial estimate. However, a more robust and comprehensive assessment, referred to as the "final reach of all consumers, " considers the preferences of all consumers within the target group, regardless of whether they have rated colors specifically identified as "similar" during the initial reach rate calculation.
[0076] For example, in an initial assessment, a particular shade might show a reach rate of 82.6%among a subset of 408 consumers who had previously rated similar colors. However, the total target consumer group (e.g., consumers from a certain country) might consist of 433 individuals, wherein 25 of these consumers had not rated colors similar to the particular test shade in the initial color evaluation. If, out of those 433 consumers, 337 are determined to find the shade suitable (based on the expanded analysis) , the "final reach of all consumers" would be 77.8% (337 / 433) . This "final reach" value, by considering the entire relevant consumer population, provides a more accurate representation of the shade′spotential appeal. This principle extends to the calculation of reach rates for entire ranges, where the final reach calculation considers all consumers, ensuring a comprehensive and robust assessment of the range′soverall appeal.
[0077] Preferably, in addition to forecasting the reach rates of individual color cosmetic shade and shade ranges, the method further comprises creating a collection of application colors, referred to herein as the "unique color library. " The unique color library is a curated set of application color representations, derived from the large scale dataset of consumer application color preferences, but significantly reduced in size and optimized for distinctiveness. The unique color library is designed to represent a wide range of perceptually distinct color cosmetic and suitable colors, while also reducing redundancy and simplifying the process of recommending inclusive shade ranges wherein the shade ranges cater to a wide and diverse spectrum of consumer skin tones and preferences. The purpose of the unique color library is twofold: to reduce computational complexity and to ensure that recommended shade ranges offer a perceptually diverse selection of suitable colors. Reference may be made to Fig. 4, which is a schematic diagram illustrating a workflow for generating a unique color library, according to one or more embodiments of the present disclosure.
[0078] The creation of the unique color library involves grouping highly similar application colors in the dataset, as illustrated at block 408 in Fig. 4. This is done to eliminate redundancy, as many colors in the full dataset may be perceptually indistinguishable from each other. The grouping process utilizes a predetermined color difference threshold, referred to as a second predetermined color difference threshold below. Preferably, this threshold is based on the CIEDE2000 color difference formula, a metric that quantifies the perceived difference between two colors. Colors from the dataset are compared pairwise, and if the CIEDE2000 color difference between two colors is less than or equal to a second predetermined color difference threshold, they are placed in the same group. Preferably, the second predetermined color difference threshold is set to a value of 1.0, as illustrated at block 408 in Fig. 4. This value is chosen because it represents a perceptually indistinguishable color difference, generally below the JND (just noticeable color difference) threshold of human perception. This grouping process results in a plurality of groups of application colors, where each group contains colors that are perceptually near identical to each other.
[0079] The creation of the unique color library further involves selecting a single, representative color from each of the groups, as illustrated at block 410 in Fig. 4. Preferably, this selection is based on the reach rate of each color within its group. The application color that has the highest reach rate within each group is selected as the representative color for that group. This ensures that the representative color is both perceptually similar to the other colors in its group and also has demonstrated appeal to consumers. Alternatively, a centroid, a median color, or a random application color data within the group can be selected as a representative color. The reach rate of the representative color of the group may be calculated as the aggregation of the reach rates of each color of the group.
[0080] Preferably, before the selection, additional filtering based on minimum reach rate is performed. For example, only colors that have reach rates above a defined minimum reach rate (e.g., >60%) and are seen by a defined minimum consumer count (e.g., >50) can be added to the unique color library, such that only high reach rate colors will be included. As shown in Fig. 4, after the grouping, selecting, and the additional filtering, about 270 colors, referred to as unique colors, are selected from the 67782 records.
[0081] The result of this method is the unique color library: a set of representative application colors that are both (1) representative of consumer preferences (due to the reach-rate-based selection and / or filtering) and (2) perceptually distinct from each other (due to the CIEDE2000 threshold) .
