How to Choose Sensor Intelligence Solutions (2026)

How to choose sensor intelligence for smart buildings

Sensor intelligence solutions read the data your buildings already produce (BMS points, energy meters, IoT sensor feeds) and turn it into decisions: what to fix, what to adjust, what to prove to an auditor. That is the category definition, and it is worth holding onto, because much of what gets sold under the label is something else. Usually a dashboard with better graphics.

This guide is for the people doing the choosing: facility managers, asset owners, and energy managers who have sat through three vendor demos that all looked the same. It covers the criteria that separate platforms once the demos blur together, what to check when a tool is pitched as BMS analytics software, and how to run an evaluation that ends in a decision instead of a longer shortlist.

If you want the wider category tour first, our smart building AI buyer’s guide covers what the technology does. This guide assumes you know what it does and need to pick one.

What counts as sensor intelligence

The term gets stretched, so draw the boundary before you shortlist anything.

A BMS is not sensor intelligence. It controls equipment and raises alarms, and a good one does both reliably, but it does not judge, rank, or explain. An analytics dashboard is not sensor intelligence either. It reorganises the same readings into charts and hands the interpretation back to you.

Sensor intelligence sits above both. It ingests readings continuously, models what each building should be doing, notices when reality diverges, and puts the divergences in front of you ranked by what they cost. The test is simple: does the system produce decisions someone can act on this week, or views someone has to study? If the answer is views, you are buying a dashboard, whatever the label says.

The six criteria that separate vendors

Once the shortlist is real, these six criteria separate the platforms that work in production from the ones that work in demos.

1. What it does with bad data

Ask what the platform does with bad data, because it will get plenty. Sensors drift, meters drop offline, a BMS point gets renamed during a service visit. A platform that reasons on top of unvalidated readings produces confident nonsense, and the failure is silent: the queue still fills and the numbers still look plausible, right up until a technician gets sent to a fault that turns out to be a dying sensor.

The platforms worth shortlisting validate first and reason second. They flag drift, mark gaps instead of smoothing them over, and tell you when a conclusion rests on a sensor they no longer trust. In the demo, ask to see what happens when a meter goes dark. If the answer is a shrug, keep looking.

2. Explanations you can check

Every insight should survive the question “where did that come from?” A recommendation that traces back to specific readings at specific times can be verified and defended to a sceptical chief engineer. A recommendation from a black box can only be believed or ignored, and in buildings it gets ignored fast.

This matters twice over when the vendor talks about AI. Ask whether the system can assert a number it never measured. For the class of tools worth buying, the honest answer is no. This is where grounded inference earns its keep: the principle that a system only claims what its data can back, and flags the gap when it cannot.

3. Decisions across the portfolio, not views per building

A platform that works per building does not survive a portfolio. Thirty buildings with thirty dashboards is thirty places to look, which in practice means nowhere. The evaluation question is whether the system can weigh a fault in building 4 against an energy drift in building 17 and tell you which one costs more by Friday.

Portfolio-wide actionability means one queue, ordered by impact, across every site. It also means each item carries its reasoning, so a regional facilities lead can hand item three to a local technician with the evidence attached. If understanding an insight requires opening that building’s view, the tool gets used for a month and then quietly abandoned.

4. Compliance evidence built in

If your portfolio carries BREEAM, LEED, or Nordic Swan certifications, or sits in scope for the EPBD’s building automation requirements or CSRD reporting, the platform should produce evidence continuously. Timestamped, traceable, exportable records are what those frameworks increasingly expect, and retrofitting an audit trail onto a tool that was never built to keep one is miserable work.

Check whether compliance tracking runs continuously or amounts to an annual export someone assembles by hand. The difference decides whether your next audit is a query against the record or a month of spreadsheet archaeology.

5. Retrofit planning fit

Sooner or later the data points at capital expenditure: a chiller past its efficiency floor, an AHU worth replacing rather than repairing again. The better platforms let you test the decision before you spend. What would a retrofit, a setpoint change, or a new schedule do to consumption and comfort? Simulated against your building’s measured behaviour, that question gets a defensible answer.

Weight this criterion if retrofit planning is on your horizon, and under the EPBD it probably is. A what-if simulation built on your own data beats an engineering estimate borrowed from a different building.

6. Works with the buildings you have

The right platform reads what is already installed: the BMS, the meters, the IoT sensors someone fitted three years ago. Be suspicious of any deployment plan that starts with new hardware, or worse, replacing the BMS. The data needs to leave the building. The systems that produce it should stay.

