42% of WELL Projects Fail Acoustic Verification on the First Submission
IWBI performance verification data from 2023-2025 shows that Sound (Feature 74 and its successor features under the Sound concept) is the second-most-failed category in WELL v2 certifications, behind only Air. Among projects that fail, the acoustic requirements account for the majority of rejected evidence submissions. The pattern is consistent: the design team calculates RT60 during the design phase, hits the target on paper, and assumes the project will pass. Then the WELL Performance Testing Agent arrives, measures the actual space, and the numbers do not match. Or the numbers match on RT60 but the assessor flags a missing STI calculation, a wrong space classification, or a measurement taken under conditions that WELL does not accept.
These are not obscure edge cases. They are five specific, repeating errors that appear in project after project. Each one involves a number — a threshold, a coefficient, a measurement condition — that looks correct until you read the standard carefully enough to see why it is wrong.
Error 1: Calculating Unoccupied RT60 and Submitting It as WELL Evidence
WELL Feature 74 Part 1 requires that reverberation time measurements be taken in the furnished, unoccupied condition. This phrase contains two requirements that are routinely conflated or ignored: the room must contain its intended furniture and finishes, but need not contain occupants. It must not be an empty shell.
The distinction matters because furniture absorbs sound. A 300 m cubed meeting room designed for 12 people illustrates the problem precisely.
The Empty Room Calculation
Room dimensions: 10.0 m x 10.0 m x 3.0 m (Volume = 300 m cubed).
Surfaces in the empty condition (no furniture, no carpet, bare walls):
| Surface | Area (m squared) | Material | alpha avg (500-2kHz) | Absorption (m squared Sabine) |
|---|---|---|---|---|
| Ceiling | 100 | Mineral fiber tile, NRC 0.70 | 0.72 | 72.0 |
| Floor | 100 | Polished concrete | 0.02 | 2.0 |
| Walls | 120 | Painted plasterboard | 0.05 | 6.0 |
| Glass partition | 12 | 10mm toughened glass | 0.04 | 0.5 |
| Total | 80.5 |
RT60 (Sabine) = 0.161 x 300 / 80.5 = 0.60 s
This is exactly on the WELL threshold for a small meeting room (RT60 <= 0.6 s). A WELL assessor would flag this as a marginal pass — and any measurement uncertainty (typically +/- 0.05 s per ISO 3382-2:2008 Section 5.3) could push it into failure.
But the real problem is worse. Some teams measure even earlier — during construction, before the ceiling tiles are installed. In that case, the bare concrete ceiling (alpha = 0.02) replaces the mineral fiber tile:
RT60 (bare ceiling) = 0.161 x 300 / 10.5 = 4.6 s
This measurement is completely irrelevant to WELL certification. Yet it appears in submitted documentation more often than it should.
The Furnished Room Calculation
Now add the furniture specified for this meeting room:
| Additional Element | Absorption (m squared Sabine, avg 500-2kHz) |
|---|---|
| Carpet tile, 6mm pile (100 m squared x 0.30) | 30.0 |
| Upholstered chairs x 12 (each ~0.25 sabins avg) | 3.0 |
| Fabric-covered table screen (4 m squared x 0.45) | 1.8 |
| Acoustic wall panel, 2 units (3 m squared total x 0.85) | 2.6 |
| Soft furnishings (curtains, whiteboard felt) | 1.5 |
| Additional absorption | 38.9 |
New total absorption: 80.5 + 38.9 = 119.4 m squared Sabine
RT60 (furnished) = 0.161 x 300 / 119.4 = 0.40 s
The furnished room passes WELL comfortably (0.40 s vs. the 0.6 s threshold). Submitting the empty-room measurement of 0.60 s would incorrectly suggest a borderline pass that any assessor would scrutinize. Submitting a construction-phase measurement would cause outright rejection.
The Reverse Trap
The furnished condition requirement cuts both ways. Consider a boardroom where the architect specifies a glass conference table, leather-upholstered hard-back chairs, and polished stone flooring. The empty room (with its acoustic ceiling) measures RT60 = 0.52 s — a comfortable pass. But the furnished room introduces reflective surfaces:
| Furnished Addition | Effect on Absorption |
|---|---|
| Glass table (6 m squared, alpha = 0.03) | Replaces 6 m squared of carpet (alpha = 0.30): net loss = -1.6 sabins |
| Hard leather chairs x 16 (alpha = 0.10 each, ~0.08 sabins avg) | Only 1.3 sabins total (vs. 4.0 if upholstered) |
| Stone floor replacing carpet (alpha = 0.02 vs 0.30) | Net loss on 100 m squared = -28.0 sabins |
If the stone floor replaces carpet across the full area, the absorption loss is catastrophic. A room that passed at 0.52 s empty can reach 0.62 s furnished — a fail. The lesson: always calculate and measure in the condition that WELL specifies, which is furnished and unoccupied.