[0082] The unique color library is then used as the basis for recommending color cosmetic shade ranges. By limiting the selection to the colors within the unique color library, the system ensures that the recommended shades are both appealing to consumers and offer a diverse range of options. Fig. 5 is a schematic diagram illustrating a workflow for utilizing a unique color library with analytical methods for applications.
[0083] As shown in Fig. 5, the method further comprises selecting, from the unique color library, a plurality of representative colors, to maximize a range reach rate of a corresponding color cosmetic shade range. The selection from the unique color library (as illustrated at block 508 in Fig. 5, wherein example unique colors 516 are shown) may take consideration into customized requests (as illustrated at block 506 in Fig. 5) . For example, a franchise might start with an existing range of shades (as illustrated as shades 502 in Fig. 5) , represented as a set of lipstick tubes. Alternatively, the brand might specify a desired target number of shades and / or a desired ratio of color families (as illustrated as the pie chart 504 in Fig. 5) , wherein the pie chart 504 shows proportions of brown, pink, red, and purple shades. These requests represent constraints on the shade selection process.
[0084] Preferably, an analytical method is employed to identify the optimal combination of unique colors from the unique color library that maximizes the overall reach rate for the target consumer group, subject to any specified constraints. This is illustrated at block 510 in Fig. 5. As shown at application color 518, one or more preferred colors in the unique color library are identified.
[0085] Preferably, the analytical method used is a Total Unduplicated Reach and Frequency (TURF) analysis, as indicated by the "TURF" label at block 510. TURF analysis is a statistical technique used in marketing research to identify the optimal combination of products, services, or marketing messages to maximize reach and exposure within a target group of consumers. In the context of lipstick shade recommendation, in an example, TURF analysis starts from a unique color that has the highest reach rate, and iteratively evaluates all possible combinations to determine the second / third / ... color that reaches the largest number of additional consumers, with the customized requests (e.g., shown at block 506) taken into consideration. Fig. 5 shows a simplified representation of the TURF process, with overlapping circles suggesting the concept of colors reaching different segments of the consumer population. The TURF analysis continues until a stopping criterion is met, such as reaching a desired number of shades or achieving a target overall reach rate.
[0086] Furthermore, the selection process can incorporate various constraints. These constraints might include, but are not limited to, a desired number of color cosmetic shades to be in the final recommended range, a desired proportion of shades from different color families (such as reds, pinks, nudes, etc. ) , or the requirement to include one or more specific existing color cosmetic shades as part of the range. By incorporating such constraints, the method allows for the creation of shade ranges that are not only optimized for reach but also align with specific brand requirements or marketing strategies.
[0087] Fig. 5 also illustrates two example application scenarios or cases. Case 1, shown at block 512, demonstrates how the method can be used to "Keep Existing Good Shades. " In this case, the brand starts with a set of existing shades that are known to perform well. The method may first forecast the reach rate of existing shades and then identifies additional shades from the unique color library that, when combined with the existing shades, maximize the overall reach rate. The example shows that three existing shades can be supplemented with three additional recommended shades to reach 97%of consumers. The phrase "Acollection reaches X%of consumers means that X%find at least one shade suitable in the collection" is included in Fig. 5.
[0088] Case 2, shown at block 514, demonstrates how the method can be used to create a new "10-Shade Range" from scratch. In this case, the brand specifies a desired number of shades (10) and may also specify other constraints, such as a desired ratio of color families (e.g., an example ratio of 4 browns, 3 pinks, 2 reds, and 1 purple) , and / or greater emphasis on certain color families (such as browns and pinks) . The TURF analysis then identifies the 10 shades from the unique color library that, in combination, are predicted to reach the highest percentage of consumers. The example shows a 10-shade range that is predicted to reach 98%of consumers.
