What is FDD, and is it worth it for a mid-size building?
Every building has faults the BAS doesn't alarm on: a valve leaking by, an economizer that never opens, a VAV box fighting the one next to it. Fault detection and diagnostics (FDD) software is built to find them. The real question is whether anyone will act on what it finds.
What FDD actually does
FDD software pulls trend data from your BAS (temperatures, valve and damper commands, fan status, setpoints) and runs rules or models against it, usually every hour or every day. When the data shows a pattern that shouldn't happen, it flags a fault. Most products then rank the faults by estimated energy, comfort or equipment impact so you know which to chase first.
It doesn't replace the BAS. Alarms tell you something is broken right now, like a fan failure or high discharge temperature. FDD finds the slow, quiet problems that waste energy and generate hot/cold calls for months without ever tripping an alarm.
Typical things FDD catches
- Simultaneous heating and cooling: a heating valve and cooling valve both open on the same AHU, or reheat running hard downstream of over-cooled supply air.
- Leaking or stuck valves and dampers: coil temperature changes when the valve is commanded closed, or an outdoor air damper that doesn't track its command.
- Economizer problems: free cooling available but the unit is running mechanical cooling with minimum outside air.
- Sensor drift and failure: a mixed-air reading that's impossible given return and outside air, or a sensor flatlined at one value.
- Schedule and override waste: equipment running nights and weekends because someone put it in hand or forgot to release an override.
- Hunting loops: valves and dampers oscillating because of bad PID tuning.
What the public data says
The largest public dataset on this comes from Lawrence Berkeley National Laboratory's Smart Energy Analytics Campaign, which tracked analytics use across more than 100 organizations and over 6,500 buildings. Per the LBNL Smart Energy Analytics Campaign report:
- Organizations using FDD tools saw a median 9% annual energy savings (about $0.24/sq ft) by the second year, with a reported range of 1% to 28%.
- Median FDD cost was about $0.06/sq ft for installation and configuration and $0.02/sq ft per year in recurring software cost.
- Median in-house labor was about 8 hours per month per building to review and act on findings.
Two caveats from the report itself: savings were measured at the whole-building level, so other projects happening at the same time also contributed. And the participants chose to be in the program, so they were probably more motivated than the average owner.
The labor number is the one that matters
Look at that last line again. FDD doesn't fix anything; people do. If nobody has a few hours a month to read the fault list and send a tech, you're paying for a very sophisticated report that gets archived. The most common FDD failure isn't bad software. It's an unread inbox.
Before buying, decide who owns the fault list. That could be in-house staff, your controls contractor under a service agreement, or the vendor's own analysts (several vendors sell FDD as a managed service).
When it's probably worth it for a mid-size building
- The building has a real DDC system with decent trending: AHUs, VAVs, a central plant or RTUs on the BAS, not just thermostats.
- Somebody can act on findings within a week or two.
- Energy is expensive, or comfort complaints keep coming back, or you're about to recommission anyway.
- You manage more than one building and want the same view across all of them.
When to wait
- The BAS network is flaky, points are mislabeled, or half the sensors are bad. Fix the basics first, or the first year of FDD will mostly be "your data is garbage" findings.
- There's no budget or labor to fix what it finds.
- The building is small and simple enough that a good tech with trend logs and a free afternoon will find most of the same things.
How to pilot it without getting burned
- Pick one building with known problems so you can check whether the tool finds them.
- Ask what data the vendor needs: points, trend interval, history length, and whether they need write access. Read-only access is enough for FDD.
- Find out who does the point mapping or tagging and what it costs. Tagging standards like Project Haystack make this faster.
- Agree up front on what success looks like after 90 days: faults found, faults fixed, and verified savings or comfort improvements.
- Make sure you can export your data and fault history if you leave.
Our analytics buyer's checklist has the full question list.
