IS4242 Intelligent Systems & Techniques L9 - Branding PDF

IS4242 INTELLIGENT SYSTEMS & TECHNIQUES L9 – Branding Aditya Karanam © Copyright National University of Singapore. All Rights Reserved. ...

IS4242 INTELLIGENT SYSTEMS & TECHNIQUES L9 – Branding Aditya Karanam © Copyright National University of Singapore. All Rights Reserved.  Announcements ▸ Quiz-1: Grades will be released by Friday ‣ Email: Simian (or me) regarding the grade if you find any discrepancies ▸ Midterm Feedback IS4242 (Aditya Karanam) 2  In this Class… ▸ What exactly is Brand? ▸ What is the Value of Brand to the Firm? ▸ How to derive Competitive Advantage through Branding? ▸ How to Create and Manage a Brand? ▸ Measuring Brand Image among consumers ‣ Convolution Neural Networks: Convolution and Pooling ‣ Hyperparameter Tuning IS4242 (Aditya Karanam) 3  What is brand? Traditional View ▸ The term “branding” comes from the practice of branding cattle ‣ Ranchers etched an identifying mark on animals to claim ownership ▸ Similarly, in the early days of market development, firms used brand marks to differentiate their products ‣ Some of these were merely the signature of the maker, while others were graphical representations and symbols ▸ Markers allowed consumers to differentiate between the products of one producer and another ‣ Identifying producers with whom they were familiar and trusted IS4242 (Aditya Karanam) 4  Brand: Culture of the Product ▸Contemporary view considers brand as the culture of the product. ‣Products are considered as cultural artifacts (borrowing the term from anthropology, sociology and history) ▸Products acquire meanings — connotations — as they circulate in society ▸Overtime, these meanings become conventional, widely accepted as “truths” about the product ‣Thus, creating a culture of the product, which is referred to as its brand IS4242 (Aditya Karanam) 5  Brand: Culture of the Product ▸ Consider a new product that has just been introduced by a new company. ▸ The product has a name and a trademarked logo, and other unique design features—all aspects that we intuitively think of as “the brand” ▸ However, the brand does not yet exist! ‣ Names, logos and designs are the material markers of the brand ‣ The product does not yet have a history – these markers are “empty” or devoid of meaning IS4242 (Aditya Karanam) 6  Brand: Culture of the Product ▸ Now think of some famous brands. ▸ These brands also have markers: Name, logo, distinctive product design feature, or any other design element that is uniquely associated with the product. ▸ These markers have been filled with customer experiences, advertisements, films and sporting events that displayed the brand in a popular culture, etc. ▸ These ideas of the product accumulate over time and “fill up” the brand markers with meaning. ‣ A brand culture is formed. IS4242 (Aditya Karanam) 7  Brand as Identity Markers of Consumers ▸ Earlier, brands played an informational role, adding value beyond its functional utility by serving as a heuristic, easing the decision-making process ‣ Thus, consumers pay extra for Heinz ketchup because they have learned to trust the Heinz brand and the company that stands behind it. ▸ In today’s culture, brands are the means to express personalities, lifestyles, ideologies, and a variety of other social identities. ‣ Brands are used by consumers as identity markers – props to construct and maintain their identities. ‣ This has created loyal customers following the culture of the product and forming communities – brand communities IS4242 (Aditya Karanam) 8  Measuring Brand Value: At a Product Level ▸ Brand equity is a measure of the brand’s value to the firm ‣ Sum of the intangible assets and liabilities linked to a brand. ▸ At a micro-level (product level), brand equity is the differential response consumers have to products when they are branded ▸ Offer consumers two identical products, one unbranded and one branded ‣ Brand equity is the consumer’s price premium: how much more consumers are willing to pay for the branded product over the unbranded product ‣ Value of the brand is the added value the brand brings during the purchase IS4242 (Aditya Karanam) 9  Brand Equity: At a Firm level ▸ On a macro-level (firm level), brand equity is measured as