Build an Amazon Product Research Tool (2026 Guide)

Thirdwatch's Amazon Scraper gives product-research SaaS builders, e-commerce operators, and retail-arbitrage teams a structured Amazon data layer — title, price, BSR, rating, reviews_count, monthly-sold, image, brand, category. Build a Jungle Scout / Helium 10 alternative with custom scoring tuned to your niche.

Why build an Amazon product-research tool

Amazon product research is a $200M+ SaaS market. According to Marketplace Pulse's 2024 Amazon Data report, more than 2 million active third-party sellers use product-research tools to identify niches, validate demand, and price competitively. The two market leaders (Jungle Scout, Helium 10) charge $50-$200/month per seat for a workflow that's fundamentally about ingesting Amazon's product, BSR, and rating data and computing opportunity scores. For platform builders and high-volume operators, owning that data layer directly is materially more flexible.

The job-to-be-done is structured. A product-research SaaS founder wants to ingest 500K+ Amazon listings weekly and compute proprietary opportunity scores. An e-commerce operator wants to evaluate 50 candidate niches per quarter for new-product launches. A retail-arbitrage team wants daily BSR snapshots on a 5K-product watchlist. An Amazon agency wants per-client niche-opportunity dashboards. All reduce to category-level or keyword-level scrapes + BSR + rating filters + custom scoring.

How does this compare to the alternatives?

Three options for Amazon product-research data:

Jungle Scout and Helium 10 are the SaaS market leaders but their pricing assumes individual-seat consumption. Amazon's official Product Advertising API is gated behind affiliate program approval and limits fields. The Amazon Scraper actor page gives you the raw data at the lowest unit cost — and you compose the scoring algorithm.

How to build a product-research tool in 4 steps

Step 1: How do I authenticate against Apify?

Sign in at apify.com (free tier, no credit card), open Settings → Integrations, and copy your personal API token. Every example below assumes the token is in APIFY_TOKEN:

export APIFY_TOKEN="apify_api_xxxxxxxxxxxxxxxx"

Step 2: How do I pull a category-level batch?

Pass keyword queries that map to the niche category you're researching.

import os, requests, pandas as pd

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

NICHES = ["silicone baking mat", "garlic press stainless steel",
          "knife sharpener", "vegetable spiralizer",
          "kitchen scale digital", "coffee scale",
          "salad spinner", "mandoline slicer"]

resp = requests.post(
    f"https://api.apify.com/v2/acts/{ACTOR}/run-sync-get-dataset-items",
    params={"token": TOKEN},
    json={"queries": NICHES, "country": "us", "maxResults": 100},
    timeout=3600,
)
df = pd.DataFrame(resp.json())
print(f"{len(df)} products across {len(NICHES)} niches")

8 niches × 100 products = up to 800 records.

Step 3: How do I score niche opportunities?

Composite score: high reviews + mediocre rating + competitive price band.

df["rating"] = pd.to_numeric(df.rating, errors="coerce")
df["reviews_count"] = pd.to_numeric(df.reviews_count, errors="coerce")
df["price"] = pd.to_numeric(df.price.str.replace("$", "").str.replace(",", ""), errors="coerce")

niche_scores = (
    df.groupby("searchString")
    .agg(
        product_count=("title", "count"),
        median_price=("price", "median"),
        median_rating=("rating", "median"),
        median_reviews=("reviews_count", "median"),
        top_brand_share=("brand", lambda x: x.value_counts(normalize=True).iloc[0] if len(x) > 0 else 0),
    )
    .assign(
        opportunity_score=lambda d: (
            (d.product_count.between(50, 500))  # not too crowded
            * (d.median_rating.between(3.8, 4.3))  # quality gap
            * (d.median_reviews >= 200)  # real demand
            * (d.top_brand_share < 0.4)  # no monopoly
        ).astype(int)
    )
    .sort_values("opportunity_score", ascending=False)
)
print(niche_scores)

The four-condition opportunity filter surfaces niches where (1) competition is tractable, (2) existing quality is mediocre, (3) demand is real, and (4) no one brand dominates.

