Predicting Consumer Tastes using Web Data Analysis for Gap Inc.

This article is based on the academic project submitted by MS BAIM graduate students at Purdue University. The case study is referred from Harvard Business School publications (9–517–115) authored by Ayelet Israeli and Jill Avery in 2018.

Overview

In the evolving landscape of fashion retail, the balance between artistic intuition and data-driven decision-making is becoming increasingly critical. Traditionally, the fashion sector has relied on the creative insight of designers to predict trends. Yet, the rise of big data is challenging this norm. Gap, a renowned entity in fashion retail, underscored this shift when the then CEO Art Peck prioritized data-driven strategies over conventional creative direction in 2017.

The central question is the efficacy of data in predicting consumer preferences compared to traditional creative methods. Our team conducted an in-depth analysis to address this, leveraging advanced web data analytics. We meticulously examined diverse data sources, from customer feedback to sentiment analysis on platforms like Reddit and sales metrics from Google Shopping. We aimed to comprehensively understand the digital presence of these brands and assess the potential impact of data on their strategic direction.

Primary Objective

The primary objective is to assess the implications of the then CEO, Peck’s transition to a data-centric strategy. Our project objective broadly expands into the following three aspects,

• Tapping into web data to shed light on Gap’s evolving direction, capturing insights on contemporary clothing trends and fashion inclinations.

• Conducting an in-depth analysis of all three brands of GAP Inc. — Old Navy, Gap, and Banana Republic by understanding their digital footprint and the role of big data in their strategies.

• Exploring data-driven methods and traditional creative approaches in marketing, aiming to identify the most effective strategy for today’s fashion retail scenario.

Analysis

Our analysis shows that the big data approach has enormous potential to drive GAP’s business ahead of competitors. Below are six methods (3 qualitative & 3 quantitative), where we have worked on live datasets and derived some valuable insights, as shown below. These methods can be prototype examples of data-driven techniques to understand customer preferences.

I. Qualitative Analysis

Method 1 — Customer Feedback Analysis

This method can be a simple yet effective data-driven technique to discover fashion trends based on GAP’s customer reviews.

1. Collected customer feedback from the website Customer Affairs

2. Utilized the GPT-3.5 model for sentiment analysis to identify popular fashion trends in the feedback

3. Provided Gap with valuable insights into customer preferences to align their clothing offerings with trends. Below is the one of the illustrations of our qualitative insight

Method 2— Google Trends Analysis

This method will help GAP understand top trends across ten clothing categories broadly used across the US retail market. Key highlights of the analysis are given below,

1. Identifying the best keywords that fit into the GAP Inc. category based on GAP’s listing on Amazon

2. Using the identified categories as keywords to collect data from Google trends

3. Analyzing the four trending parameters — Trending queries, Most searched queries, Most searched topics, and top state searching for keywords.

In this way, GAP Inc. can use Google Trends to analyze the top trends for each clothing category to stay ahead of the competition in identifying the products that can be in demand.

Method 3— Competitor Analysis on Macy’s website

This method can help GAP monitor the competing brands' pricing strategies, trends, best-selling products, and promotions.

1. Use Selenium to extract content from competitors’ websites, including paragraphs, headers, links, spans, articles, etc.

2. Employee ChatGPT, LLM to analyze the extracted content and provide insights into the latest fashion trends and popular items

   
   

All three brands can efficiently combine web parsing and AI-powered natural language processing to stay updated on trends within the fashion industry.

II. Quantitative Analysis

Method 1 — Insights from 3rd party trading channels like Amazon

• Sales data from the direct partner, Amazon, is extracted to find out the most sold item, category, and how the customer loved the item

• This Amazon data can also be used in further stages to validate our prediction model to identify trends or customer preference

• Since Gap’s sales are predominant in Amazon, this data can also be an excellent source to analyze the brand-wise preferences and their strengths for the products listed on Amazon

Method 2 — Sentiment analysis of the customer using Redditt data

This method is used to understand the consumer sentiments of the consumers from one of the famous fashion reviews on Redditt's page — “malefashionadvice”

1. Collect post descriptions from the Reddit “malefashionadvice” subreddit, specifically targeting GAP brands

2. Conduct sentiment analysis using Azure API post assigning scores to specific keywords

3. Aggregate sentiment scores to identify distinct brand-specific trends for each of the three brands

Method 3— Building regression model based on Google shopping

This method effectively highlights how factors like price, discount, promotions, no. of reviews, store category, seller, and product category influence the product rating, assuming GAP Inc. has incorporated a big data strategy. This method can also evaluate whether the same big data strategy works for all three GAP brands. Below is a simple flow chart of the approach.

Applications: Role of Big data in GAP’s marketing strategy

By integrating data-driven insights with existing creative interpretation, Gap can effectively balance the art of emotional connection with customers and the science of precision and efficiency in marketing efforts. Following our analysis and derivation of helpful market insights, prioritizing a science-based approach rather than just sticking to traditional creative instincts seems advantageous to GAP.

Conclusion

In summary, the fashion retail industry is evolving, emphasizing the blend of creative intuition and data-driven decision-making. Our analysis explored the data’s effectiveness in predicting consumer preferences versus traditional methods. Key findings highlight leveraging tools like Selenium, ChatGpt, and Pytrends offer valuable insights. Data analysis across these brands underscores the potential of big data strategies.