The $30,000 Refund Nobody Caught: What Restaurant Data Is Really Worth
Here’s a number that should make every multi-unit operator go look at their own books.
A company with 27,000 employees ran an audit of their security camera footage. They were looking for one thing: cash refunds issued when there was no customer standing at the register. What they found was four employees — four, out of twenty-seven thousand — quietly ringing refunds in cash after the guest had already walked out the door, and pocketing the money. That single audit surfaced $30,000 in fraudulent refunds.
I’ve had a lot of conversations on this podcast about restaurant technology, but the one I just had with Tom Woodbury from MachineQ kept coming back to a theme I don’t think we talk about enough: the money is already leaking out of your business. You just can’t see it. And in almost every case, the reason you can’t see it is that the data doesn’t exist yet — or it’s sitting on a clipboard in a filing cabinet where it does you exactly zero good.
MachineQ is the IoT connectivity platform inside Comcast Business. Tom has spent the last 15 years figuring out how technology can make restaurants more profitable — and, in true restaurant fashion, he started as a 12-year-old dishwasher and came up through the ranks. (He jokes that “CUTTR” — came up through the ranks — is his certification, the one you don’t see stitched on the chef’s jacket.) I came up the same way, so this was a fun one. Let’s dig into the money.
Start with the return, not the technology
Whenever I talk to operators about IoT, the first objection is always the same: it’s too expensive, or it’s too complicated to figure out. So I asked Tom to start with the R in ROI — the return — before we ever touched the tech. He broke it into three buckets, and the math is more compelling than most people expect.
Labor reallocation. Your managers are spending 30 to 45 minutes a day walking around with a clipboard — checking the walk-in temperature, the reach-in, the freezers, the dumpster. The FDA wants those temperatures logged periodically throughout the day, so most stores do it at the start of or during every meal period. Tom’s point is simple: there is no reason a human needs to do that. Now, you’re not going to cut a manager’s salary by 45 minutes a day, so this isn’t hard-dollar savings. But it’s 45 minutes reallocated to the thing that actually drives revenue — taking care of customers. Happier customers come back more often, and they spend more.
Food waste reduction. This is where it gets real. A catastrophic walk-in cooler failure — mechanical, power, or just a door left open — means throwing out roughly $10,000 of food. Assume that happens once every five years, and you’re looking at about $2,000 a year in avoidable food waste. And that’s before you count the cost of replenishing the cooler, the labor to do it, the downtime, and the guests who walk because you’re 86’d on half the menu.
Energy. This is the one that blew me away. If you tune your walk-in cooler to run at food-safe temperatures — say, moving from 36°F to 38°F — you cut the utility cost for that single device by about 9%. Two degrees. Nine percent. Multiply that across every piece of refrigerated equipment in every store and it stops being a rounding error.
Here’s the framing I keep coming back to: in this industry, most of us are working on high-single-digit to low-double-digit margins. Every dollar you save is roughly ten dollars of top-line sales you don’t have to go chase. Cost savings isn’t the boring cousin of revenue growth. At these margins, it’s the faster path.
The clipboard is a liability, not a record
Tom said something that reframed the whole paper-log conversation for me. The clipboard isn’t a management tool — it’s a compliance artifact. That paper goes into a filing cabinet, sits there for 90 days (or, as he put it, “sometimes 90 years, it feels like”), and then somebody throws it away. Nothing is ever done with it. Its one and only job is to satisfy a health inspector if one shows up.
Which means you’re doing all that work and getting almost none of the value. Digitize the same data and suddenly it does something. You can compare walk-in cooler brands across your fleet and see which one is actually the most efficient, which one has the longest mean time between failures — real data to inform your next equipment purchase instead of a sales rep’s brochure. You simply cannot extrapolate that from paper.
And then there’s the reliability problem, which Tom told with a story I loved. He’s seen paper logs where every reading is in the same pen, the same handwriting, always 38 degrees — and where the manager was apparently so clairvoyant he’d already filled in next week’s temperatures. Pardon the levity, but it makes a serious point. If you ever have to prove in a legal setting that food was stored safely, and opposing counsel puts your manager on the stand and asks, “Have you ever written the wrong number? Ever gone back and added a reading you forgot to take?” — the moment the answer is yes, every paper log in your building is suspect. An IoT sensor can’t be back-filled. It’s always on, always recording. Heaven forbid you’re ever pulled into a food-safety outbreak investigation, that difference matters enormously.
What actually happens between 11 a.m. and 4 p.m.
I pushed Tom on the behavioral gap, because this is where I’ve seen brands get hurt. You check the walk-in at 11 before the lunch rush. You don’t check it again until 4. In between, somebody does inventory and leaves the door cracked, or a delivery comes in the back and the door sits open. Cut leafy greens — one of the highest-risk items in any kitchen — go bad the fastest and are the most likely to support pathogen growth. Worse, picture a key drop at 3 a.m. when nobody’s on site: the delivery driver props the cooler door open for four hours moving product in and out. You walk in the next morning with no idea that food sat unsafe half the night. A sensor reporting every 15 minutes, with an alert when the walk-in has been over 50°F for 30 minutes, closes exactly that blind spot.
