Price Anomaly Detection: Spotting Counterfeit Products Early with Data Science
The earliest indicator of counterfeit products in digital markets is a price anomaly. An abnormal drop in the regional or temporal price movements of the same SKU often signals activity…
Data Science & Counterfeit Detection • 2025
The earliest indicator of counterfeit products in digital markets is a price anomaly. An abnormal drop in the regional or temporal price movements of the same SKU often signals activity involving counterfeit products or the grey market. Data science detects these anomalies to protect brands proactively.
B10 CounterShield™ • 2025 • Reading time: ~7 min
1. What Is a Price Anomaly?
An unexpected price deviation for a product (SKU) in certain regions, platforms or periods. For example, a product normally priced at 1000 TL falling to 600–650 TL across marketplaces in the same week indicates a risk of counterfeit or unauthorised supply.
“Price is the quietest but most reliable indicator of counterfeiting.”
2. The Basic Logic of Anomaly Detection
Data science uses statistical modelling and machine learning to detect these deviations:
- Z-Score Analysis: Values differing from the mean price by 2 or 3 standard deviations are considered anomalies.
- IQR (Interquartile Range): Prices outside the lower and upper quartile boundaries are considered suspicious.
- Isolation Forest: Isolation-based detection, especially in multivariate datasets (region, date, seller rating, stock quantity, etc.).
- LSTM (Time Series Forecasting): Significant deviations from time-series predictions are labelled as anomalies.
3. Data Sources
- Marketplace APIs (Amazon, Trendyol, AliExpress, Temu)
- Seller history and price history (price history endpoints)
- Competitor brands and similar SKU prices
- Currency / regional exchange-rate data (to normalise regional price differences)
4. Model Workflow: Anomaly Pipeline
| Stage | Description | Technology / Method |
|---|---|---|
| Data Collection | SKU price data from API and web sources | Python, Scrapy, REST API |
| Cleaning and Normalisation | Normalizing on a currency, date and platform basis | Pandas, NumPy |
| Anomaly Detection | Z-Score, Isolation Forest, LSTM | Scikit-learn, PyOD, TensorFlow |
| Warning & Reporting | Alerts for SKUs exceeding the anomaly threshold | Power BI, Looker Studio, Slack Bot |
5. Application Example
The system developed by B10, “PriceShield”, monitors more than 100,000 SKUs daily to analyse regional price distributions. For each SKU, the model generates an anomaly_score as follows:
{ "sku": "B10-COS-2025", "mean_price": 999.00, "current_price": 649.00, "z_score": -3.42, "region": "TR-West", "anomaly_score": 0.91, "flag": "high-risk" }
“A cheaper SKU does not necessarily mean a greater likelihood of counterfeiting; yet no warning is as systematic as a price deviation.”
6. Regional Deviation and Risk Score
The Price Variance Index (PVI) is calculated for each SKU. This score measures the product’s price stability.
| PVI Range | Status | Comment |
|---|---|---|
| 0.0 – 0.2 | Stable | No price difference / market consistent |
| 0.2 – 0.5 | Warning | A regional promotion or stock difference may be involved |
| 0.5 – 1.0 | Critical | Risk of counterfeit products or unauthorised sellers |
7. 2025 Trends: Predictive Counterfeit Detection
- AI models now explain not only an anomaly, but also why it occurs (Explainable AI / SHAP).
- Blockchain-linked SKU identities verify original product price data on the chain.
- Graph-based fraud detection: detecting coordinated price manipulation through seller connection networks.
Spot price deviations early and stop counterfeiting before it starts.
The B10 Digital Agency offers data-based price analysis, anomaly detection and counterfeit monitoring solutions for brands.
Sources and related documents
These documents provide references for the methods and rules discussed in this article.