additional profits obtained from the sale of branded products over the sale of the same products unbranded ‣ At a firm level, brand equity drives not only a revenue premium but also cost savings ▸ Products with strong brands have lower marketing costs as consumers are more receptive and responsive to the brand’s messages IS4242 (Aditya Karanam) 10  Brand Equity: Dynamic and Difficult to Measure ▸ Brands, as assets, can both increase and decrease in value over time. ‣ Negative news about your brand can decrease its asset value ▸ Calculating the asset value of a brand is quite difficult ‣ Difficult to separate out the intangible value associated with their products from the tangible value derived from functional attributes of the product ▸ In most countries, brands do not appear on financial statements as assets unless they are purchased from another firm IS4242 (Aditya Karanam) 11  Branding as Competitive Advantage ▸ Companies can leverage brand equity in several ways ▸ The simplest is to create extensions ‣ New versions of the product within the same product category or extending the brand into other product categories. ‣ Example: Dyson has a good brand in vacuum cleaners and extended further into hair dryers IS4242 (Aditya Karanam) 12  Branding as Competitive Advantage ▸ Firms can also leverage their brand through co-branding, franchising, etc. ‣ Co-branding example: Prada and Adidas ‣ Several franchises in apparel, food, etc. IS4242 (Aditya Karanam) 13  Branding: Creation of Brand Culture ▸Brand cultures are created through stories from various “authors” ▸Companies: Shaped by all product-related activities of the firm that touch customers ‣ E.g.: All elements of the marketing mix: product, communication, pricing, etc. “tell stories” about the product. ▸Popular culture: Products are frequently used in media: films, television, books, etc. These representations can have a powerful influence on brands. ‣ E.g.: Ian Fleming - “Bond had been offered the Aston Martin (DB III) or a Jaguar 3.4. Either of the cars would have suited his cover – a well-to-do, adventurous young man with a taste for the good, fast things in life. But DB III had the advantage of an inconspicuous colour – battleship grey – and certain extras which might or might not come in handy.” ‣ Describes the image of Jaguar and Aston martin IS4242 (Aditya Karanam) 14  Branding: Creation of Brand Culture ▸Customers: Create consumption stories involving the product, which they often share with friends ‣Example: People share photos of the food they consume, pictures of themselves with brands, share their experience with the firm on social media, and so on ▸Influencers: In many categories, noncustomers’ opinions are influential. ‣Example: Product reviews on YouTube, Influencer Posts on TikTok, Instagram, etc. ▸Stories created by all of these “authors” interact in complex ways IS4242 (Aditya Karanam) 15  Branding: Creation of Brand Culture ▸ Customers watch ads and listen to influencers as they use the product. ▸ The media monitor how customers use the product and consider this in how they represent the product ▸ Quantity and complexity of these interactions makes it difficult to isolate the influence of each author The Firm Popular Culture Brand Stories Brand Culture Brand Shared, taken-for-granted brand stories, Brand Stories images, and associations Stories Brand Stories Influencers Customers IS4242 (Aditya Karanam) 16  How do you Manage Brand? ▸ Know your brand: The first step is always to understand your brand’s image in the minds of consumers ▸ Analyze what consumers have been hearing and interpreting about your brand (your received brand image) ‣ NOT what you have been saying in your brand communications (your intended brand identity) ▸ Understanding how your brand is currently perceived by consumers offers you insight into how to craft your brand messages going forward IS4242 (Aditya Karanam) 17  How do you Manage Brand? ▸ Brand management is image or personality management » ▸ Radically changing the meaning of a brand when consumers are using that meaning in their lives can be detrimental ‣ Ex: Twitter (X): Changes in name and logo ‣ Bird icon was friendly and welcoming, unclear about the meaning of ‘X’ – modern and edgy? ‣ This led to massive spike in Twitter Lite installs in the first week of the rebranding. IS4242 (Aditya Karanam) 18  How do you Manage Brand? ▸ Manage your brand’s authors: Meaning is authored by many people other than the brand’s managers ‣ Authors include consumers, media, influencers, etc. ▸ As a result, managers need to consider who the prominent authors of their brands are ▸ Work to positively influence those authors to compose and deliver desired brand messages IS4242 (Aditya Karanam) 19  Image-Based Brand Image: Measuring Brand from Images Technique: Convolution Neural Networks © Copyright National University of Singapore. All Rights Reserved. 