Tom’s team has gone well beyond the walk-in. They’ve got clients with sensors in every piece of refrigerated or frozen equipment in the store — make lines, make tables, even a machine that dispenses cream into coffee, where they figured out how to track temperature inside the device. One client put IoT sensors on their dumpster and discovered the trash never once exceeded 70% full; they went from two pickups a week to three every two weeks. Ice makers, fryer oil, banquet-room HVAC that’s been cooling an empty room for six weeks — it’s all fair game. As Tom put it, it might be $10 a month here and $500 a month there, but those micro-improvements add up to a materially more profitable enterprise.
“But my CIO tried IoT and it failed”
If you evaluated this five, seven, or ten years ago and walked away, I don’t blame you. I had those same conversations back then — it looked great on the whiteboard and fell apart in real life. Tom talked to a VP of IT recently whose CIO won’t touch IoT because he tried it. How long ago? About five years. As Tom said: there’s your problem.
Fifteen years ago your options were Wi-Fi, Bluetooth, and Zigbee. Batteries died every six to twelve months if you were lucky, and signal propagation was a nightmare — especially with freezers nested inside coolers. An early adopter told Tom that on a good day he’d have 75% of his sensors reporting in.
What changed is a protocol called LoRaWAN — an open standard that came out of Europe for utility metering, where a single access point covers a huge area. Applied to restaurants, it erases the old problems. Battery life up to 10 years. Reporting rates of 99.99%. Tom described a large chain whose requirement was 99% of sensors reporting at any given time; his team’s response was essentially, “We can add two more nines to that and still hit your metric.”
And it barely touches your network. A good LoRaWAN deployment uses less than half a gig of data a month — with hundreds of sensors across the enterprise. Download one song on Spotify and you’ve used more data in a day than the whole IoT network uses. When Tom sits down with network managers, they look at him wondering why he’s even bringing it up. Stamp it approved, move on.
Security is the other question he gets, and it’s a good one. Remember the breach where an HVAC contractor plugged a malware-laden USB into a rooftop unit that had an IP address on the network — and it worked its way into payment card data and cost a major retailer a nine-figure settlement? LoRaWAN sensors don’t carry an IP address. Each one has its own 128-bit encryption key, the link from sensor to gateway is encrypted, and the hop from gateway to cloud is encrypted again. Tom’s analogy: it’s like changing a car’s license plate and then driving it into a box truck — double-encrypted in transit. And even if someone cracked both layers? Congratulations, you now know the walk-in is at 38 degrees.
Where this is going: data first, then AI
Here’s the line from Tom I’m going to clip and use everywhere: your AI experience as a restaurant is only as good as the data you put into it. Everybody wants AI to solve everything, but AI on top of bad data — or no data — is just an expensive guess.
Good data makes the response actionable, and increasingly automatic. Alerts route to the store manager: shut the walk-in door; your fryer’s total polar materials hit 24, change the oil. Some clients escalate to a lockout — warn, warn, warn, then shut the equipment off if nobody acts. My favorite example: a client’s rapid-cook oven kept throwing an error code. The manager would call the service desk, who’d dispatch a technician, who’d drive out and discover the fix was to turn the oven off and back on again. Hundreds of dollars and real downtime to flip a switch. Now the first response to that trigger is an automatic power cycle. Layer AI on top and the system can recommend a fix, or dispatch a tech to the store, before the manager even knows there’s a problem.
The bigger prize is preventative. A manager will never notice that an electric heating element is quietly drawing more power than it used to — but that’s a leading indicator of failure. The right sensor catches it, triggers the service call, and the element gets replaced before the oven dies the day before Thanksgiving when the pies are your whole business.
Tom did flag one caution on the future, and I’m 100% with him. A CIO he heard recently warned that if you use AI to eliminate the customer interaction, you’re risking the very thing that differentiates your restaurant. The right move is AI that augments your people — vision intelligence watching the cameras that nobody’s staring at (queue length, slip-and-fall risk, the cash-refund fraud we opened with), IoT handling the mundane, and your team freed up to do what humans do best. Is the best use of a manager’s time writing “38” on a clipboard, or engaging a guest and having the judgment to react when the walk-in reads 40?
The takeaways
If you take nothing else from the conversation with Tom, take these five:
- Lead with the return. Labor reallocation, food-waste avoidance (~$2,000/year per cooler failure risk), and ~9% energy savings from a two-degree tune-up are real dollars — and at restaurant margins, a dollar saved is worth about ten in sales.
- Your clipboard is a liability. Paper logs do nothing but satisfy an inspector, and they fall apart the moment their reliability is questioned. Digital data can’t be back-filled and actually informs decisions.
- Judge IoT by what it is today, not what it was. LoRaWAN killed the old battery, reliability, and network problems — 10-year batteries, 99.99% reporting, less than half a gig a month, and encryption that makes the data useless to steal.
- Data first, then AI. AI is only as good as the data feeding it. Get the sensors right and the responses become automatic — and increasingly preventative.
- You don’t have to boil the ocean. Start with one or two “hero” use cases that pay for the network and infrastructure, then add on. As Tom’s former guest line goes: it’s not a proof of concept, it’s a proof of value.
Huge thanks to Tom Woodbury for coming on and educating me and all of you. You can find MachineQ at machineq.com — they’re part of Comcast Business — and Tom’s happy to be reached directly at thomas_woodbury@comcast.com to figure out how to make your restaurant as efficient as possible.
If you haven’t subscribed yet, do it on your favorite podcast player or on YouTube, and jump on the monthly newsletter while you’re at it. Until next time — make it a great day.