Step 4: How do I track BSR drift over time for shortlisted products?

Persist (asin, snapshot_date, bsr) for ongoing tracking.

import psycopg2

shortlist = df[df.bsr.notna() & (df.reviews_count >= 100)].copy()
shortlist["bsr"] = pd.to_numeric(shortlist.bsr, errors="coerce")

with psycopg2.connect(...) as conn, conn.cursor() as cur:
    for _, p in shortlist.iterrows():
        cur.execute(
            """INSERT INTO bsr_history (asin, title, bsr, price, snapshot_date)
               VALUES (%s, %s, %s, %s, current_date)
               ON CONFLICT (asin, snapshot_date) DO NOTHING""",
            (p.asin, p.title, int(p.bsr), p.price)
        )

print(f"Persisted BSR snapshot for {len(shortlist)} products")

Daily BSR snapshots over 4-8 weeks build the time-series needed for sales-velocity estimation and seasonality detection.

Sample output

A single Amazon product record looks like this. Five rows weigh ~6 KB.

{
  "asin": "B07ABC1234",
  "title": "Premium Silicone Baking Mat - 16x12 inch Set of 2",
  "price": "$14.99",
  "rating": 4.6,
  "reviews_count": 12450,
  "bsr": 234,
  "category": "Kitchen & Dining > Bakeware > Baking Mats",
  "brand": "Acme Kitchen Co",
  "image_url": "https://m.media-amazon.com/images/I/...",
  "in_stock": true,
  "prime_eligible": true,
  "monthly_sold": "10K+ bought in past month"
}

asin is Amazon's canonical natural key — stable across price, title, and image changes. bsr is the headline ranking metric; lower = better. monthly_sold is the new-in-2024 field exposing approximate units sold (Amazon previously only inferred this from BSR). prime_eligible: true indicates Amazon-fulfilled inventory which converts at higher rates than third-party-fulfilled. category provides hierarchical category context for cross-niche normalization.

Common pitfalls

Three things go wrong in product-research pipelines. BSR fluctuation noise — BSR can swing 20-30% within a single day based on hourly purchase volume; for stable analysis, average BSR across 5+ daily snapshots before treating as a niche signal. Sponsored-product contamination — Amazon's search results mix organic and sponsored listings; the actor returns both, but for organic-niche analysis filter out rows where is_sponsored: true. Brand-name normalization — sellers often spell their brand inconsistently across listings ("Acme Kitchen Co" vs "ACME Kitchen" vs "Acme"); for accurate top-brand-share computation, normalize via case-folding and a controlled vocabulary.

Thirdwatch's actor handles proxy rotation so you can focus on the data. Pair Amazon with Flipkart Scraper for cross-marketplace pricing intelligence in India and AliExpress Scraper for sourcing-cost research. A fourth subtle issue worth flagging: Amazon's "X bought in past month" field is bucketed (50+, 100+, 500+, 1K+, 5K+, 10K+) rather than exact, so for fitted BSR-to-sales-volume regressions use bucket-midpoints rather than treating the threshold as the value. A fifth pattern unique to product-research work: certain seasonal niches (holiday gifts, summer outdoor) show 5-10x BSR volatility around Q4; for niche-opportunity scoring, exclude rows scraped between October-January from your year-round opportunity model and treat seasonal-window data as a separate scoring dimension. A sixth and final pitfall: Amazon's search ranking sometimes injects 1-2 unrelated products into category searches (e.g., a knife in a "garlic press" search); for tight category aggregation, post-filter on category_path containing the expected category token before computing per-niche statistics.

Related use cases

• Scrape Amazon products for price monitoring
• Track Amazon bestsellers by category
• Monitor Amazon competitor pricing and ratings
• The complete guide to scraping ecommerce
• All Thirdwatch use-case guides