How Much Does Building Fault Detection Software Cost in 2026?

The bands a fault detection build falls into

Under $90,000 you are building reporting and workflow around a packaged analytics product you already licence, typically pushing findings into your maintenance system and adding portfolio reporting. That is a legitimate and often sensible spend. Between $90,000 and $200,000, over 14 to 20 weeks, you get a first release proven on a pilot group of five to ten buildings: edge collection across your controls vendors, a normalisation pipeline into a semantic equipment model with human review, a core fault rule library, and a triage queue ranked by estimated cost rather than by severity label. Between $250,000 and $600,000, phased over 8 to 14 months, you add maintenance system integration with round trip verification that reopens a fault when the behaviour returns, weather normalised measurement and verification, tenant comfort correlation, contractor performance reporting and portfolio analytics.

The reason the first release is defined as a pilot rather than a rollout is hard won. Turn on a decent rule library across a large portfolio and you will get thousands of faults in the first week. Most will be real. Almost none will get fixed, because a facilities team with fixed headcount cannot triage an unrankable list. Prove prioritisation on eight buildings first, then scale.

What drives a fault detection build up

• Controls diversity and vintage. Each additional protocol or gateway is real integration work. BACnet over internet protocol is straightforward. BACnet over a serial trunk, Modbus devices with no useful naming, and older LonWorks installations behind a gateway are three separate problems. Buildings without a supervisory layer are markedly harder than buildings with one.
• Point count and polling interval. Fifteen minute data on five thousand points and one minute data on fifty thousand points are different engineering problems, not the same system with a bigger bill.
• Site network access. Frequently the slowest item on the plan and almost never for software reasons. Security review, firewall changes and remote access approvals are calendar time you have to budget.
• Central plant complexity. Chiller and boiler plant rules are considerably more demanding than air side rules, and plant sequencing faults need higher resolution data to detect reliably.
• Equipment record quality. A portfolio without reliable equipment inventory will spend real time simply establishing what is installed, and that work is unavoidable because the rules need to know what they are looking at.

Building count matters less than people assume once the vendor mix is fixed. Adding the fortieth building of the same type as the first thirty is an onboarding cost, not a development cost.

What keeps the number down

Pilot on five to ten buildings that represent your worst controls diversity rather than your best. If normalisation and collection work on the awkward sites, the rest of the rollout is predictable. Piloting on your newest building proves nothing.

Start at fifteen minute trend data. It catches most scheduling, setpoint, economiser and simultaneous heating and cooling faults, and it keeps both storage cost and controls network load reasonable. Apply one to five minute collection selectively, to equipment where a control loop or valve fault would be expensive.

Adopt an existing tagging standard such as Haystack or Brick rather than inventing your own. It costs nothing extra during the build and it makes every future integration cheaper, including any decision to move to a packaged product later.

Build the rule library incrementally. Ten well tuned rules that produce actionable findings beat sixty rules that produce noise, and the tuning only happens once real engineers work the queue.

And defer measurement and verification to phase two. Estimated avoided cost is enough to rank work. Defensible measured savings needs weather and occupancy normalisation against a baseline period, and that work is worth doing properly rather than quickly.

A worked example that adds up

A portfolio operator with roughly 60 commercial buildings across three controls vendors, about half with a Niagara supervisory layer and half without, in house engineering staff who self perform maintenance. First release, proven on eight buildings.

• Discovery, controls survey across the estate and selection of a tagging standard: 3 weeks, $18,000.
• Edge collectors covering BACnet over internet protocol, BACnet over serial trunk and Modbus, with local buffering through network outages and health telemetry reported as monitored data: 4 weeks, $32,000.
• Time series storage and query sized for fifteen minute data across roughly 30,000 points: 2 weeks, $16,000.
• Normalisation pipeline with automated suggestions from names, units, ranges and behaviour, a technician confirmation step and versioned mappings: 5 weeks, $40,000.
• Core fault rule library covering scheduling, setpoint drift, economiser operation, simultaneous heating and cooling, and sensor plausibility: 3 weeks, $24,000.
• Cost ranked triage queue with symptom grouping and suppression during commissioning or planned shutdown: 3 weeks, $22,000.
• Pilot rollout across eight buildings and engineer training: 2 weeks, $13,000.