[0089] The output of this process is a recommended color cosmetic shade range. The recommended shades are preferably presented as CIELAB color values, allowing for precise color matching during the manufacturing process. Optionally, the shades can be simulated on a model’s lip and visualized using color-calibrated displays to provide a realistic representation of their appearance.
[0090] By leveraging this data-driven approach, brands can make informed decisions about which shades to be included in a new range, ensuring maximum appeal and minimizing the risk of launching poorly performing products. This not only improves the chances of commercial success but also promotes inclusivity by ensuring that the shade range caters to a wider spectrum of skin tones and preferences.
[0091] The benefits extend beyond product development and portfolio management. The insights generated by this disclosure can also inform marketing and communication strategies. Brands can leverage the reach rate predictions to make compelling, data-backed claims about the inclusivity and suitability of their shades. For example, a claim such as "This shade range is statistically validated to complement 97%of consumers from Country A" or "This shade is a true universal color for 98%consumers from Country A and 90%consumers from Country B" are directly supported by the reach rate forecasting capabilities of the disclosure.
[0092] Furthermore, the disclosure can be used to personalize the shopping experience. While the core focus is on group-level reach rate forecasting, the underlying data and methods can be adapted to provide individual shade recommendations. By analyzing a consumer′sskin tone and comparing it to the dataset, the system can suggest shades that are highly likely to be a good match for that individual, enhancing customer satisfaction and driving sales.
[0093] The ability to predict shade popularity also has implications for inventory management. By forecasting demand for different shades, brands can optimize their stock levels, minimizing waste and ensuring that popular shades are readily available.
[0094] Finally, the disclosure′sadaptability to different geographic regions allows for the creation of customized shade ranges tailored to the specific preferences and demographics of consumers in various markets. This level of localization can significantly enhance a brand′sglobal appeal and competitiveness.
[0095] In practice, the recommended colors may be provided as CIELAB values and, optionally, as AdobeRGB images. To ensure accurate color representation during the product development process, these color specifications should be viewed on color-calibrated displays (e.g., EIZO screens) .
[0096] In an example, the system may be deployed as an Application Programming Interface (API) for deployment as an inclusive shade recommendation tool.
[0097] Fig. 6 illustrates a block diagram of a system 600 for forecasting a reach rate of a color cosmetic shade, according to one or more embodiments of the present disclosure. The system 600 may be configured to perform the methods according to the embodiments of the disclosure. The system 600 comprises at least one memory 604 and at least one processor 602 coupled, for example by a bus 606, to the at least one memory 604. The at least one memory 604 is configured to store data and instructions, and the at least one processor 602 is configured to execute instructions to perform the operations of the method for forecasting a reach rate of a color cosmetic shade.
[0098] For example, the processor 602 accesses and executes instructions stored in the memory 604 to perform the operations of: obtaining test application color data of the color cosmetic shade when applied to a keratin material; obtaining a set of records associated with a consumer attribute from a dataset, each record including application color data and a suitability metric indicating a perceived suitability of the color cosmetic shade; identifying, within the set of records, a subset of records with application color data that fall within a first predetermined color difference threshold of the test application color data; and calculating the reach rate as m / n, where n is a total number of records within the subset of records, and m is the number of records, within the subset of records, that have a suitability metric exceeding a predetermined suitability threshold.
[0099] The memory 604 may store the dataset, sets of records and its subsets, test application color data, intermediate data generated during the calculations, and the calculated reach rate, among other data required for the system′soperation. The memory may also store the unique color library, if used.
[0100] It is understood that the system 600 may further comprise other components not explicitly shown in Fig. 6, such as input / output interfaces for receiving data (e.g., image data from a camera, consumer data from a dataset) and displaying results (e.g., reach rate predictions, recommended shades) , network interfaces for communication, and user interfaces for user interaction. These components would be readily understood by one skilled in the art. The system may include a display for displaying, for example, the predicted reach rate.
