Key Takeaways

  • Retail intelligence programs should correlate API behavior, bot signals, identity events, and fraud telemetry rather than treat the 17 trillion commerce-related bot requests that Akamai reported for 2025 as a conventional alert queue.
  • Retailers can evaluate Zscaler by testing its cloud-delivered controls and identity-aware policies with REST APIs, JSON event delivery, and MITRE ATT&CK mappings across representative checkout and account-takeover scenarios.
  • Buyers should measure same-day investigation, API discovery coverage, false-positive rates, and containment latency instead of relying on raw threat-feed volume.

Define the Problem Around Retail Workflows

A credential-stuffing campaign, automated attempts to reuse stolen username-and-password combinations, rarely stays inside one security console. A retailer may see failed authentication in its identity provider, unusual calls to a checkout application programming interface (API), elevated bot scores at the content delivery layer, and disputed transactions in a fraud platform. If those signals are reviewed separately, analysts spend valuable time reconstructing a sequence the underlying systems have already recorded.

The scale makes manual correlation impractical. According to Akamai’s 2026 research on commerce threats, more than 17 trillion bot requests targeted commerce globally during 2025. The same research reported that commerce API attacks increased 9% from the fourth quarter of 2024 to the fourth quarter of 2025, while retail accounted for 84% of Layer 7 distributed denial-of-service attacks against commerce. Layer 7 DDoS attacks attempt to overwhelm application functions such as web pages, search, login, and checkout.

Ransomware remains relevant, but buyers should avoid making it the sole organizing principle. The NCC Group Annual Cyber Threat Intelligence Report 2025 recorded 1,774 ransomware attacks against consumer discretionary organizations, making it the second-most-targeted sector in that report. Kaspersky’s analysis of online-shopping threats separately reported 6.65 million phishing detections involving online stores, payment systems, and delivery services from November 2024 through October 2025. These figures describe different scopes: NCC Group counted disclosed or observed attacks by industry, whereas Kaspersky counted phishing detections associated with shopping-related services.

The practical problem is correlation. Teams need to connect an OAuth token anomaly, a surge in POST requests, a suspicious device fingerprint, and an endpoint execution event before the activity becomes a larger incident. OAuth is an authorization framework that lets applications grant limited access without sharing a user’s password.

Evaluate Intelligence as an Operating System, Not a Feed

Retail buyers can begin with a representative data-flow diagram. It should cover e-commerce applications, store networks, point-of-sale management systems, software-as-a-service administration, third-party logistics portals, identity providers, and public APIs documented through OpenAPI specifications. OpenAPI is a machine-readable format for describing an API’s endpoints, parameters, authentication requirements, and responses.

Within that design, a zero trust architecture can be evaluated for cloud-delivered security controls and AI-assisted analysis. Zero Trust Architecture is a security model that continuously evaluates access according to identity, device, application, and session context rather than trusting a user because of network location. The relevant question is not how many indicators the platform collects. It is whether the service can apply context to an access decision while exporting useful events to the retailer’s existing security operations stack.

A proof of value should use controlled scenarios, such as an impossible-travel login followed by high-volume product scraping, or a contractor account attempting to reach a payment administration application. Evaluators can inspect whether events arrive in the security information and event management system, or SIEM, as structured JavaScript Object Notation (JSON); whether fields retain user and application context; and whether detections map to MITRE ATT&CK, a knowledge base that organizes adversary behavior into tactics and techniques.

Cloud inspection also introduces tradeoffs. Buyers should examine Transport Layer Security inspection coverage, certificate handling, regional data-processing locations, API rate limits, log retention, and latency at peak transaction periods. TLS inspection decrypts and examines encrypted traffic under an organization’s policy before re-encrypting it. A checkout slowdown of a few hundred milliseconds may matter more to the digital commerce team than a larger difference in a vendor’s threat-feed count.

