A framing crew in Houston installed R-13 batts in every exterior wall cavity of a 2,400-square-foot production home in 2015. Passed the code inspector, passed the HERS rater, met the R-value requirements of the 2015 IECC to the letter. The insulation was also compressed behind electrical boxes in eleven cavities, gapped along the top plates in six more, and misaligned against the sheathing in a way that created convective loops in both bathrooms. The right amount of insulation, installed wrong, in a home that would bleed heat through its walls for the next thirty years while the energy bill told the homeowner everything was fine because they had nothing to compare it to.
That home was one of hundreds inspected in the SPEER/EEP Houston field study, which found that 39 percent of wall insulation installations in new Texas homes failed to meet RESNET Grade I quality standards. Not missing insulation, not the wrong product. Correctly specified material, incorrectly stuffed into cavities by crews working fast on a Thursday afternoon.
Why Nobody Catches It
Drywall goes up fast. On a production home, the gap between insulation completion and drywall hang is measured in days, sometimes hours on a tight schedule, and once those sheets are screwed to the studs the only way to verify what is behind them involves cutting holes in finished walls, which nobody does unless water is pouring through the ceiling. Code inspectors check that insulation exists and meets R-value spec. They do not squeeze behind batts to verify contact with the air barrier. They do not measure compression ratios at wiring penetrations. A visual inspection of fiberglass in a cavity looks correct from four feet away even when it is performing at 60 percent of rated value due to installation defects that would take less than ten minutes to fix.
Thermal imaging changes the math. An infrared camera sees temperature differentials through insulation, revealing compression, gaps, voids, and thermal bridges that are invisible to the naked eye. A trained thermographer with a $2,000 FLIR E-series camera has been able to do this for decades. What changed is that a FLIR ONE Pro phone attachment now costs $300, and AI-powered analysis software can flag anomalies that would previously require an ASNT Level II certification to interpret.
What AI Adds to a Camera
Thermal cameras produce data. Interpreting that data is the hard part.
A stud shows up as a cold stripe in winter because wood conducts heat faster than the insulation beside it. Normal. A compressed batt shows up as a slightly warmer stripe because its effective R-value has been reduced by the compression. Also normal, also a defect, and the difference between the two patterns is subtle enough that an untrained contractor holding a FLIR will misidentify structural thermal bridging at studs as insulation failures roughly a third of the time. AI changes this by training on thousands of labeled thermal images to distinguish normal structural patterns from actual installation defects, and a 2026 study published in MDPI Sensors using deep learning on infrared thermography achieved 90.6 percent precision and 91.4 percent recall for building envelope defect classification using a YOLO-based architecture.
That 90.6 percent precision means roughly 1 in 11 flagged defects is a false positive. Annoying but not catastrophic, because you pull the batt, check, put it back, and lose five minutes. The alternative is missing the defect entirely and discovering it never, because the homeowner will attribute their high energy bill to the HVAC system, the windows, or the weather, in that order, and never to the insulation that passed inspection.
Lamarr.AI, a Georgia Tech spinoff led by associate professor Tarek Rakha, combines drone-mounted thermal cameras with AI to create 3D defect maps and energy simulations that quantify how much energy each defect wastes in dollars per year. They raised $1.1 million in pre-seed funding from Hazelview Ventures in October 2024 and are targeting multi-family and commercial buildings, though the underlying technology transfers directly to single-family residential at a fraction of the cost because you do not need a drone to scan a 2,400-square-foot home. You need a phone, a $300 attachment, and twenty minutes.
The Money, Specifically
Here is the ROI calculation nobody in the insulation industry publishes, because the numbers are embarrassing.
| Scenario | Cost | Payback |
|---|---|---|
| DIY scan (FLIR ONE Pro, reusable) | $300 one-time | First defect found |
| Professional thermal scan | $400 per home | < 3 years on energy savings alone |
| Fix insulation pre-drywall | $200 to $500 | Immediate vs. post-drywall |
| Fix insulation post-drywall | $2,000 to $5,000+ | Demo, reinstall, repaint |
A builder scanning 10 homes per year with a $300 FLIR ONE Pro will statistically encounter defects in 4 of them, based on the 39 percent failure rate. Fixing those defects pre-drywall costs roughly $350 each, or $1,400 total. Fixing them after drywall would cost $2,000 to $5,000 per home in demolition, reinstallation, and finishing, which means the builder avoids $8,000 to $20,000 in potential callbacks and warranty claims for a $1,700 investment. The camera pays for itself on the first defective home. Every subsequent catch is pure margin protection.