[0101] Fig. 7 illustrates a block diagram of an example device 700 for forecasting a reach rate of a color cosmetic shade, according to one or more embodiments of the disclosure. The device 700 is a functional apparatus, implemented for example as a dedicated hardware device or as a software module executing on a computing device.
[0102] The device 700 comprises test application color data obtaining module 702. The test application color data obtaining module 702 is configured to perform the operation of obtaining test application color data of the color cosmetic shade when applied to a keratin material as described above. In an example, the application color obtaining module 702 may include an interface to receive test application color data from an external source, such as a camera system or a data storage unit. In another example, it may include components to perform the test application color data generation internally, such as image processing units and color calculation units.
[0103] The device 700 further comprises a record obtaining module 704. The record obtaining module 704 is configured to perform the operation of obtaining a set of records associated with a consumer attribute from a dataset, as described above. The record obtaining module 704 may be configured to retrieve or generate a set of records associated with a specific consumer attribute from the dataset, where each record includes application color data and a suitability metric.
[0104] The device 700 also comprises an identifying module 706. The identifying module 706 is configured to perform the operation of identifying, within the set of records, a subset of records with application color data that fall within a first predetermined color difference threshold of the test application color data.
[0105] The device 700 further comprises a reach rate calculation module 708. The reach rate calculation module 708 is configured to calculate the reach rate based on the subset of records identified by the identifying module 706. The calculation is performed as m / n, as described above, where m is the number of records in the subset with suitability metrics exceeding a predetermined threshold, and n is the total number of records in the subset.
[0106] The reach rate obtained by the reach rate calculation module 708 is outputted by the device 700. The device may further include an output interface (not shown) to display or transmit the reach rate to a user or another system. The present disclosure also discloses a non-transitory computer-readable medium storing instructions that, when executed by a processor, cause the processor to perform operations of the method for forecasting a reach rate of a color cosmetic shade, and other operations described herein. The computer-readable medium may be any suitable medium for storing computer-executable instructions, such as a magnetic disk, optical disk, semiconductor memory device, or other storage medium.
[0107] While the preferred embodiments described above focus on specific implementations of the disclosure, various alternative embodiments are also within the scope of the present disclosure.
[0108] For example, while the preferred embodiments focus on skin tone as the primary segmentation variable, other factors, such as age of consumer, country of consumer, could also be used. For example, consumer preferences for application colors often vary significantly between different age groups or different countries. The methods described herein could be adapted to calculate reach rates for specific age groups (e.g., 18-25, 26-40, 41-55, 56+) or a specific country by categorizing consumer records based on age or country, in addition to or in place of skin tone. Further, the present disclosure can apply TURF analysis to recommend shades for various consumer groups, for instance, recommending shades for specific age groups, for a specific country, etc.
[0109] In conclusion, the present disclosure provides a novel and effective method and system for forecasting a reach rate of a color cosmetic shade. By leveraging a large dataset of consumer preferences, a carefully calibrated pseudo-spectral camera system, and a robust data-driven approach, the disclosure overcomes the limitations of traditional methods and offers significant advantages in terms of accuracy, efficiency, and inclusivity. The system can be used to improve both the shade development process for cosmetic companies and the color cosmetic selection experience for consumers.
[0110] It is to be understood that the present disclosure may be implemented in various forms of hardware, software, firmware, special purpose processors, or a combination thereof. For example, the functions described herein as being performed by the "processor" may be performed by a single processor, multiple processors, a digital signal processor (DSP) , an application-specific integrated circuit (ASIC) , a field-programmable gate array (FPGA) , a general-purpose computer, or any other combination of hardware and / or software capable of performing the specified functions. The present application may be implemented as a computer program product for performing the functions or operations described herein. The present application may be implemented as a system including means for performing the functions or operations described herein.
