TestNow Operational Monitoring Platform

Problem

Uptime checks prove a page responds, not that the business function on it still works — a contact form can return HTTP 200 for weeks while silently delivering nothing, and nobody finds out until a customer complains. Reasoning about health across services, vendors, and downstream systems is also hard when every tool reduces it to a human-readable status page.

Solution

Exercised the real function rather than the page around it: synthetic submissions run as durable, crash-recoverable Workflows, and region-aware probes cover the conventional uptime signals from nine locations. Operational state is modeled on Open Operational State so results stay machine-readable rather than ending at a status page.

Role

Founder, system architect, and implementation lead

Technologies

Impact

• Created a product implementation path for Open Operational State concepts
• Built durable, crash-recoverable form-submission testing on Cloudflare Workflows, with per-monitor parallelism, configurable field permutations, and template-driven payloads, so a form that silently stops delivering is caught rather than assumed working
• Ran HTTP, keyword, SSL-certificate, and DNS checks at one-minute intervals from Durable Object probes across nine Cloudflare regions
• Modeled multi-tenancy with organization-scoped data, role-based access, and invitation workflows, deliberately decoupling billing from the organization hierarchy so pricing changes do not require reshaping tenancy
• Delivered state-change alerting through an API-to-queue-to-email pipeline with failure categorization, so a transient blip and a sustained outage are not reported identically
• Added an absence detector running outside Cloudflare to catch the scheduler itself failing, so a monitoring platform that stops running is noticed by something that does not share its failure domain