20  How to Know Brand Image? ▸ Simply conduct survey of the customers ▸ More accurate: Listen to consumer experiences through social media such as Instagram, Twitter, etc. ▸ Consumers share photos and often tag brands in their posts ‣ Consumers communicate about brands with each other using these images ‣ For example, a search of #nike on Instagram returns over 92 million photos tagged with the brand name Nike (in 2023). IS4242 (Aditya Karanam) 21  Consumer Experiences: Example ▸ In the images posted on social media, consumers often link brands with usage context, feelings, and consumption experiences. ▸ Understand the brand image by calculating average consumption experiences in the consumer images IS4242 (Aditya Karanam) 22  Data Mining Task 1. Identify the consumption experience of apparel and beverage brands ‣ Classes: Glamor, Fun, Healthy, and Rugged ‣ Supervised learning problem ‣ Consider the task as image classification: 4 Class Classification problem 2. Calculate the average experience of all the consumers in a given brand ▸ How to obtain the data? ‣ Data with class labels for training the model ‣ Consumer data related to the brands IS4242 (Aditya Karanam) 23  Data ▸ Data for model building: ‣ Obtain tags from mTurk Annotators ‣ Build an image classifier to predict the experience ▸ Data on apparel and beverage brands: ‣ Identify brand images using the brand hashtags in the post on Instagram ‣ Brands: Adidas, Nike, Gucci, Levi's, Coca-Cola, Pepsi, Fanta, etc. IS4242 (Aditya Karanam) 24  Image Classification ▸ Can you identify the consumer experiences in these images? ▸ The human visual system is incredibly good: the hallmark of intelligence ▸ It is extremely difficult to make a rule-based computational system that recognizes experiences in the images. IS4242 (Aditya Karanam) 25  Image Classification: ML Approach ▸ Use a large number of labeled images to train a classifier Fun Rugged IS4242 (Aditya Karanam) Glamor Healthy 26  Image Representation ▸ Each image is a matrix of pixel values. ▸ Greyscale image: single value per pixel, 0.0 represents black, 255 represents white, and intermediate values have darkening shades of grey. ‣ E.g.: MNIST data: 28 x 28 pixel images ▸ RGB image: Each pixel is represented by three integers ranging from 0 to 255, reflecting the contribution of each of the three (red, green, blue) color channels ‣ Our dataset: 227 pixels wide and 227 pixels tall in each dimension: 227 x 227 x 3 ▸ Represent each image by a single dimensional vector of RGB values, i.e., a vector of 154,587 (=227 X 227 x 3) numbers. IS4242 (Aditya Karanam) 27  Neural Networks ▸ Example: Multi-label classifier Fun 154,587 vector Rugged Healthy Glamour IS4242 (Aditya Karanam) 28  Deep Neural Networks In the recent decade 1980s (deep: number of hidden layers) ▸ Increasing number of hidden layers ⇒ Increasing number of weights to be learnt ⇒ Optimization (Training) becomes harder IS4242 (Aditya Karanam) 29  Recap: Feed Forward Neural Network ▸ Neural Network Architecture ‣ Different layers of neurons ‣ Different types of activation functions ▸ Loss Function: Determined by the task ▸ Training: Backpropagation using gradient descent ‣ Learning rate and initialization matter ▸ CNN: Major change is in the architecture, especially feature mappings IS4242 (Aditya Karanam) 30  Convolution Neural Networks (CNN) ▸ Brief overview with intuition on the salient aspects of CNN ▸ Similar to Neural Networks but utilizes spatial correlations in grid-like data (e.g. images) ▸ Reduces the number of weights through convolution and pooling operations and parameter sharing (tied weights) IS4242 (Aditya