That totals $165,000 and about 22 weeks of effort, delivered in 19 calendar weeks with three developers. Onboarding each additional building after the pilot typically runs $1,500 to $4,000 depending on point count and how much normalisation the naming convention resists.

How the spend phases

Phase zero is discovery at $12,000 to $20,000 over three weeks, and it has to include a controls survey rather than a questionnaire. What the estate documentation says is installed and what is actually installed diverge in almost every portfolio, and the difference is the schedule risk.

Phase one is collection, normalisation, the rule core and triage, proven on the pilot. Budget 45 to 55 percent of first year spend here. Do not roll out to the whole portfolio until the engineers working the pilot queue tell you the ranking is right, because the ranking is the product.

Phase two is the closed loop: work order creation in the maintenance system your technicians already use, with evidence and point references attached, and automatic re evaluation after the work order closes so a fault reopens if the behaviour returns. This is the phase that separates a system that pays for itself from a dashboard.

Phase three is measurement and verification, comfort correlation and contractor performance reporting. Sequence contractor reporting last and deliberately, because it is the report that changes behaviour and it needs clean data behind it before you put it in front of a service provider.

Roll out buildings in waves after phase one, funded as an operating cost rather than as project scope.

The ongoing costs nobody quotes

Infrastructure scales with points and interval rather than with buildings. For a portfolio of this size at fifteen minute resolution, expect $600 to $3,000 a month, rising sharply if you move a large share of points to one minute collection. Design retention tiering early: recent data hot, older data compressed but queryable.

Edge hardware is a per site capital and replacement cost, commonly $800 to $2,500 per building depending on how many networks you have to reach. It also needs a maintenance path, because a collector that quietly went offline produces silence that looks exactly like a portfolio with no faults.

Mapping maintenance is the recurring cost specific to this category. A controls upgrade that renames half a building's points has to be a versioned diff rather than a crisis, and somebody has to review it. Budget engineering time for this every time a site is retrofitted.

Support and enhancement runs 15 to 20 percent of build cost a year, and in this category most of it goes into rule tuning rather than defects.

The largest ongoing cost is not software at all. It is the engineering capacity to act on findings, and a portfolio that adds detection without adding capacity has bought a longer list.

Comparing a build against your current renewal

Packaged analytics is usually priced by building or by point, often with an onboarding fee per site through an integrator, so get both numbers from your current provider before comparing anything.

Multiply the recurring licence across five years for your full portfolio, not just the buildings live today, and add the onboarding fee for every building you intend to bring on. That second number is the one that grows quietly, and it is the reason large portfolios with heterogeneous controls end up pricing a build in the first place.

Against that, put $165,000 of first release, $1,500 to $4,000 per building of onboarding you perform yourself, infrastructure, edge hardware and 15 to 20 percent a year.

Then weigh the part that is not on either invoice. The normalised model of your portfolio is years of encoded engineering knowledge about your specific buildings, and whether you own it or rent it determines what a future migration costs. If you are a service provider delivering analytics to clients, that model is also part of what you are selling, which changes the calculation from cost to margin.

When buying beats building

Buy if you operate fewer than about 15 buildings. SkySpark deployed through a competent integrator, or Clockworks Analytics or KGS Buildings with their existing fault libraries, will find your faults and cost far less than building. Their libraries reflect years of practice you would otherwise develop from scratch.

Buy if you have no in house engineering capacity to act on findings. The constraint in that case is not detection, and adding a system will produce a longer list nobody works. Fix the capacity question first.

Buy Facilio or Switch Automation if you want analytics inside a broader operations platform and you are content to work the way those products work. That is a reasonable trade and it is cheaper than owning everything.

Build when several of these are true. Your portfolio is large enough that a percentage point of energy cost is a serious number. Your controls estate is heterogeneous enough that per building onboarding pricing has become a significant recurring cost. You self perform maintenance and want findings inside your own dispatch workflow rather than in a portal your technicians will not open. You deliver analytics to clients as part of your own service offering. Or you have already tried a packaged deployment and it stalled at the point where thousands of faults met a team of six.

If you want a second opinion before signing anything, Digital Heroes builds and runs its own products, so the people choosing your architecture live with those decisions on their own revenue. The document is yours whichever way you go.