Mogul Wireless

2016

Taking It Over

Mogul didn’t start with us. Early in my career I’d worked as a consultant at another firm, and one of its owners had built this product for Mogul, going back to around 2012. By 2016 they were no longer interested in looking after it, and Rob Bell, Mogul’s owner, had struggled to find a firm to work with that he was happy with.

So we took it over. The app was running without major issues, but there were performance problems to look at, and Rob had a lot of functionality he wanted to add. More than anything, he was looking for a partner to move forward with. We signed our agreement in February 2016, my first commit was on the 23rd, a small one adding notes to a sensor, and the retainer began on 1 March.

What it does

Watching Every Five Minutes

Wireless sensors sit in the fridges, freezers and rooms that matter, and read the temperature, the humidity or the room’s pressure about every five minutes. A gateway on site collects their readings and sends them on to the collector, a small app of Mogul’s own, which queues them for the web app to import.

The web app is where everything else happens. Each hospital or pharmacy is organised into groups, each group has its sensors and their limits, and each has an alert tree: who hears about a problem, by text, phone call or email, and how long after it starts. Everything is recorded, because an auditor will want to see it.

Today that’s about 2,500 sensors on 91 gateways, sending in something like 950,000 readings a day. The app has raised more than 450,000 alerts since 2012, and sent 1.68 million notifications about them.

The hard part

When the Gateways Come Back

Some of the hardest moments over the years have come from a gateway doing exactly what it should. When one has been offline and comes back, it sends the collector everything it has been holding, all at once. When enough of them come back together, the collector is flooded with readings, all of which still have to be imported, checked and charted.

So we made it scale. In 2017 we moved the app and its collector to Heroku, with Sidekiq behind them, and protected our services so a flood of readings wouldn’t take them down. In late 2018 we took the import apart. A gateway’s log is now parsed into its readings, each reading is imported as its own job, in parallel, and only once the last one lands are the aggregates and the displays brought up to date, once per import instead of once per reading.

2018

Readable From Across the Room

Not everyone who needs to know a fridge is warm is logged in. In 2018 we built the sensor displays: small, independent screens that sit beside the sensors, each showing a rendering of its readings that’s meant to be read from across the room.

They’re served by the web app, can be embedded anywhere, and have been tuned over the years for the small browsers they run in, including a stylesheet all of its own for one of them. In 2019 we darkened their colours for contrast, and this year I raised it again, with WCAG’s contrast guidance in mind. Rob is installing more and more of them.

Compliance

The Reports Auditors Ask For

A lot of what Mogul sells is peace of mind at inspection time, so over the years we’ve kept rolling out new kinds of sensors and more of the compliance reporting its customers need: door sensors in 2020, compliance reports in 2022, and daily and monthly reports sent on a schedule.

In early 2025 I built support for Vaccines for Children, the federal programme with its own rules for checking and recording vaccine fridges. A sensor can be enrolled, someone performing a check sees the lowest and highest temperatures of the last 24 hours, reminders go out twice a day until the checks are done, and the whole record comes out as a PDF.

2026

An Alert You Can Trust

An alert is only as good as the phone it reaches, and a wrong or dead number means somebody stops getting alerts and nobody finds out. This year’s biggest piece of work, and the one I’m proudest of, was closing that gap, with much deeper auditing and a much closer integration with Twilio.

Everyone now has their own record of their own number, and it isn’t relied on for alerts until a code texted to it comes back. Twilio’s word on every message is recorded against the number. A STOP or a landline is written off at once, a phone that’s just switched off has to fail ten times in a month before it is, and a failure that’s really about us, not the phone, never writes anyone off, because acting on it would silence everyone at once. When a number stops working, its owner is emailed to say so.

A client’s admin can see who won’t get their alert texts, and why, and a chart of delivery errors over the last thirty days. And the phone calls got better too: alerts are read in a neural voice, and door alerts finally have something to say on the phone.

Keeping up

Every Upgrade, as It Comes

Looking after an app for ten years mostly means keeping it current, and the work has an ebb and flow to it. Between August 2024 and December 2025, I took it from Rails 6.1 through every version to Rails 8.1, turning on each version’s new defaults one at a time, and Ruby up to 3.4. Along the way it moved to GitHub Actions, Sidekiq 7, Active Storage and Pundit.

This summer the admin, where Mogul’s own team works, was restyled in Mogul’s own palette, and the login was made to look as professional as the product is. I’m also building Mogul’s new marketing website, which isn’t live yet.