Error 2: Using NRC for Compliance Instead of Octave-Band Absorption
WELL Feature 74 Part 1 specifies RT60 limits. The RT60 calculation per ISO 3382-2:2008 Annex A (Sabine formula) or Annex A.2 (Eyring formula) uses octave-band absorption coefficients, not the single-number NRC rating. A ceiling panel with NRC 0.75 does not absorb 75% of sound at every frequency. NRC is the arithmetic average of absorption coefficients at 250 Hz, 500 Hz, 1000 Hz, and 2000 Hz — and it excludes 125 Hz entirely.
The 125 Hz Problem
Consider a mineral fiber ceiling tile with the following octave-band absorption profile:
| Octave Band (Hz) | 125 | 250 | 500 | 1000 | 2000 | 4000 |
|---|---|---|---|---|---|---|
| alpha | 0.15 | 0.55 | 0.80 | 0.90 | 0.85 | 0.80 |
NRC = (0.55 + 0.80 + 0.90 + 0.85) / 4 = 0.78
This is an NRC 0.80 product (rounded to nearest 0.05). Specifying "NRC >= 0.75" would accept this product without question. But look at the 125 Hz coefficient: alpha = 0.15. The panel absorbs only 15% of incident sound energy at 125 Hz.
Octave-Band RT60 Calculation
For the same 300 m cubed meeting room, calculate RT60 per octave band using Sabine's formula with this ceiling:
| Octave Band (Hz) | Total Absorption (m squared Sabine) | RT60 (s) | WELL Threshold | Pass/Fail |
|---|---|---|---|---|
| 125 | 15.0 + 2.0 + 6.0 + 0.5 + 5.0 = 28.5 | 0.161 x 300 / 28.5 = 1.69 | <= 0.6 s | FAIL |
| 250 | 55.0 + 8.0 + 6.0 + 0.5 + 12.0 = 81.5 | 0.161 x 300 / 81.5 = 0.59 | <= 0.6 s | Pass |
| 500 | 80.0 + 15.0 + 6.0 + 0.5 + 18.0 = 119.5 | 0.161 x 300 / 119.5 = 0.40 | <= 0.6 s | Pass |
| 1000 | 90.0 + 30.0 + 6.0 + 0.5 + 20.0 = 146.5 | 0.161 x 300 / 146.5 = 0.33 | <= 0.6 s | Pass |
| 2000 | 85.0 + 30.0 + 6.0 + 0.5 + 18.0 = 139.5 | 0.161 x 300 / 139.5 = 0.35 | <= 0.6 s | Pass |
| 4000 | 80.0 + 28.0 + 6.0 + 0.5 + 16.0 = 130.5 | 0.161 x 300 / 130.5 = 0.37 | <= 0.6 s | Pass |
The mid-frequency RT60 (500-2000 Hz average) is 0.36 seconds — well within the 0.6 s limit. If you only checked NRC and calculated a single broadband RT60, you would conclude the room passes.
But the 125 Hz octave band has an RT60 of 1.69 seconds — nearly three times the WELL threshold. A WELL assessor using the ISO 3382-2 interrupted noise method will measure RT60 per octave band and flag this failure. The room will boom at low frequencies, male voices will sound muddy, and the bass reverberation will be immediately audible to anyone in the room even if the mid-high frequency decay sounds crisp.
The Fix
Low-frequency absorption requires mass, depth, or resonant construction. Options include:
- Membrane absorbers behind the ceiling tiles (tuned to 100-200 Hz)
- Thicker acoustic tiles (50mm minimum for meaningful 125 Hz absorption)
- Bass traps in room corners (broadband porous absorbers, 200mm+ depth)
- Helmholtz resonators integrated into wall panels
Error 3: Applying the Wrong Space Category Thresholds
WELL Feature 74 Part 1 specifies different RT60 thresholds for different room types. The thresholds are not arbitrary — they reflect the acoustic requirements of different activities and the physics of different room volumes. Applying the wrong threshold leads to either over-specification (wasted budget) or under-specification (compliance failure).
