How ForeverPet hears about a stuck order before the customer does

ForeverPet makes custom pet sculptures from a photo. A customer uploads one picture of their dog or cat, sees a preview of the sculpture in seconds, and orders a keepsake that is sculpted and shipped to their door. Many of those orders are memorials for a pet who has just died.

That makes a quiet failure expensive. If a photo upload breaks, or an order stalls halfway through, the person on the other end is often grieving, and they won't file a bug report. They just leave. A small team can't sit and watch dashboards all day, so ForeverPet needed something that would tell them the moment anything went wrong.

What CanaryHub watches

The site itself

CanaryHub checks foreverpet.ai around the clock against its /api/healthendpoint. That endpoint doesn't only say "the server is up". It also reports how many credits are left with the 3D sculpting service ForeverPet relies on. When the balance drops below a set level, the check turns from ok to degraded, so the founder tops up before customers start seeing failed previews.

Every step of an order

The app sends a short event to CanaryHub at each important step: a photo uploaded, a photo retaken, an email or phone number entered. Failures are sent as errors, each named for the step that broke:

• a photo that couldn't be turned into a preview
• a sculpture that failed to start, or failed its quality check
• a multi-pet keepsake that couldn't be put together
• an order that needs a human to look at it

Each of these is set up as a watch. The moment one fires, CanaryHub pings every channel the founder has connected, with the error message attached so it's clear what to fix without opening a log.

A daily digest

Once a day, CanaryHub sends a summary: uptime, response times and what happened in the pipeline. On a calm day it simply says all is quiet, which is its own kind of reassurance.

Why it fits a small shop

ForeverPet didn't add an agent, a dashboard to babysit, or a monitoring stack. It's one health endpoint and a tiny client that sends events and never throws, so a monitoring hiccup can't break a customer's order. The alerts come to the founder, wherever they are.

We run CanaryHub on ForeverPet ourselves; it's made by the same team. Watching a real shop with real customers is how we find what's missing: when an error watch on ForeverPet once stayed silent, we traced it to a bug in CanaryHub's own event pings and fixed it for everyone.