Karanam) 31  Convolutional Neural Networks History: Alexnet ▸ Alexnet: CNN designed by Krizhevsky, Sutskever and Hinton., won the ImageNet Large Scale Visual Recognition Challenge, 2012. ‣ Marks the rise of deep neural networks IS4242 (Aditya Karanam) 32  Image Net Challenge ▸ Other architectures have improved performance further. IS4242 (Aditya Karanam) 33  Convolutional Neural Networks ▸ CNN can be considered to have two components: Feature Mappings and Classifier layers ▸ Feature mappings: Convolution, Activation functions (ReLU (max(0, 𝑥))) and Pooling ▸ Classifier: Fully connected layers (neurons) with some activation function IS4242 (Aditya Karanam) 34  Convolution ▸ A convolution operation uses a filter/kernel that ‘slides’ along the input volume and transforms it. ‣ The same weights are used for the entire input volume ▸ Stride: the step-size as the filter moves through the image (default value is 1) IS4242 (Aditya Karanam) 35  Convolution ▸ Padding: We may utilize pixels less on the perimeter of images compared to the one in the middle ▸ Padding the image with 0’s in the border ‣ Performed especially in the final layers IS4242 (Aditya Karanam) 36  Pooling ▸ Pooling transforms the input volume by resizing it using operations such as min, max, avg., etc. ‣ E.g., Max pooling takes the maximum of every block IS4242 (Aditya Karanam) 37  Convolutional Neural Networks ▸ Several convolutions are applied to the same image to extract different types of features ▸ Flatten the final layer of convolution ▸ Provide it as an input to the fully connected layer to obtain class IS4242 (Aditya Karanam) 38  Application: Consumer Experience ▸ 5 convolution layers and Max Pooling ‣ Activation function: ReLU ▸ Two fully connected layers for classification ▸ Efficiently implement using PyTorch library IS4242 (Aditya Karanam) 39  Application: Consumer Experience ▸ ConvNet with 5 layers and two fully connected layers Feature mapping Classifier IS4242 (Aditya Karanam) Included to highlight the architecture exact dimensions are in the code (prev. slide) 40  Calculating Image-Based Brand Image ▸ Obtain average probability of each class across all the images of a brand shared by users and sponsors ▸ Compare the probabilities to understand the brand image among consumers and the brand image that the firm wishes to portray Users’ posts Sponsored posts IS4242 (Aditya Karanam) 41  Hyper-Parameter Tuning ▸ Identify the optimal values of the parameter for the model ▸ 1. Create 3 splits of your data: Train(70%), Validation(10%) & Test Sets (20%) IS4242 (Aditya Karanam) 42  Hyper-Parameter Tuning ▸ 2. Use different hyperparameter values and train different models using Training Data IS4242 (Aditya Karanam) 43  Hyper-Parameter Tuning ▸ 3. Evaluate performance of each model on Validation Set and choose the best hyper parameters ▸ One validation set may not be representative of the data ‣ Combine training and validation data, and partition them into ‘k’ folds IS4242 (Aditya Karanam) 44  Hyper-Parameter Tuning ▸ K-Fold Cross Validation ▸ Divide the entire data into K folds ▸ In each fold, train on the training set and evaluate on the validation set ▸ Average the performance over all the folds Estimate of the “true” test error IS4242 (Aditya Karanam) 45  Hyper-Parameter Tuning ▸ Using the best Hyperparameter values, train the model on Training + Validation Set to learn a new model IS4242 (Aditya Karanam) 46  Hyper-Parameter Tuning ▸ Grid SearchCV ‣ Evaluates every possible combination of hyper-parameter values along with cross validation ▸ Randomized search CV ‣ Random samples of hyperparameter values are evaluated ‣ Optimizes for time ▸ Existing packages (sklearn) perform these efficiently IS4242 (Aditya Karanam) 47  References ▸ Churn management: ‣ https://hbsp.harvard.edu/product/509041-PDF-ENG ‣ Liu et al. 2022, Visual Listening In: Extracting Brand Image Portrayed on Social Media, Marketing Science ▸ Neural Networks: ‣ http://cs231n.github.io ‣ http://neuralnetworksanddeeplearning.com/index.html ‣ http://www.deeplearningbook.org ▸ Hyper-Parameter Tuning: ‣ https://sebastianraschka.com/blog/2016/model-evaluation-selection- part1.html IS4242 (Aditya Karanam) 48  Thank You © Copyright National University of Singapore. All Rights Reserved.