Background Noise's Per-Head Performance Loss Outweighs Fit-Out

For the CRE and workplace decision-maker

Get the acoustics right before you defend the capex, because staff cost per person per year dwarfs rent and energy per head, and background speech that pushes distraction distance out across an open floor buys you measurable performance loss on every desk. Haapakangas and colleagues (2020, Indoor Air) place the onset of that loss at a speech transmission index of 0.21, with the maximum decrease reached at 0.44, so the threshold is a measured number you can specify against. Kristiane Roed Jensen and colleagues (2019, PLoS ONE), working in a healthcare rather than an office setting, found teams unexposed to noise had a 91.3 per cent chance of high performance against 58.9 per cent for exposed teams; read it as directional. The effect on stress and error rate is backed by solid evidence, with the effect on cognitive performance more moderate. This sheet sets out what to require in an acoustic spec, what the STI and ISO 3382-3 target values mean for a fit-out or a lease-versus-relocate paper and where the case is strong enough to enter a GRESB or CSRD social pillar.

Version V2 · evidence checked July 2026 · open-plan offices

The three steps to the staff-cost case

Three steps lead from acoustics to revenue. They differ in how well they are proven, and brokers run them together.

Step one, the environment to people. Solid.

Background speech reaches people at their desks and degrades the work they are doing with words. Banbury and colleagues (1998, British Journal of Psychology) found that both speech and office noise disrupt performance on memory for prose and mental arithmetic, and that the effect holds independently of what the irrelevant speech means. Sörqvist and colleagues (2012, Scandinavian Journal of Psychology) recorded the same pattern in written output: normal speech impaired the number of characters produced and raised uncorrected typing errors while lengthening the pauses between words. Keus van de Poll and colleagues (2015, The Journal of the Acoustical Society of America) traced where the harm concentrates, showing that a single intelligible voice impaired memory relative to silence whereas masking that voice with multiple voices brought performance back to the level of silence.

The physiological side is measured with comparable confidence. Lee and colleagues (2024, Journal of Toxicology and Environmental Health, Part A) found that short-term noise during task performance elevates stress hormone levels, with the extent of the rise varying by exposure across 35, 45 and 75 dB. Haapakangas and colleagues (2020, Indoor Air) located a threshold rather than a single trigger point: performance begins to fall at roughly STI = 0.21 and reaches its maximum decline at STI = 0.44.

This step is backed by solid evidence. The direction is consistent across verbal-task accuracy, output quality and stress markers, and it survives the counter-findings. Hua and colleagues (2014, Journal of the American Academy of Audiology) observed no significant differences in cognitive performance across four background conditions, and Zhou and colleagues (2023, Cognition, Technology & Work) reported minimal evidence for a negative effect of moderate broadband noise. Both point to the same distinction the supporting work draws. The reliable harm comes from intelligible speech rather than steady broadband sound, which is why the derived thresholds turn on speech transmission and distraction distance.

Step two, quality to price. Moderate, and only on average.

• Green-label rent and price premium (bundled): “‘rated’ buildings command substantially higher rents and selling prices than otherwise identical buildings”. Measures prices not human effect; a bundled green label (Energy Star or LEED), not an acoustic measure (Eichholtz, Kok & Quigley 2010, American Economic Review 100(5):2492-2509, DOI 10.1257/aer.100.5.2492 (academic, peer-reviewed)). Market price evidence, a category average, not the single measure.
• Green-certification rent and sale premium (bundled, working paper): “There is a rental premium of approximately 6% for LEED and Energy Star certification. A sale price premium of approximately 35% was found for … LEED rated buildings and 31% for … Energy Star rated buildings”. Measures prices not human effect; bundled certification not acoustics; US sample, academic working paper not peer-reviewed (Fuerst & McAllister 2009, Working Papers in Real Estate & Planning 07/09, University of Reading (academic working paper, not peer-reviewed)). Market price evidence, a category average, not the single measure.

Step three, the single measure to one building’s revenue. Thin. No supplied market item isolates what a single acoustics measure adds to one building’s revenue; that step stays the weakest link.

What to require

1. Set the design target for speech intelligibility at the occupant’s position at STI 0.21 or below. Haapakangas and colleagues (2020, Indoor Air) found that performance begins to decrease approximately above STI = 0.21 while the maximum decrease is reached at STI = 0.44, so treat 0.21 as the onset threshold and hold the field between it and 0.44 rather than accepting a single nominal value.