For homeowners, the EPA estimates that proper air sealing and insulation saves 15 percent on heating and cooling costs, which translates to roughly $180 per year on a typical $1,200 annual HVAC bill. A $400 professional thermal scan that catches a Grade II installation defect pays for itself in under three years on energy alone, before you factor in the $2,000+ avoided remediation cost. The 2026 Angi pricing data puts full-home thermal scans at $300 to $500, with partial scans starting at $150.
The Catch
Thermal imaging requires a temperature differential between indoor and outdoor air. The standard minimum is a 15-degree Fahrenheit differential between indoor and outdoor air. In Houston in July, the delta between a 75-degree interior and a 100-degree exterior gives you 25 degrees to work with. Perfect. In San Diego in October, you might have 72 inside and 74 outside. Useless. This means thermal scans during construction are seasonally constrained in mild climates unless the builder runs a blower door to pressurize the building and create artificial temperature differentials, which adds $200 to $300 to the scan cost and requires a door to already be installed.
Consumer-grade cameras also have lower thermal resolution than professional units. The FLIR ONE Pro shoots at 160 by 120 pixels in the thermal band, while an FLIR E8-XT professional camera delivers 320 by 240. That resolution gap matters when you are trying to distinguish a compressed batt from a thermal bridge at a stud, and it matters more in the AI analysis pipeline because the deep learning models trained on high-resolution commercial imagery may degrade on consumer-camera inputs in ways nobody has benchmarked yet. No peer-reviewed study has validated AI thermal analysis specifically for residential new construction insulation using consumer-grade hardware. The 90.6 percent accuracy figure from the MDPI study was achieved on high-rise facade imagery with professional equipment.
Then there is the labor problem the camera cannot solve. A FLIR image that shows compressed insulation in cavity 47 is only useful if someone fixes cavity 47, and the crew that installed it wrong the first time is usually the same crew asked to fix it. The camera diagnoses, but it does not treat. The 39 percent defect rate exists because installation quality is a supervision problem, not a detection problem, and adding a camera to a process without adding supervision is an expensive way to generate a defect list that nobody acts on.
If You Are Building a Home
Request a pre-drywall thermal scan. It is the single highest-ROI inspection you can add to a new construction project, and most builders will not object because it costs them nothing if their insulation is installed correctly. If they object, that tells you something. Budget $400 for a professional scan or $300 for a FLIR ONE Pro that you can reuse during the build for window seal verification, HVAC duct leak detection, and post-occupancy comfort checks.
Do not assume the code inspector caught installation quality problems. Code inspectors verify compliance with minimum R-value requirements but do not grade installation quality. ENERGY STAR certification requires Grade I, but conventional code does not, and most production homes are built to code, not to ENERGY STAR standards.
If you are a builder running more than five homes per year: buy the camera. At $300 it is the cheapest quality assurance tool in your truck. Use it on every home after insulation and before drywall. Photograph the thermal images alongside the visible-light cavity shots and include them in your warranty documentation. This protects you from callbacks and gives your buyers evidence that their walls are performing as designed, which is a sales advantage nobody in production homebuilding is currently exploiting because the industry has collectively decided that insulation installation quality is someone else's problem.
Limitations of This Analysis
The 39 percent defect rate comes from the SPEER/EEP study of new construction in Texas, climate zones 3 and 4. Cold-climate builders may achieve better installation quality because the stakes of poor insulation are more immediately visible in heating costs and ice dam formation, though no comparable field study exists for northern markets. The DOE's $100 billion envelope waste figure covers all building types, not residential alone. AI thermal analysis accuracy figures are drawn from high-rise facade research using professional equipment, and performance on consumer-grade cameras in residential framing environments has not been independently validated. Cost calculations use average EIA household energy data and ORNL insulation performance estimates. Actual savings depend on climate zone, fuel prices, home size, and the specific nature and location of installation defects. Lamarr.AI's residential application is inferred from their technology platform rather than demonstrated in published residential case studies.