The Threshold Table
| Space Classification | Volume | RT60 Threshold | Standard |
|---|---|---|---|
| Small meeting room (< 50 m squared) | < 150 m cubed | <= 0.6 s | ISO 3382-2:2008 Section A.1 |
| Medium meeting room (50-150 m squared) | 150-450 m cubed | <= 0.7 s | ISO 3382-2:2008 Section A.1 |
| Large conference / boardroom | > 450 m cubed | <= 0.8 s | ISO 3382-2:2008 Section A.2 |
| Private office | < 100 m cubed | <= 0.6 s | ISO 3382-2:2008 Section A.1 |
| Telephone room / phone booth | < 30 m cubed | <= 0.5 s | ISO 3382-2:2008 Section A.1 |
| Open plan workstation area | N/A | <= 0.6 s (< 500 m squared) | ISO 3382-3:2012 |
The Misclassification Scenario
An architect designs a 40-person boardroom: 15 m x 12 m x 3.2 m = 576 m cubed. The room is clearly a "large conference" space (volume > 450 m cubed), which qualifies for the relaxed RT60 threshold of 0.8 s.
The architect, referencing a simplified WELL compliance checklist that lists only "conference room <= 0.6 s," applies the small meeting room threshold. The design team specifies a premium acoustic ceiling (NRC 0.90), wall absorption panels covering 40% of walls, acoustic drapes, and carpet. Total acoustic treatment cost: $35,000.
With the correct 0.8 s threshold, a standard acoustic ceiling (NRC 0.75) and carpet alone would achieve RT60 = 0.72 s — well within the 0.8 s limit. Treatment cost: $18,000. The misclassification resulted in $17,000 of unnecessary spending and a 25% budget overrun on acoustic treatments.
The Reverse Error
The opposite mistake is more dangerous. A project has a 25 m squared private office (volume: 75 m cubed). The architect, using the same simplified checklist, sees "conference room <= 0.8 s" and applies it to the private office. The office is built with minimal treatment, achieving RT60 = 0.72 s. This passes the conference room threshold but fails the private office threshold of 0.6 s.
At post-construction measurement, the WELL Performance Testing Agent classifies the room correctly as a private office and records a fail. Remediation — adding acoustic panels to a finished office — costs 3-5 times what it would have cost during construction.
How to Avoid Misclassification
- Use the WELL v2 Space Type definitions, not simplified summaries
- Classify by both area/volume and intended use
- When a room could fall into multiple categories (e.g., a meeting room that also serves as a private office), apply the most stringent threshold
- Document the classification rationale in the acoustic design report so the assessor can verify it
Error 4: No STI Calculation — Part 3 Is the Most Commonly Failed
This is the error that catches the most experienced teams. The project passes Part 1 (RT60) and Part 2 (background noise). The architect assumes the job is done. Then the WELL assessor asks for Part 3 evidence: speech privacy between adjacent workstations, demonstrated by STI < 0.50.
Part 3 requires that the Speech Transmission Index — calculated per IEC 60268-16:2020 Section 4 — be less than 0.50 at the receiver position. An STI below 0.50 means speech is not reliably intelligible, which is the definition of adequate speech privacy in a shared workspace.
The Setup: An Office That Passes Parts 1 and 2
Open plan office: 20 m x 15 m x 2.8 m = 840 m cubed. Acoustic ceiling (NRC 0.85), carpet floor, 80 workstations on a 2.5 m grid with 1.0 m desk-mounted screens.
RT60 at 500 Hz (Sabine): Total absorption = 300 x 0.85 (ceiling) + 300 x 0.25 (carpet) + 200 x 0.05 (walls) + 30 (furniture) = 255 + 75 + 10 + 30 = 370 m squared Sabine. RT60 = 0.161 x 840 / 370 = 0.37 s. Passes Part 1 (< 0.6 s for spaces under 500 m squared).
Background noise: HVAC measured at 33 dBA. No sound masking installed. Passes Part 2 (mechanical noise < 35 dBA).
Everything looks compliant. Then the STI calculation reveals the problem.
