Executive Summary
"How many units is this product selling?" is the most valuable question in commerce intelligence and the one most likely to be answered dishonestly.
The honest answer starts with a boundary: marketplaces do not publish exact unit sales for most products, so exact sales volume cannot be scraped. What can be captured is a set of public signals — bestseller rank, review velocity, "X bought recently" indicators where shown, stock movement, and rating accumulation — that correlate with sales and can be combined into a defensible estimate. The difference between "we scrape sales volume" and "we build a validated demand proxy from public signals" is the difference between a claim that collapses under scrutiny and a dataset a team can actually use.
This report covers marketplace sales volume data scraping done honestly: which signals exist, how they combine into an estimate, where the limits are, and what the data looks like.
This report is published by Product Data Scrape. Sample figures are illustrative of structure and method, not audited sales figures.
The Honesty Boundary — Stated First
This category is defined by what it cannot do, so that belongs at the top.
Exact units sold is generally not public. Most marketplaces do not publish per-product unit sales. Anyone claiming to scrape exact sales volume for arbitrary products is overstating what the data supports.
What is public are signals that correlate with sales. Bestseller rank, review count and velocity, "bought in past month" badges where surfaced, stock decrements, and rating accumulation are all observable and all related to sales, imperfectly.
Combining signals produces an estimate, not a measurement. A well-built demand proxy is a validated estimate with a stated confidence, not a precise count. Treating it as a count is the error that gets teams into trouble.
Product Data Scrape captures the public signals and can build a demand-proxy estimate from them, with the methodology and its limits stated. It does not claim to read exact sales that marketplaces do not publish. This honesty is not a weakness of the offer; it is what makes the estimate trustworthy enough to act on.
The Signals That Exist
Bestseller rank. Where a marketplace publishes category rank, it is a strong ordinal signal — rank 5 outsells rank 500 — though the rank-to-units relationship is non-linear and category-specific.
Review velocity. The rate of new reviews correlates with sales, adjusted for a review-rate that varies by category and price.
"Bought recently" indicators. Some platforms surface "X bought in the past month" badges — a coarse but direct signal where available.
Stock movement. Observed decrements in stock indicators, where surfaced, proxy sales between captures.
Rating accumulation. The pace at which total ratings grow tracks purchase volume over time.
Each is noisy alone. The method is combining them.
The Traps
Trap one: presenting an estimate as a count
The single biggest error. A demand proxy is an estimate with uncertainty. Reporting it as "units sold" invites decisions the data cannot support.
Trap two: linear rank assumptions
Bestseller rank is non-linear — the gap between rank 1 and 10 is far larger in units than between 100 and 110. A model that treats rank linearly is wrong in the direction that matters most.
Trap three: ignoring category calibration
Review rates, rank dynamics, and badge availability differ by category and price band. A proxy calibrated on one category misleads in another.
Trap four: no confidence reporting
An estimate without a confidence band is used as if it were exact. Every estimate has to carry its uncertainty.
What a Usable Demand-Proxy Dataset Captures
| Field group |
Fields |
| Identity |
product_id, category, brand, platform |
| Raw signals |
bestseller_rank, review_count, review_velocity, bought_recently_badge, stock_movement, rating_count |
| Estimate |
demand_proxy_score, estimated_units_band, confidence |
| Method |
model_version, category_calibration, signals_used |
| Capture |
captured_at, capture_window |
The Estimate group reports a band with confidence, and the Method group states how it was derived — the two things that keep the dataset honest.
Sample Data: A Demand-Proxy Record
An illustrative record.
| Field |
Value |
| Product |
BrandA Snack 150g |
| Category |
Snacks |
| Bestseller rank |
12 |
| Review velocity |
45/week |
| "Bought recently" |
500+ |
| Demand proxy score |
82 / 100 |
| Estimated units/month |
2,000–3,500 (band) |
| Confidence |
Medium-high |
Illustrative — an estimate, not an audited count.
The structured record:
{
"product_id": "MKT-SNACK-BRANDA-150G",
"category": "snacks",
"platform": "marketplace_a",
"captured_at": "2026-07-15T10:00:00+05:30",
"raw_signals": {
"bestseller_rank": 12,
"review_count": 4130,
"review_velocity_per_week": 45,
"bought_recently_badge": "500_plus",
"rating_count": 5200
},
"estimate": {
"demand_proxy_score": 82,
"estimated_units_band_monthly": [2000, 3500],
"confidence": "medium_high"
},
"method": {
"model_version": "v2",
"category_calibration": "snacks_in",
"signals_used": ["rank", "review_velocity", "bought_badge"],
"note": "estimate derived from public signals; not an exact sales figure"
}
}
The record never claims an exact number. It reports a band (2,000–3,500/month), a confidence, the signals used, and an explicit note that it is an estimate. That structure is what lets a category team use it for relative comparison and market sizing without over-trusting a false precision.
What the Data Enables
Relative demand ranking. Even as estimates, the signals reliably order products by demand — which is selling more, which less — for competitive and assortment decisions.
Market sizing and entry. For a brand entering a category or market, banded estimates across the category size the opportunity and identify the top-performing products and platforms.
Bestseller and trend identification. Rising velocity and improving rank flag products gaining demand.
Competitive benchmarking. A brand compares its own demand signals against competitors' on a consistent basis.
Who Uses Sales Volume Signal Data
Brands and sellers estimate how their products and competitors' are selling, in relative terms, for assortment and competitive decisions — the need behind "how many units are competitors selling" and "units sold, price range."
Market-entry analysts size categories and identify top products and platforms before entering — as in requests to see "what's in demand, how many being sold, and top-performing platforms."
Product-research tools surface demand estimates and bestsellers for sellers.
Category strategists track demand shifts over time from consistent signals.
Limitations
These are estimates derived from public signals, not exact sales figures, which marketplaces generally do not publish. Accuracy varies by category, price band, and signal availability, and every estimate carries a confidence band. Rank-to-units relationships are non-linear and calibrated per category. We capture publicly available signals only; sample figures illustrate method rather than audited sales.
About the Data
This report was produced using marketplace sales volume data scraping methods from Product Data Scrape. We capture public demand signals — bestseller rank, review velocity, "bought recently" indicators, stock movement, rating accumulation — across marketplaces, and combine them into category-calibrated demand-proxy estimates with stated confidence bands.
Delivered as JSON, CSV, or via API, with methodology and limits documented, because an estimate is only useful when its uncertainty is honest.
Need to estimate demand across a category? Product Data Scrape will build a demand-proxy sample on your products and competitors', with confidence bands and stated method — a defensible estimate, not a false precise count.
Product Data Scrape — turning marketplace complexity into decision-ready data.
Singapore
EMAIL
PHONE
.webp)