2. Require a measured spatial decay rate of speech D2,S of at least 7 dB per distance doubling (ISO 3382-3:2012, Annex A). This is an informative example target, so record the measured decay rather than a designer’s predicted curve.

3. Require the A-weighted sound pressure level of speech at 4 m, Lp,A,S,4m, to sit at or below 48 dB (ISO 3382-3:2012, Annex A). Verify it against a measurement at the receiving position, not against a modelled estimate.

4. Require the distraction distance rD to stay at or below 5 m (ISO 3382-3:2012, Annex A). Beyond that distance speech should have decayed below the distraction threshold, and the point where it does should be measured on the built floor.

5. Hold the background level from building services to the ASHRAE ranges (2023, Ch. 49, Table 1): NC/RC 30, roughly 35 dBA, for an enclosed space, and NC/RC 40, roughly 45 dBA, for an open shared space. These are steady-state service levels and belong on a sound level meter reading, not on a specification sheet alone.

Target values

Each value gets two separate labels: what kind of number it is, a measured threshold or an adopted standard convention, and how strong the evidence behind it is. Most published targets record what a standards committee agreed; only a few trace to a measured point where behaviour changes. The measured human threshold here, STI, shows a dose-response gradient: in open-plan offices, an increase in distraction distance (Speech Transmission Index-based) predicts an increase in disturbance; cognitive performance begins to decrease above STI = 0.21 and reaches maximum decrease at STI = 0.44. That is what earns it a solid rating under GRADE, with the open question of how far it holds on a real floor held against it.¹

Parameter	Target	Unit	Condition	Kind of number	Strength
STI	threshold at 0.21 (max effect at 0.44)	—	observed boundary; see What to require	measured threshold	solid
D2,S	≥ 7 dB	—	international (ISO 3382-3:2012)	standard convention	adopted
Lp,A,S,4m	≤ 48 dB	—	international (ISO 3382-3:2012)	standard convention	adopted
rD	≤ 5 m	—	international (ISO 3382-3:2012)	standard convention	adopted
Private office NC/RC	30 (≈ 35 dBA)	—	international (ASHRAE)	standard convention	adopted
open-plan office NC/RC	40 (≈ 45 dBA)	—	international (ASHRAE)	standard convention	adopted

Each row names its source: a measured threshold points to the cited study, an adopted convention to a standards reference. Each standard’s full class bands sit in the cited reference.

What the evidence supports, and what it does not

Reducing background speech raises satisfaction and lowers disturbance in shared open spaces. Haapakangas and colleagues (2020, Indoor Air) found that irrelevant background speech causes dissatisfaction and impairs cognitive performance, and Sakellaris and colleagues (2016, International Journal of Environmental Research and Public Health) found the highest association with overall comfort for noise, ahead of air quality, light and thermal satisfaction. The evidence establishes the direction of the effect more firmly than any single decibel figure.

Speech intelligibility drives the effect more than loudness alone. Banbury and colleagues (1998, British Journal of Psychology) found that office noise with speech disrupts both a memory task and a mental arithmetic task while noise without speech disrupts arithmetic only, and Keus van de Poll and colleagues (2015, The Journal of the Acoustical Society of America) found that a single voice impaired memory against silence while masking that voice with multiple voices returned performance to the level of silence. A claim that a fixed dBA figure guarantees a quiet floor sits against this: the intelligibility of what is heard matters alongside its level.

Raised background noise probably increases physiological stress. Lee and colleagues (2024, Journal of Toxicology and Environmental Health, Part A) found significant exposure-dependent differences in stress hormone levels between 35, 45 and 75 dB, and Seddigh and colleagues (2015, Journal of Environmental Psychology) found enhanced acoustical conditions associated with lower perceived disturbance and cognitive stress.

The effect of moderate broadband noise is weaker than the speech case. Discount any promise built on masking alone. Zhou and colleagues (2023, Cognition, Technology & Work) found minimal evidence for a negative effect of moderate broadband noise on cognitive performance, and Hua and colleagues (2014, Journal of the American Academy of Audiology) observed no significant differences in cognitive performance across four background conditions. Reinten and colleagues (2017, Building and Environment) noted that evidence for the effectiveness of room acoustic strategies to control the sound environment is lacking, which is a caution against selling a specific treatment as a solved problem.