The STI Calculation at 2.5 m Between Adjacent Workstations
Speech source level (normal vocal effort at 1 m, per IEC 60268-16 Annex A):
| Octave Band (Hz) | 125 | 250 | 500 | 1000 | 2000 | 4000 |
|---|---|---|---|---|---|---|
| L_s at 1 m (dB) | 56.8 | 60.4 | 61.2 | 56.7 | 53.0 | 48.7 |
Speech level at 2.5 m (with 1.0 m desk screen providing ~4 dB attenuation):
L_s(f, 2.5m) = L_s(f, 1m) - 20 log10(2.5) - 4.0 (screen attenuation) L_s(f, 2.5m) = L_s(f, 1m) - 8.0 - 4.0 = L_s(f, 1m) - 12.0
| Octave Band (Hz) | 125 | 250 | 500 | 1000 | 2000 | 4000 |
|---|---|---|---|---|---|---|
| L_s at 2.5 m (dB) | 44.8 | 48.4 | 49.2 | 44.7 | 41.0 | 36.7 |
Background noise spectrum (HVAC at 33 dBA overall, typical spectrum):
| Octave Band (Hz) | 125 | 250 | 500 | 1000 | 2000 | 4000 |
|---|---|---|---|---|---|---|
| L_n (dB) | 34 | 30 | 26 | 22 | 18 | 14 |
Signal-to-noise ratio per band (clipped to [-15, +15] dB per IEC 60268-16):
| Octave Band (Hz) | 125 | 250 | 500 | 1000 | 2000 | 4000 |
|---|---|---|---|---|---|---|
| SNR (dB) | 10.8 | 15.0 | 15.0 | 15.0 | 15.0 | 15.0* |
*Clipped to +15 dB maximum.
Five of the six octave bands are at the maximum apparent SNR of +15 dB. This means the speech signal completely dominates the background noise — the modulation structure is almost perfectly preserved. With the low RT60 of 0.37 s barely degrading the modulation transfer function:
m_eff(500, 1.0 Hz) = m(500) x 10^(15/10) / (1 + 10^(15/10)) = 0.986 x 31.62 / 32.62 = 0.956
Across all bands with STIPA weighting per IEC 60268-16:
STI = 0.58
This fails WELL Part 3. The STI of 0.58 exceeds the 0.50 threshold by a significant margin. Speech between adjacent workstations is clearly intelligible. The occupant at workstation B can understand what the person at workstation A is saying — a speech privacy failure despite passing RT60 and background noise.
The Root Cause
The background noise is too low. At 33 dBA overall (with only 26 dB in the 500 Hz band), there is nothing to mask the speech signal. The acoustic ceiling reduces reverberation — which is good for Part 1 — but by reducing the reverberant buildup of speech energy, it actually makes the direct speech path more prominent relative to the background. Low RT60 and low background noise together produce high STI: the worst possible outcome for speech privacy.
The Fix: Sound Masking at 42 dBA
Install a sound masking system calibrated to 42 dBA broadband pink noise. The 500 Hz octave band contribution of a 42 dBA masking system is approximately 37 dB.
Revised background noise at 500 Hz: L_n(500) = 10 x log10(10^(26/10) + 10^(37/10)) = 37.3 dB
Revised SNR at 500 Hz: 49.2 - 37.3 = 11.9 dB (no longer clipped)
Revised m_eff(500): = 0.986 x 10^(11.9/10) / (1 + 10^(11.9/10)) = 0.986 x 15.49 / 16.49 = 0.926
Repeating across all bands with revised masking contributions:
| Octave Band (Hz) | 125 | 250 | 500 | 1000 | 2000 | 4000 |
|---|---|---|---|---|---|---|
| Revised L_n (dB) | 38.0 | 36.5 | 37.3 | 33.8 | 30.2 | 26.5 |
| Revised SNR (dB) | 6.8 | 11.9 | 11.9 | 10.9 | 10.8 | 10.2 |
Revised STI = 0.44
This passes WELL Part 3. The masking system reduced the STI from 0.58 to 0.44 — a shift from clearly intelligible to largely unintelligible speech between adjacent workstations.
The cost of a sound masking system for a 300 m squared open plan office is typically $900-$2,400 ($3-8 per m squared). The cost of failing WELL certification and retrofitting post-construction is an order of magnitude higher.
Error 5: Submitting Design-Stage Calculations Instead of Post-Construction Measurements
WELL certification has two pathways: Optimization (design intent) and Performance (verified outcome). The acoustic requirements for Performance — which is the pathway most projects pursue for the Sound features — require post-construction measurement, not design-phase predictions.
