Build an India Fashion Trend Pipeline with Myntra (2026)

Thirdwatch's Myntra Scraper lets you turn India's largest fashion marketplace into a longitudinal trend dataset — weekly snapshots across 20-30 categories, indexed on colour, silhouette, brand and price-band. Built for fashion researchers building seasonal forecasts, founders sourcing private-label inventory, marketplace analysts benchmarking assortment, and brand teams tracking competitor catalogue ramps.

Why build a trend pipeline on Myntra

Myntra is the leading-edge of India fashion assortment. According to Statista's 2024 India online fashion report, Myntra is the country's largest dedicated fashion marketplace by GMV, with 6,000+ brands and several million live SKUs across apparel, footwear, beauty and accessories. Brands stage new-season inventory on Myntra before pushing to AJIO, their own DTC sites, or offline retail — typically 4-8 weeks ahead. For trend research, that staging window is the entire value proposition: you can see what's about to be commercially important rather than what already is.

The job-to-be-done is structured. A fashion researcher tracks colour-and-silhouette trends across men's, women's and kids' apparel for a seasonal forecast deck. A founder evaluating a private-label launch needs price-band density per category to size the white-space opportunity. A marketplace ops team benchmarks assortment depth versus Myntra to plan category investment. A brand team tracks competitors' new-season ramps in near-real-time. All reduce to weekly longitudinal capture of brand, colour, article_type, price, sizes and rating_count across a fixed category list, then trend computation on top.

How does this compare to the alternatives?

Three options for building an India fashion trend dataset:

Official reports are the right macro context layer but they don't operate at SKU granularity or weekly cadence. The Myntra Scraper actor page gives you the underlying records to build your own trend layer — see the e-commerce scraping guide for adjacent patterns.

How to build a Myntra trend pipeline in 4 steps

Step 1: How do I structure the weekly category sweep?

Pull a fixed category list every Monday at 09:00 IST. Use Apify's scheduler or your own cron. Keep the input shape boring and deterministic for downstream stability.

export APIFY_TOKEN="apify_api_xxxxxxxxxxxxxxxx"

import os, requests, datetime, json, pathlib

ACTOR = "thirdwatch~myntra-scraper"
TOKEN = os.environ["APIFY_TOKEN"]

CATEGORIES = [
    "men-tshirts", "men-shirts", "men-jeans", "men-trousers", "men-footwear",
    "women-tops", "women-dresses", "women-kurtas-kurtis", "women-jeans",
    "women-footwear", "women-handbags",
    "ethnic-wear", "sportswear", "beauty-personal-care",
    "accessories", "bags-backpacks",
]

resp = requests.post(
    f"https://api.apify.com/v2/acts/{ACTOR}/run-sync-get-dataset-items",
    params={"token": TOKEN},
    json={
        "queries": CATEGORIES,
        "sortBy": "popularity",
        "maxResults": 300,
    },
    timeout=1800,
)
records = resp.json()
week = datetime.date.today().isocalendar()
stamp = f"{week.year}-W{week.week:02d}"
pathlib.Path("trends").mkdir(exist_ok=True)
pathlib.Path(f"trends/myntra-{stamp}.json").write_text(json.dumps(records))
print(f"{stamp}: {len(records)} records across {len(CATEGORIES)} categories")

16 categories × 300 products = ~4,800 records per weekly pull. Over 12 weeks, that's ~60K records — a usable sample size for India fashion trend analysis at the SKU level.

Step 2: How do I compute a colour trend index?

Aggregate listings by (week, article_type, primary_colour) and compute the share of listings per article_type. Week-over-week share growth is the trend signal.

import pandas as pd, glob

frames = []
for f in sorted(glob.glob("trends/myntra-*.json")):
    stamp = pathlib.Path(f).stem.replace("myntra-", "")
    for j in json.loads(pathlib.Path(f).read_text()):
        frames.append({
            "week": stamp,
            "sku": j.get("sku"),
            "brand": j.get("brand"),
            "article_type": j.get("article_type"),
            "primary_colour": j.get("primary_colour"),
            "price": j.get("price"),
            "rating_count": j.get("rating_count") or 0,
        })

df = pd.DataFrame(frames).dropna(subset=["article_type", "primary_colour"])

colour = (
    df.groupby(["week", "article_type", "primary_colour"])
      .size().reset_index(name="listings")
)
totals = colour.groupby(["week", "article_type"])["listings"].transform("sum")
colour["share"] = colour.listings / totals

# Week-over-week share growth per (article_type, colour)
colour = colour.sort_values(["article_type", "primary_colour", "week"])
colour["share_wow"] = (
    colour.groupby(["article_type", "primary_colour"])["share"].pct_change()
)

trending = colour[
    (colour.share_wow >= 0.20)
    & (colour.listings >= 30)
].sort_values("share_wow", ascending=False)
print(trending.head(20))

A 20%+ week-over-week growth in listings share, with at least 30 listings to filter noise, is a robust trending-colour signal. Cross-check against brand breadth in step 3 to filter single-brand bets.

