The category sells four different things under one name. Knowing which one you're buying is the whole job.
Search for an IoT platform for retail facility management and you'll get a dozen vendors who appear to sell the same product. The products differ more than the category page suggests. The phrase covers four distinct capability layers, and a given platform usually delivers one or two of them well, gestures at a third, and markets all four.
Sorting that out is the buyer's actual job. Here's a way to do it.
Sense. Collecting telemetry off equipment and spaces. Temperature, humidity, power draw, door state, equipment runtime. This is the foundation and it's where the demo tends to linger, because live data on a dashboard looks impressive.
Control. Changing equipment behavior remotely, on a schedule, across a whole portfolio, without sending a technician. This is a much smaller field than the marketing suggests.
Diagnose. Turning telemetry into a named condition. Not "site 47 used more power last week" but "the rooftop unit at site 47 is drifting toward failure." This is the layer where vendor language gets loosest, because "insights" and "analytics" can describe almost anything.
Resolve. Getting the named condition actually fixed, and confirming it stayed fixed. Very few platforms claim this honestly, GlacierGrid included.
Each layer depends on the one before it. But the reverse isn't true: a platform can sense beautifully and control nothing.
Question: what happens when a device goes offline, and how would I know?
Sensing failures are quiet. A sensor that stops reporting doesn't send an alert saying it stopped reporting, unless somebody built that. Ask how offline devices surface, how quickly, and to whom.
The answer that should worry you: a demo that shows only the happy path. If every screen in the demo has complete data, ask to see a site with a dead sensor.
Also ask what the sensing actually costs at scale, per point, per month. Sensing is priced per device, and a 200-site portfolio with four monitored points per site is 800 devices. A GlacierGrid door sensor, for instance, runs $10 per month per sensor. Whatever the vendor's number is, multiply it by your real device count before you get attached to the demo.
Question: can you change a setpoint at 400 sites tomorrow morning, and can you prove it took?
This is the question that separates the field, and the gap between "we connect to your thermostats" and "we set your thermostats" is where buyers get surprised. Read-only integrations are common. Write access that works reliably across a mixed fleet is not.
The answer that should worry you: any version of "it depends on your equipment" without a specific list. It does depend on your equipment, which is exactly why the vendor should be able to tell you which of your models they can write to and which they can only read.
Ask a second question here that buyers skip: what happens when 400 sites all recover their setpoints at the same moment? A fleet that snaps back in unison builds one enormous coincident demand peak and hands you a demand charge that eats the savings. Paced recovery is a real engineering problem. GlacierGrid handles it with Smart Setpoints and Adaptive Recovery. Ask whoever you're evaluating how they handle it, and be skeptical if the question seems new to them.
Question: show me a condition you named before the equipment failed, and tell me your false-positive rate.
Diagnosis is only useful if it's early and if it's trustworthy. Early without trustworthy produces alert fatigue, and a facilities team that has learned to ignore your alerts is worse off than one that never had them.
The answer that should worry you: dashboards presented as diagnosis. A chart showing consumption over time is sensing with better graphic design. Diagnosis names a condition, on a specific unit, and says what's wrong.
GlacierGrid's HVAC Diagnostic Dashboard flags units drifting toward failure, which is the shape of the thing to look for. Whatever platform you evaluate, ask for the named condition, not the trend line.
Question: after the alert fires and somebody does something, how does the system know the problem went away?
This is the honest weak point of the whole category, and it's worth naming plainly rather than pretending otherwise.
Across this space, vendors have spent years getting better at telling operators something is wrong. Far less attention has gone to closing the loop. An alert fires, a work order gets written somewhere else, a technician visits, something happens, and no system anywhere verifies that the condition actually cleared and stayed cleared. The operator finds out it didn't when the same alert fires again in six weeks.
The answer that should worry you: a confident yes with no mechanism behind it. Ask what specifically gets measured after the fix, and for how long.
A buyer's guide that quietly resolves to "and that's why you should pick us" isn't a buyer's guide. So here is where GlacierGrid actually sits against its own framework.
GlacierGrid controls HVAC and lighting. It monitors refrigeration and cold chain. That distinction is load-bearing and worth stating directly: we do not control refrigeration. Anyone evaluating us for that should know it before the demo, not after.
On diagnosis, the HVAC Diagnostic Dashboard names units drifting toward failure. On resolution, the category-wide gap described above applies to us as much as anyone.
Published benchmarks: roughly 10% energy savings, roughly a one-month payback, and 15% fewer service calls.
Ask every vendor on your list to draw the same map of themselves. The ones who can't, or won't, have told you something.
Take the four layers into your next three vendor calls. For each layer, ask the question, then ask for the mechanism behind the answer. Write down which layers each platform genuinely delivers.
The field sorts quickly. Nearly everyone senses. Fewer control. Fewer still diagnose in a way a technician would act on. And the resolve column will have more blanks in it than anyone's homepage suggests.
That map is worth more than any feature comparison, because it tells you what you're actually buying.
Want to see where a platform lands on your own equipment? Book a demo and bring your hardest site.