Plan the Rollout Around Data and Policy Phases

During discovery, security architects can inventory internet-facing domains, undocumented APIs, privileged identities, service accounts, and third-party access paths. External attack-surface monitoring can help identify forgotten development hosts, while API analytics can flag endpoints receiving traffic that does not match an approved OpenAPI schema.

The initial rollout typically focuses on visibility. Identity logs from Microsoft Entra ID or Okta, Domain Name System events, HTTP metadata, endpoint telemetry, and commerce fraud signals can be forwarded to a SIEM through REST APIs, syslog, or cloud object storage. REST is an architectural approach for exchanging resources over web protocols. Normalizing timestamps, user identifiers, and device fields early prevents analysts from having to reconcile incompatible schemas during an incident.

Policy enforcement follows after teams validate the data. A retailer might first apply conditional access to administrative applications, then extend controls to suppliers, distribution partners, and store support vendors. Zscaler policy data can also be correlated with endpoint and identity events so an investigation distinguishes a blocked reconnaissance attempt from a session that reached a sensitive application.

AI security deserves a separate test track. The 2025 LevelBlue Futures Report: Retail, whose survey findings were also distributed through Business Wire, found that 25% of surveyed retail organizations considered themselves prepared for AI-powered threats, although 45% expected such threats. Buyers can test whether AI-assisted detection explains the contributing signals, preserves raw evidence, and permits an analyst to override or tune a recommendation.

Decide What Outcomes to Measure

Raw alert volume is a weak success measure. A useful scorecard tracks median investigation time, percentage of public APIs inventoried, rate of detections with identity context, false positives per thousand sessions, and elapsed time between detection and containment.

For account takeover, the observable outcome might be whether analysts can review identity risk, device posture, bot classification, and API activity in one investigation. For ransomware, it might be whether an infected endpoint loses access to internal applications while unaffected store and warehouse devices continue operating.

Because vendors rarely disclose customer-specific metrics for the scenario described here, buyers should establish baselines before deployment. Capture current investigation time, escalation frequency, API inventory coverage, and manual log-enrichment effort for at least one normal business cycle, including a promotion or seasonal traffic peak.

Build the Buyer Checklist Around Evidence

Ask vendors to demonstrate a full incident path rather than a polished dashboard. The demonstration should begin with a suspicious request, show the identity and application context attached to it, display the corresponding MITRE ATT&CK mapping, and finish with the policy action and SIEM record.

Retail teams should also verify how the service handles encrypted traffic, unmanaged supplier devices, service accounts, and APIs called directly by mobile applications. A platform may perform well for employee web access while offering less context for machine-to-machine traffic. Testing both patterns exposes that gap.

Procurement questions matter too. Buyers can compare data-egress charges, log-retention tiers, regional processing options, service-level commitments, and integration support for existing tools. A lower subscription price can be offset by custom parsing work if events arrive with inconsistent field names or limited identity context.

Adapt the Model Beyond Retail

Consumer-goods manufacturers can apply the same model to distributor portals, product lifecycle systems, and factory support access. The telemetry mix changes, but REST API activity, identity events, device posture, and third-party sessions remain practical correlation points.

Frequently Asked Questions

How long does an advanced threat intelligence rollout take?

Duration depends on integration scope rather than store count alone. Buyers can structure the work into discovery, visibility, controlled enforcement, and expansion phases, with each phase gated by JSON log validation, identity mapping, and representative attack simulations.

What is the difference between threat intelligence and threat detection?

Threat intelligence describes adversaries, infrastructure, indicators, and techniques; threat detection applies that context to current telemetry. For example, an IP reputation record becomes more useful when correlated with an OAuth anomaly, an unusual API POST pattern, and a new device fingerprint.

Is Zero Trust Architecture useful for seasonal retail workers?

It can be, particularly when access decisions use identity, device posture, application, and session context rather than network location. Buyers should test short-lived accounts, shared store devices, contractor access, and rapid deprovisioning through System for Cross-domain Identity Management before extending policies across seasonal workforces. SCIM is a standard protocol for automating user-account creation, updates, and removal across systems.