Business Challenge

A popular D2C fashion brand faced major challenges with:

• Frequent stockouts of trending styles.
• Overstock of less-popular SKUs.
• Delayed design changes based on slow sales data.

A client partnered with Datazivot to carry out Myntra Product Reviews Scraping to gain actionable insights from customer feedback across various fashion categories. The goal was to analyze user reviews at scale and uncover trends in preferences and satisfaction levels.

The review scraping focused on the following categories:

• Women’s ethnic wear (kurtas, sarees)
• Men’s footwear
• Casual tops and dresses

Their goal: use customer sentiment and emerging keyword trends to forecast fashion trends before peak demand.

Objectives

Datazivot’s mandate:

• Scrape customer reviews, star ratings, and keywords from 10,000+ Myntra products.
• Perform sentiment analysis and detect fashion preferences.
• Identify seasonal style trends based on rising positive mentions (colors, fits, fabrics).
• Help the client adjust marketing, design, and inventory decisions based on data—not guesswork.

Our Approach

1. Review Scraping at Scale

We built a custom crawler to collect:

• Review text, star rating, product title, and category
• Timestamps and seasonal tags (if available)
• Color/style/fit mentions (from product title and review body)

Platforms:

• Myntra (Primary)
• Cross-check with Amazon Fashion for trend overlap

Dataset Size: 220,000+ reviews from Jan to Oct 2024

Refresh Frequency: Bi-weekly scraping during festive/seasonal periods

2. Sentiment + Keyword Pattern Analysis

We processed data with:

NLP models for sentiment classification (positive, neutral, negative)

Custom phrase extractors for style mentions:

• Style: "flared", "slim-fit", "A-line", "oversized"
• Color: "pastel", "mustard", "lavender", "neon"
• Fabric: "cotton", "rayon", "linen", "silk"

We tracked month-on-month keyword growth to flag rising trends.

Sample Data Snapshot

Key Insights & Outcomes

1. Trend Spotting Before Sales Data Catches Up

• “Pastel” kurtas surged in positive reviews in July—5 weeks before they peaked in sales.
• “Lavender” and “neon green” saw rising mentions for Gen Z styles.
• “Cotton-linen blends” had consistent praise for comfort in summer.

2. Negative Sentiment Flags for Fit

• “Slim-fit” men’s shirts had the highest negative ratio (33%), mostly due to sizing mismatch.
• “Free-size” ethnic wear received neutral to negative reviews citing lack of shape.

Helped the brand avoid overstocking these variants for Diwali season.

3. Festive Surge Prediction

• Based on October review patterns, we predicted a 15–20% uptick in embroidered and silk SKUs for Diwali.
• This led to faster production timelines and smart inventory placement in North India zones.

Visual Dashboard Samples

Monthly Positive Keyword Trends (Top 5 – July to October 2024)

Sentiment by Style (Q3 2024)

• A-line Dresses: 79% positive
• Slim-fit Shirts: 57% positive
• Flared Kurtas: 82% positive

Tech Stack Used

• Scraping: Python (Selenium + BeautifulSoup), rotating proxies
• NLP: spaCy, TextBlob, BERT for sentiment + custom regex for keyword parsing
• Trend Modeling: Time series forecasting in Python (Prophet + ARIMA)
• Visualization: Power BI + Google Looker Studio

Business Impact

• +22% Faster Trend Adoption:

The client launched trending styles 3–4 weeks earlier than competitors during the festive season.

• -18% Overstock on Low-Rated SKUs:

Poorly-reviewed designs were discontinued early based on negative review volume.

• +31% Inventory Optimization:

Data-driven fashion trend forecasting allowed improved warehouse planning across regional distribution centers.

• New Design Feedback Loop:

Designers used real-time review dashboards to modify fit, fabric, and patterns before mass production.

Conclusion

This case study proves that reviews are more than feedback—they're fashion signals.

By mining and analyzing Myntra reviews at scale, Datazivot enabled a fashion brand to move from reactive retail to proactive trend leadership.