Begin with the journeys that matter
Describe how your people and locations reach the applications they need. Include a managed laptop at home, a contractor using a browser, a branch reaching a private application, and a voice or contact-center workflow. These journeys expose different dependencies and user-experience requirements. Use them to define the evaluation instead of starting with a long feature checklist.
Map identity and device decisions
For each journey, record how the user or device is identified and what evidence is required before access is allowed. Include exceptions such as unmanaged devices, service accounts, and applications that cannot use modern authentication. The implementation plan should show who owns those exceptions and when they will be reviewed.
Exercise the complete data path
Test the application, the route, the security decision, and the return traffic together. Include private applications, SaaS, internet browsing, and access from representative locations. Cloudflare One’s documentation separates private-application access, network connectivity, and implementation patterns; a real pilot should likewise make its traffic paths explicit rather than assuming a single test covers them all.
Introduce a failure deliberately
Within an approved test window, examine what happens when a circuit, tunnel, connector, or identity dependency becomes unavailable. Record what the user sees, what the operator sees, and how service is restored. Agree acceptable behavior before testing. A successful failover means the important application still works within the agreed requirements.
Review data controls and exceptions
Use permitted sample data to evaluate the policies that matter to the business. Check where inspection applies, how exceptions are approved, and how incidents are investigated. Bring application owners into the review when a control changes a workflow. Include the effort required to explain and maintain policies across different user and device groups.
Compare the cost of operating the design
Alongside licensing and circuits, account for migration, support ownership, logging, training, and the systems being retired. Document contract dependencies and a rollback plan before a broad rollout. Finish with a decision record that names the tested journeys, limitations, operational responsibilities, and next migration wave. That record is more durable than a score based on feature count alone.