Cost order, per fit-out measure

From the supplied market evidence, marked as price evidence:

• Sound masking cost per area (vendor): “an AtlasIED sound masking system costs around $1 - $2 per square foot” (broader vendor range about $1.50 to $6.00 per square foot; retrofit higher than new build). A cost input, not a revenue or human-effect figure (AtlasIED, manufacturer and vendor with a sales interest)
• Acoustic ceiling cost per area (vendor): “Typical installed range: $6 to $20 per square foot; average installed cost: $10 to $14 per square foot”. A cost input, not a revenue or human-effect figure (Acoustical Surfaces, Inc., vendor and installer with a sales interest)

Measurement and verification

Verify an existing floor with the ISO 3382-3:2012 protocol, which yields the four quantities directly from measurement at occupied positions. Measure the spatial decay rate of speech D2,S, the A-weighted speech level at 4 m Lp,A,S,4m and the distraction distance rD along a line of positions where people spend time, and read them against the Annex A example targets of D2,S at least 7 dB, Lp,A,S,4m at or below 48 dB and rD at or below 5 m. Take the readings on the furnished floor in its operating state so the result reflects the space as used.

Measure the steady background level from services separately with a sound level meter at the occupant’s position and compare it to the ASHRAE ranges (2023, Ch. 49, Table 1) of NC/RC 30, roughly 35 dBA, for an enclosed space and NC/RC 40, roughly 45 dBA, for an open shared space. This level is the floor beneath the speech metrics and should be logged with the ventilation running.

Carry the speech intelligibility check against the measured STI threshold from Haapakangas and colleagues (2020, Indoor Air), with the onset at 0.21 and the maximum effect at 0.44. On an occupied floor, measure STI at representative receiving positions and record where it falls within that range rather than reporting one averaged value.

In a new design the same quantities can be modelled before construction, and the model gives a prediction that stands until the floor is built. Repeat the full ISO 3382-3 measurement after fit-out and handover, since a designer’s predicted decay curve and the level measured on the finished floor can diverge once furniture, occupancy and services are in place.

ESG reporting, social pillar

Background noise in open-plan offices enters the social pillar through three human effects with named evidence: cognitive performance, error rate and stress. Haapakangas and colleagues (2020, Indoor Air) measured a threshold where performance begins to fall around STI = 0.21 and reaches its maximum decrease at STI = 0.44. This gives the social pillar a measured acoustic threshold to report against rather than a subjective complaint count.

• Verbal work degrades under intelligible background speech, which is the dominant noise in open-plan settings. Banbury and colleagues (1998, British Journal of Psychology) found office noise with speech disrupts both memory for prose and mental arithmetic, while noise without speech disrupted arithmetic only. Sörqvist and colleagues (2012, Scandinavian Journal of Psychology) found normal speech increased uncorrected typing errors and lengthened pauses between words compared with quieter conditions.
• Error rate rises with noise level rather than merely with annoyance. Golmohammadi and colleagues (2019, Applied Acoustics) recorded higher average commission errors in a 1-Back task under louder conditions. Jensen and colleagues (2019, PLoS ONE) reported that teams unexposed to noise reached high clinical performance in 91.3 per cent of cases against 58.9 per cent for exposed teams (p < 0.001), a healthcare setting rather than an office, so read it as directional for office knowledge work.
• Stress shows an exposure-dependent physiological signal. Lee and colleagues (2024, Journal of Toxicology and Environmental Health, Part A) found short-term noise during task performance elevated stress hormone levels with significant differences between 35, 45 and 75 dB. Colenberg and colleagues (2020, Building Research & Information) reported that open-plan offices, shared rooms and higher background noise are negatively related to health.

One supplied item touches absence directly. Palacios and colleagues (2020, PLoS ONE) found a 2 per cent reduction in absenteeism after relocation to a building with improved perceived environmental conditions and lower sick building syndrome symptoms. That figure belongs to a whole-environment relocation covering air quality, thermal comfort, lighting and noise together. Gap: the supplied evidence does not isolate the share of that 2 per cent attributable to background noise alone.