Step 3: How do I separate real trends from single-brand bets?

Compute brand breadth per trending colour. A real trend has many brands ramping; a single-brand bet has one.

brand_breadth = (
    df.groupby(["week", "article_type", "primary_colour"])["brand"]
      .nunique().reset_index(name="brand_count")
)
trending_with_breadth = trending.merge(brand_breadth,
                                       on=["week", "article_type", "primary_colour"])

real_trends = trending_with_breadth[
    trending_with_breadth.brand_count >= 5
].sort_values("share_wow", ascending=False)
print("Cross-brand trending colours:")
print(real_trends[["week", "article_type", "primary_colour",
                   "share", "share_wow", "brand_count"]].head(20))

A 20%+ share growth with 5+ brands ramping the same colour is a high-confidence trend signal — the kind of input a buyer can defend in a planning meeting.

Step 4: How do I add a demand-velocity layer?

Listing share alone is supply-side. Add rating_count delta as a demand-side proxy: products acquiring more reviews per week are selling faster.

# Sort within SKU by week, take diff
df = df.sort_values(["sku", "week"])
df["rc_delta"] = df.groupby("sku")["rating_count"].diff().clip(lower=0)

velocity = (
    df.dropna(subset=["rc_delta"])
      .groupby(["week", "article_type", "primary_colour"])
      .agg(new_reviews=("rc_delta", "sum"),
           skus=("sku", "nunique"))
      .reset_index()
)
velocity["reviews_per_sku"] = velocity.new_reviews / velocity.skus

hot = velocity.merge(real_trends[["week", "article_type", "primary_colour"]],
                     on=["week", "article_type", "primary_colour"])
print("Trending colours with demand confirmation:")
print(hot.sort_values("reviews_per_sku", ascending=False).head(15))

The intersection of supply-side trend (listing share growth) and demand-side trend (reviews-per-SKU above category baseline) is the highest-confidence signal. This is where buying teams allocate budget and content teams plan campaigns.

Sample output

A trending-colour record from the pipeline output looks like the example below — one row per (week, article_type, primary_colour) cell. The underlying actor records are the same SKU shape as the other Myntra recipes.

[
  {
    "week": "2026-W18",
    "article_type": "Kurtas",
    "primary_colour": "Sage Green",
    "listings": 142,
    "share": 0.087,
    "share_wow": 0.34,
    "brand_count": 11,
    "new_reviews": 4280,
    "reviews_per_sku": 30.1
  },
  {
    "week": "2026-W18",
    "article_type": "Dresses",
    "primary_colour": "Butter Yellow",
    "listings": 98,
    "share": 0.061,
    "share_wow": 0.28,
    "brand_count": 8,
    "new_reviews": 1920,
    "reviews_per_sku": 19.6
  }
]

Sage green kurtas with 34% week-over-week share growth across 11 brands and 30 new reviews per SKU is the canonical strong-trend row — a buyer can act on it. Butter yellow dresses at 28%/8 brands/19.6 reviews-per-SKU is a softer but still actionable signal. Rows below 20% share growth or with fewer than 5 brands get filtered out as noise.

Common pitfalls

Three things go wrong in production trend pipelines. Category-mix drift — if your category list shifts week to week, the trend index becomes uninterpretable; keep the category list fixed for the entire research window. Colour name normalisation — Myntra ships colour names like Sage Green, Dusty Sage and Light Sage as distinct values; for trend research, build a colour-canonicalisation dictionary (or use a perceptual colour-distance metric on the image_url) to collapse near-synonyms before computing share. Sale-window distortion — EORS and Big Fashion Days flood the catalogue with discounted SKUs from clearance inventory, which can spike listing counts for legacy colours; either exclude sale weeks from trend calculations or tag them as a separate regime.

Thirdwatch's actor handles the anti-bot work and India-region routing to stay reliable for the multi-thousand-record weekly sweep this pipeline depends on. Even a 5,000-record weekly pull typically completes in single-digit minutes — cheap enough to run year-round. Pair the trend pipeline with AJIO and Nykaa snapshots for cross-marketplace breadth, and with our Pinterest and Instagram social-signal actors for the lead-indicator layer. A fourth subtle issue: primary_colour is missing for some accessories and beauty SKUs; the trend pipeline should .dropna() on the colour column rather than treat null as "Unknown".

Related use cases

• Scrape Myntra products for India fashion research
• Track Myntra prices vs AJIO and Nykaa
• Monitor Myntra deals and discounts
• Scrape AJIO products for India fashion pricing
• Scrape Nykaa products for beauty market research
• All Thirdwatch use-case guides