What Riskified and Zendesk announced

On September 17, 2026, Riskified announced that its Identity Risk Intelligence system would be connected to Zendesk’s industry-specific agents. When a refund, product return or order change is requested, the support team is intended to receive a risk category and explanatory indicators before responding to the customer. [1 · Business Wire / MarTech Cube · Riskified and Zendesk integration announcement, September 17, 2026]

The signals are built from order, claim and chargeback histories across Riskified’s network. The company says this helps identify multiple accounts, address mismatches and recurring claims that may be invisible in a single store’s data. General availability is planned for November 2026: at this stage, it is accurate to describe an integration announcement, rather than a widely operational product. [1 · Business Wire / MarTech Cube · Riskified and Zendesk integration announcement, September 17, 2026]

Balancing losses and service

The proposed arrangement separates inquiries: low risk allows an automatic or accelerated decision, while high risk triggers additional review. The economic objective is to reduce abuse without subjecting all legitimate shoppers to the same burdensome checks. [1 · Business Wire / MarTech Cube · Riskified and Zendesk integration announcement, September 17, 2026]

Riskified cites an estimate of $849.9 billion in US retail returns in 2025 and a fraudulent-return share of about 9%, then calculates potential losses of $76.5 billion. These figures convey the scale of the problem but are not evidence of the future effectiveness of this specific integration; public data on decision accuracy, false positives and savings after deployment is not yet available. [1 · Business Wire / MarTech Cube · Riskified and Zendesk integration announcement, September 17, 2026]

Expert commentary

The integration matters because it moves risk assessment from a separate anti-fraud system into the moment of conversation with a shopper. Return decisions are often made quickly and under pressure to maintain service quality; context in the same interface could reduce manual searches. The sources confirm the intended workflow and timing but do not yet provide operational results. [1 · Business Wire / MarTech Cube · Riskified and Zendesk integration announcement, September 17, 2026] [2 · Riskified · official website] [3 · Zendesk · official company website]

The economic mechanism has two sides. Detecting recurring abuse reduces direct losses, while faster service for legitimate customers lowers the cost of each contact and the risk of losing them. But the benefit is easily overstated if every rejected request is counted as a prevented loss. A sound calculation includes wrongful denials, repeat inquiries, compensation, agents’ time and the customer’s subsequent value. [1 · Business Wire / MarTech Cube · Riskified and Zendesk integration announcement, September 17, 2026]

Network data gives Riskified a potential advantage over a model used by a single merchant: one person’s history can emerge across multiple stores. At the same time, that very scale raises the demands on identity matching. A shared family address, a move or multiple accounts do not equate to fraud. A risk category should prompt proportionate checks, rather than an unexplained automatic accusation. [1 · Business Wire / MarTech Cube · Riskified and Zendesk integration announcement, September 17, 2026] [2 · Riskified · official website]

For customer relationships, the ability to correct an error is critical. If a support agent sees only a final label, automation reinforces bias and makes appeals more difficult. It is more useful to show verifiable indicators, record the agent’s decision and measure differences across customer groups. The announcement mentions supporting indicators but does not disclose an appeals procedure or how false positives will be controlled. [1 · Business Wire / MarTech Cube · Riskified and Zendesk integration announcement, September 17, 2026] [3 · Zendesk · official company website]

The social effect of this news is confined to returns practices, but involves an important trade-off: faster service in exchange for combining behavioral data. Merchants will need clear retention periods, access controls and an explanation of how a signal influences a decision. The companies’ official pages describe their products, but the published announcement does not provide a complete data governance framework for the new integration. [1 · Business Wire / MarTech Cube · Riskified and Zendesk integration announcement, September 17, 2026] [2 · Riskified · official website] [3 · Zendesk · official company website]

The editorial team’s conditional forecast is that the first reliable signs of benefit will emerge after general availability in November and several months of seasonal returns. Indicators to monitor include losses per thousand inquiries, the share of manual reviews, resolution time, wrongful denials, repeat contacts, customer retention and the number of appeals. Until that data appears, only the integration announcement and its proposed architecture are established. [1 · Business Wire / MarTech Cube · Riskified and Zendesk integration announcement, September 17, 2026]

Sources

  1. Business Wire / MarTech Cube · Riskified and Zendesk integration announcement, September 17, 2026 — Report on features and planned availability; estimates of the scale of losses and the effect come from the provider.
  2. Riskified · official website — Description of the company and its e-commerce risk management solutions.
  3. Zendesk · official company website — Description of the customer service platform and company profile.