Our Methodology

When you submit a product URL, trustd fetches every review from Takealot's public API and runs two layers of anomaly detection:

• Duplicate reviews — the same customer account posting multiple reviews on the same product under the same name. This can be caused by double-clicks, re-purchase prompts, or deliberate attempts to boost a rating.
• Identity manipulation — the same customer account posting reviews under different display names on the same product. This ranges from innocent profile updates to deliberate attempts to appear as multiple people.

Every review is cross-referenced against all other reviews on the product using customer account IDs from Takealot's API. This is the same data visible to any shopper on the product page — trustd does not access any private information.

trustd classifies each detection by severity so you can see the difference between innocent glitches and genuine manipulation:

• Duplicate reviews are treated as innocent — extra copies are removed, but the first review is kept at full weight.
• Name changes with the same initial (e.g. "K" to "Kayla") are likely profile updates, not manipulation. Extra reviews are downweighted to 70%.
• Name changes with different initials (e.g. "John" to "Sarah") are suspicious. Extra reviews are downweighted to 50%.
• 3+ different names on the same product from one account is a strong manipulation signal. Extra reviews are removed entirely.

This tiered approach means trustd never overreacts to innocent behaviour while still catching genuine manipulation. The goal is accuracy, not false alarms.

The Trustd Rating uses weighted averaging. Every review starts with a weight of 1.0. When an anomaly is detected, the first review from that account always keeps full weight. Extra reviews are either removed (weight 0) or downweighted (weight 0.5–0.7) depending on severity.

The result is a rating that reflects genuine customer sentiment more accurately than a simple keep-or-remove approach. Products with no detected anomalies will have a Trustd Rating identical to their Takealot rating.

All data comes directly from Takealot's public product review API. Reviews are fetched in real time when you submit a URL. Nothing is cached or pre-computed. You're seeing the same reviews visible on the Takealot product page, just analysed for manipulation.

trustd has also conducted a full-scale analysis of the Takealot marketplace — over 6.4 million reviews across 1.28 million products. This dataset powers the aggregate statistics shown on the site and informs our understanding of marketplace-wide review patterns.

This analysis detects duplicate reviews and identity manipulation based on customer account data visible in the public API. It does not yet detect other forms of manipulation such as:

• Incentivised reviews (reviews given in exchange for free products or payment)
• Coordinated review networks across multiple products
• Bot-generated review text
• Review velocity anomalies (sudden floods of reviews)

We're continuously expanding our detection methods. As new patterns are identified and validated, they are added to the ReviewShield engine.