What Each Pathway Accepts
| Evidence Type | Optimization | Performance |
|---|---|---|
| Sabine/Eyring RT60 calculation | Accepted (with caveats) | Not accepted |
| Computer acoustic model (CATT, Odeon) | Accepted (with caveats) | Not accepted |
| ISO 3382-2 RT60 measurement | Accepted | Required |
| IEC 60268-16 STI measurement | Accepted | Required |
| IEC 61672-1 noise level measurement | Accepted | Required |
| Material specification sheets | Accepted as supporting evidence | Supporting evidence only |
The caveats for Optimization submissions are significant: the assessor will check that the calculation methodology is correct, that the material data comes from ISO 354 or ASTM C423 test reports (not manufacturer marketing), and that the room geometry matches the architectural drawings. Approximations such as "average absorption = 0.5" or "NRC 0.75 for all surfaces" will be rejected.
Why Calculations and Measurements Diverge
There are systematic reasons why design-stage calculations do not match post-construction measurements:
1. Installation quality. An acoustic ceiling specified as NRC 0.85 achieves its rated performance only when installed with the manufacturer-specified airspace, edge profile, and grid system. Tiles installed tight against the soffit (zero airspace) can lose 0.10-0.20 in absorption coefficient at low frequencies. Tiles with visible gaps at the grid have reduced performance at high frequencies.
2. As-built geometry. The calculation assumes the dimensions on the architectural drawings. In practice, walls move during value engineering, ceiling heights change to accommodate ductwork, and additional glazing is added for daylighting. Each change affects the surface area and the total absorption calculation.
3. Flanking transmission. Sabine and Eyring models assume a closed room with no sound leakage. Real rooms have doors with 5-10 mm undercuts, partitions that stop at the suspended ceiling (not the slab), and HVAC penetrations through walls. These flanking paths reduce the measured isolation and can increase the effective RT60 by 0.05-0.15 seconds compared to the closed-room prediction.
4. Furniture and fit-out. The calculation typically uses generic absorption coefficients for "upholstered chair" or "carpet." The actual furniture specified by the interior designer may have different absorption characteristics — particularly if cost-driven substitutions replace acoustic-grade products with standard commercial alternatives.
The Measurement Protocol
Post-construction measurements for WELL Performance verification must follow:
- RT60: ISO 3382-2:2008, interrupted noise method or integrated impulse response. Minimum 3 source-receiver positions per room. Results reported per octave band (125 Hz to 4000 Hz).
- STI: IEC 60268-16:2020, STIPA method. Source at 1.5 m height (standing talker), receiver at 1.2 m height (seated listener). Minimum 3 source-receiver pairs per open plan zone.
- Background noise: IEC 61672-1, L_Aeq over a representative 15-60 minute period. All building systems at normal operating condition. Minimum 5 measurement positions per zone.
The Practical Lesson
Design-phase calculations are essential for specifying the right treatments. But they are not evidence of compliance for the Performance pathway. The project timeline must include a post-construction measurement phase — typically 2-4 weeks after furniture installation and HVAC commissioning — and budget for remediation if the measurements reveal discrepancies.
Budget $3,000-$8,000 for a full WELL acoustic measurement survey of a typical 1,000-2,000 m squared office floor plate (RT60, STI, and background noise across all applicable spaces). This is less than 0.1% of the total fit-out cost and is the only way to demonstrate compliance for Performance certification.
The Common Thread: Specificity
All five errors share a root cause: treating WELL acoustic certification as a single-number check when it is actually a multi-parameter, condition-specific, measurement-verified process. RT60 alone is necessary but not sufficient. The correct NRC is necessary but not sufficient. The right threshold is necessary but not sufficient. STI compliance is necessary but not sufficient. And calculations are necessary but — for Performance certification — they are not evidence.
The projects that pass on the first submission are the ones that address all five parameters together, from the design phase through post-construction verification.
Calculate Your WELL F74 Compliance Before You Submit
AcousPlan calculates RT60 per octave band (not just the broadband average), STI between any source-receiver pair in your room, and flags which WELL Part your design passes or fails — all before you submit to the assessor.
Check your WELL F74 compliance with AcousPlan — enter room dimensions, select materials from 5,600+ products with full octave-band absorption data, and generate a compliance report that covers Parts 1, 2, and 3.
Related Articles
- WELL v2 Feature 74 Decoded: Every Acoustic Requirement, Every Calculation, Every Clause — the complete reference for all three Parts of Feature 74
- Open Plan Office Acoustic Design: The Complete Guide — ISO 3382-3 parameters, ABC rule, cost analysis
- NRC 0.75 Does Not Mean 75% Absorption — why single-number ratings hide octave-band failures