No defensible single productivity or wellbeing percentage can be placed on background noise as a standalone measure. The strongest supplied effects (Schiller and colleagues, 2021, JSLHR, Cohen’s d between -0.67 and -2.65) come from children’s listening tasks across signal-to-noise ratios, not adult office output, and the direction is contested: Zhou and colleagues (2023, Cognition, Technology & Work) found minimal evidence for a negative effect of moderate broadband noise, and Hua and colleagues (2014, Journal of the American Academy of Audiology) observed no significant cognitive differences across four background conditions. For the social pillar, report the acoustic targets met rather than a converted productivity gain: STI at or below 0.21, D2,S at least 7 dB, Lp,A,S,4m at or below 48 dB, rD at or below 5 m and an open-plan NC/RC of 40 against a private-office 30, all as adopted standard conventions from ISO 3382-3:2012 Annex A and the ASHRAE Handbook HVAC Applications (A23, 2023). These are the figures that can be audited without inventing a per-head number the evidence will not support.

Sources

Human effects, the evidence core.

• Background Noise lowers Satisfaction (thin, measured). Sandra A. Sherman and colleagues, 2005, Children Youth and Environments; Piers MacNaughton and colleagues, 2016, Building and Environment; Haapakangas A and colleagues, 2017, The Journal of the Acoustical Society of America; Ellen Taylor and colleagues, 2018, HERD; Haapakangas A and colleagues, 2020, Indoor Air; Andrea Bertuzzi and colleagues, 2023, BMJ Open.
• Background Noise lowers Comfort (moderate, measured). Sandra A. Sherman and colleagues, 2005, Children Youth and Environments; Haapakangas A and colleagues, 2017, The Journal of the Acoustical Society of America; S. Torresin and colleagues, 2018, Indoor Air; Ellen Taylor and colleagues, 2018, HERD; H. Gaminiesfahani and colleagues, 2020, HERD: Health Environments Research & Design Journal; Andrea Bertuzzi and colleagues, 2023, BMJ Open; Matteo Pellegatti and colleagues, 2023, Building and Environment; Simone Torresin and colleagues, 2025, Building and Environment.
• Background Noise Impairs Cognitive Performance (moderate, measured). Simon Banbury and colleagues, 1998, British Journal of Psychology; Aram Seddigh and colleagues, 2015, Journal of Environmental Psychology; Haapakangas A and colleagues, 2020, Indoor Air; Isabel S. Schiller and colleagues, 2021, Journal of Speech, Language, and Hearing Research : JSLHR; Nazi Niazmand-Aghdam and colleagues, 2021, La Medicina del Lavoro; Michail Evangelos Terzakis and colleagues, 2022, International Journal of Environmental Research and Public Health; Matteo Pellegatti and colleagues, 2023, Building and Environment; Arianna Latini and colleagues, 2024, SSRN Electronic Journal. Against: Hang Zhou and colleagues, 2023, Cognition, Technology & Work.
• Background Noise raises Stress (solid, measured). Roger Ulrich, 2008, The Journal of the Acoustical Society of America; Aram Seddigh and colleagues, 2015, Journal of Environmental Psychology; H. Gaminiesfahani and colleagues, 2020, HERD: Health Environments Research & Design Journal; Susanne Colenberg and colleagues, 2020, Building Research & Information; Juan Palacios and colleagues, 2020, PLoS ONE; Daan Kropman and colleagues, 2022, Ergonomics; Michail Evangelos Terzakis and colleagues, 2022, International Journal of Environmental Research and Public Health; Sung-Hee Lee and colleagues, 2024, Journal of Toxicology and Environmental Health, Part A.
• Background Noise lowers Productivity (moderate, measured). Tina Bringslimark and colleagues, 2007, HortScience; A. Kaarlela-Tuomaala and colleagues, 2009, Ergonomics; Aram Seddigh and colleagues, 2015, Journal of Environmental Psychology; Ioannis Sakellaris and colleagues, 2016, International Journal of Environmental Research and Public Health; Jikke Reinten and colleagues, 2017, Building and Environment; Kristiane Roed Jensen and colleagues, 2019, PLoS ONE.
• Background Noise raises Error Rate (solid, measured). Laura J. Loewen and colleagues, 1992, Environment and Behavior; Simon Banbury and colleagues, 1998, British Journal of Psychology; S. J. Schlittmeier and colleagues, 2008, Ergonomics; Patrik Sörqvist and colleagues, 2012, Scandinavian Journal of Psychology; Marijke Keus van de Poll and colleagues, 2015, The Journal of the Acoustical Society of America; Marijke Keus van de Poll and colleagues, 2016, Applied Cognitive Psychology; Lilin Xiong and colleagues, 2018, International Journal of Environmental Research and Public Health; Rostam Golmohammadi and colleagues, 2019, Applied Acoustics; Manuj Yadav and colleagues, 2023, Applied Acoustics. Against: Håkan Hua and colleagues, 2014, Journal of the American Academy of Audiology.