Then ask where the data lives once it leaves. For European portfolios, EU hosting under GDPR is a due-diligence requirement, and your data protection officer will raise it even if you don’t.

Vendor lock-in belongs on the same checklist. If leaving the platform means losing the history, you are renting your own building data back.

Evaluating BMS analytics software specifically

A large share of the category pitches itself as BMS analytics software: tools that sit on top of the building management system and analyse what it sees. The six criteria above still apply, and three more questions matter.

First, vendor breadth. A portfolio rarely runs one BMS make. If the analytics layer only speaks one vendor’s dialect, you are buying a per-estate tool and will be back in procurement within two years.

Second, point mapping. Ask who does the work of mapping thousands of BMS points into the platform’s model, how long it takes, and what it costs. This is where deployments stall, and vendors know it, which is why the demo rarely mentions it.

Third, reach beyond the BMS. A building management system sees a lot, but meters and standalone IoT sensors see more, and an analytics layer that stops at the BMS boundary leaves energy and compliance value on the table.

One naming note while you compare: the market uses BMS analytics, building analytics, and fault detection and diagnostics (FDD) almost interchangeably. Judge tools by what they ingest and what they output. The label tells you very little.

What energy managers should weight differently

If your title says energy manager, the criteria shift. Data quality and explainability still lead, but three capabilities move up the list.

Forecasting with confidence bounds comes first. A single-number forecast is a guess with good posture. A forecast with an honest range tells you when consumption is off-plan and when it is just inside normal variation, which is the difference between chasing noise and catching drift early.

Then ranking by cost. Across a portfolio, opportunities should be ordered by money and kilowatt hours, not by alarm severity, so the budget conversation writes itself.

And reporting-grade output. Scope 2 figures built from measured, timestamped consumption survive an auditor; figures built from flat annual factors increasingly do not. The energy manager is more and more often the person feeding the sustainability report, so the platform’s evidence quality becomes your evidence quality.

We wrote up the portfolio version of this problem on our energy management for commercial buildings page.

Running the evaluation

Shortlists are cheap. Evaluations cost weeks, so structure them.

Start from a comparison rather than a search engine. Our roundup of the best building energy management software names where each major platform is strong, including where competitors beat us, and gets you to three or four candidates quickly.

Then insist the demo runs on your data, not the vendor’s demo building. Any platform looks clairvoyant on a dataset it was tuned for. Give each candidate the same export from one real building, a messy one for preference, and see who finds something you didn’t already know. You are buying the finding, not the interface.

Finally, pilot small and read-only. One building, a few weeks, no control integration. A short pilot on live data answers the questions a slide deck cannot: how the platform handles your data quality, whether its queue earns your team’s trust, and whether anyone still opens it in week three without being reminded to.

All six criteria are versions of the same question: will your team trust what the system tells them? Validated data, explanations you can check, evidence an auditor accepts, a queue your technicians act on. Buy the platform your team still believes in month six. That is the one that ends up saving the money.

Want to see this on your own building? Book 20 minutes and we will show you Explore on a sample of your data. Not ready for a call? Get your Building Intelligence Score and see where your portfolio stands today.

Frequently asked questions

What are sensor intelligence solutions?
Software that continuously reads the data a building already produces, from BMS points to energy meters to IoT sensors, then detects, ranks, and explains problems and opportunities as a prioritised queue of decisions. The category differs from dashboards, which visualise the same data but leave the interpretation to you.

How is sensor intelligence different from a BMS?
A building management system controls equipment and raises alarms in the moment. Sensor intelligence sits on top of it, cross-references what the BMS and other sources are reporting, and judges what matters most across the whole portfolio. It complements a BMS rather than replacing it, which is why deployment should never require touching the control layer.

Do we need to install new sensors first?
Usually not. Most commercial buildings already produce more data than anyone reads, through the BMS, utility meters, and existing IoT devices. Start the evaluation with the data you have, and add hardware only where the pilot shows a gap worth closing.

What questions should we ask in a vendor demo?
Ask what happens when a sensor drifts or a meter drops offline. Ask where any given number came from and whether the system can assert something it never measured. Ask how insights are ranked across buildings, whether compliance evidence exports cleanly, and whether the demo can run on your own data. Vendors comfortable with those questions are usually comfortable in production too.

How long should a pilot take?
A few weeks on a single building, connected read-only, is normally enough to see the things that matter: how the platform handles your data quality, whether its recommendations hold up when your engineers check them, and whether the team keeps using it unprompted.