[0111] While the embodiments have been illustrated and described herein, it will be understood by those skilled in the art that various changes and modifications may be made, and equivalents may be substituted for elements thereof without departing from the true scope of the present technology. In addition, many modifications may be made to adapt to a particular situation and the teaching herein without departing from its central scope. Therefore, it is intended that the present embodiments not be limited to the particular embodiment disclosed as the best mode contemplated for carrying out the present technology, but that the present embodiments include all embodiments falling within the scope of the appended claims.
Claims
1.A method for forecasting a reach rate of a color cosmetic shade, the method comprising:obtaining test application color data of the color cosmetic shade when applied to a keratin material;obtaining a set of records associated with a consumer attribute from a dataset, each record including application color data and a suitability metric indicating a perceived suitability of the color cosmetic shade;identifying, within the set of records, a subset of records with application color data that fall within a first predetermined color difference threshold of the test application color data; andcalculating the reach rate as m / n, wheren is a total number of records within the subset of records, andm is a number of records, within the subset of records, that have a suitability metric exceeding a predetermined suitability threshold.2.The method of claim 1, wherein the consumer attribute is a skin tone, the method further comprising:calculating the reach rates of the color cosmetic shade for a plurality of different skin tones; andcalculating a group reach rate of the color cosmetic shade for the target group based on the reach rates of the color cosmetic shade for the plurality of different skin tones.3.The method of claim 1 or 2, wherein the consumer attribute is at least one of age of consumer or country of consumer.4.The method of any of claims 1-3, further comprising:calculating the reach rates of a plurality of color cosmetic shades in a color cosmetic shade range; andcalculating a range reach rate of the color cosmetic shade range by aggregating the reach rates while accounting for overlapping consumers across the plurality of color cosmetic shades.5.The method of any of claims 1-4, wherein obtaining the test application color data of the color cosmetic shade when applied to a keratin material comprises:applying the color cosmetic shade to the keratin material;capturing image data of the color cosmetic shade as applied to the keratin material; andprocessing the image data to determine the test application color data of the color cosmetic shade.6.The method of claim 5, wherein processing the image data comprises determining the test application color data as a median color value from a plurality of pixel color values within a detected application region in the image data.7.The method of any of claims 1-6, wherein the test application color data is represented in CIELAB color space, and / or wherein the first predetermined color difference threshold is based on the CIEDE2000 color difference formula.8.The method of any of claims 1-7, further comprising creating a unique color library, wherein creating the unique color library comprises selecting a representative color from each of a plurality of groups of application color data to be included in the unique color library.9.The method of claim 8, wherein the plurality of groups of application color data are formed by grouping application color data from the dataset based on a second predetermined color difference threshold.10.The method of any of claims 8-9, wherein selecting the representative color further comprises selecting application color data that has the highest reach rate within each of the plurality of groups of application color data as the representative color.11.The method of any of claims 8-10, further comprising:selecting, from the unique color library, a plurality of representative colors, to maximize a range reach rate of a corresponding color cosmetic shade range, based on a constraint, preferably, wherein the constraint specifies at least one of: a desired number of color cosmetic shades in the color cosmetic shade range, a desired proportion of color cosmetic shades from different color families, and one or more existing color cosmetic shades that must be included in the color cosmetic shade range, more preferably, the selecting is performed using a Total Unduplicated Reach and Frequency (TURF) analysis.12.The method of claim 11, further comprising providing the plurality of color cosmetic shades corresponding to the plurality of representative colors as a recommendation.13.The method of any of claims 1-12, the color cosmetic is selected from lip color products, face color products, hair color products, and nail color products, preferably selected from lipsticks, lip glosses, lip stains.14.A system for forecasting a reach rate of a color cosmetic shade, the system comprising:at least one memory; andat least one processor coupled to the at least one memory, wherein the at least one processor is configured to perform operations of the methods of any one of claims 1-13.15.A non-transitory computer-readable medium storing instructions that, when executed by a processor, cause the processor to perform operations of the methods of any one of claims 1-13.