Related topics and trade-offs

Speech intelligibility drives the distraction, so the room-acoustic levers work against each other and against the fit-out budget. Sound masking lowers the signal-to-noise ratio of a distant voice by raising the ambient level, which trades a lower absolute noise figure for control of intelligible speech. Haapakangas and colleagues (2020, Indoor Air) locate the onset of measurable performance loss at STI = 0.21 with the maximum reached at STI = 0.44, so the target is a speech transmission index kept low across the floor rather than a quiet room in the abstract sense. A floor tuned only for a low sound-pressure reading can still carry speech clearly across desks and cost attention.

• Desk density works directly against the distraction distance. The ISO 3382-3:2012 example targets set the distraction distance rD at 5 m or below and the spatial decay rate D2,S at 7 dB or above, both of which a tighter desk grid erodes as more speaking colleagues fall inside the range where speech stays intelligible. Haapakangas and colleagues (2017, The Journal of the Acoustical Society of America) report that an increase in distraction distance predicts an increase in disturbance by noise. Adding workstations to the same footprint raises the rent recovered per square metre while shifting the acoustic case in the wrong direction, and the staff-cost side of the ledger is the larger number per head.

• Lower noise targets pull against HVAC sizing and running cost. The ASHRAE Handbook (A23, 2023) sets the open-plan design point at NC/RC 40, roughly 45 dBA, and the private office at NC/RC 30, roughly 35 dBA, and reaching the lower figure means larger ducts, lower air velocities and quieter terminal units. Pellegatti and colleagues (2023, Building and Environment) found a negative effect of fan noise on comfort, which argues for the quieter plant. Gap: the supplied evidence gives no figure for the energy or capital cost of moving from the open-plan to the private-office noise target, so the trade against operating cost cannot be sized here.

• Sound masking itself sits on a threshold. Raising the masking level buys speech privacy by pushing intelligible speech below the STI = 0.21 onset, yet an ambient level set too high becomes its own source of dissatisfaction. Haapakangas and colleagues (2020, Indoor Air) describe irrelevant background speech as the mechanism that causes dissatisfaction and impairs cognitive performance, which is the case for masking; a masking level set too high is itself an ambient sound the occupant hears. The workable band is narrow, and the STI 0.21 onset marks the point the masking has to push intelligible speech below.

Related sheets drawn from the supplied factor conflicts:

• Background Noise and Satisfaction (Factor 674), where the open-plan setting and the same distraction-distance mechanism make the occupant experience case rather than the cognitive-performance case.
• Background Noise and Comfort (Factor 675), which shares the fan-noise conflict from Pellegatti and colleagues (2023) and the occupant-control argument.
• Background Noise and Stress (Factor 679), where Palacios and colleagues (2020, PLoS ONE) report a 2 per cent reduction in absenteeism after relocation, the figure closest to the staff-cost line on this topic.
• Background Noise and Productivity (Factor 682), where Jensen and colleagues (2019, PLoS ONE) record high clinical performance in 91.3 per cent of unexposed teams against 58.9 per cent of exposed teams, the sharpest performance gap in the set.

The same factor material underpins the developer and investor sheet on this topic, framed to the capital case rather than the occupier’s staff-cost case. There the acoustic targets set fit-out specification and HVAC sizing against build cost and rentable area. This sheet keeps them on cognitive performance, error rate and staff cost per person per year.

Change log

V1, July 2026. First release, built from the factor evidence (674, 675, 676, 679, 682, 1191). New, changed, raised or unchanged lines carry a version marker from the next version on.

V2, July 2026. Factor 674 downgraded from moderate to thin, Factor 676 relabelled and downgraded from solid to moderate (Background Noise Lowers Sustained Attention is now Background Noise Impairs Cognitive Performance) and Factor 682 downgraded from solid to moderate, after a citation review.

Footnotes

1. The strength labels follow GRADE: solid covers High and Moderate certainty, moderate covers Low, thin covers Very Low (Cochrane Handbook, chapter 14, 2023). GRADE starts observational evidence low but allows an upgrade for a large effect, a dose-response gradient or confounders that would work against the effect found, so a strong field result is not forced down to thin. ↩