Skip to content
case study · workforce wellness

Measuring wellness the survey never sees

A rigorous 10-day internal validation study — seven participants, three daily voice assessments, a controlled stress phase and a recovery phase — testing whether conversational voice check-ins can measure workforce wellness in real time.

Design
Three-phase, 10-day controlled study
Participants
7 · 30 interactions each
Assessment
3–5 min conversational call
Features analyzed
30+ acoustic features
89%Employee adherencevs. 40–60% for surveys
97%System accuracyAcross 1,076 tasks
75%Time saved per assessment3–5 min vs. 15–20
0.84Correlation (r)AI vs. self-report
the challenge

Investment without measurement

Despite significant investment in wellness programs, most organizations lack real-time, objective insight into employee mental health and cognitive performance.

Traditional wellness surveys suffer from low completion rates of 40 to 60%, delayed insights and subjective bias — making it impossible to identify struggling employees before crises develop.

Seven Amplifier team members took part in a 10-day study with three daily voice assessments. Each 3–5 minute call included natural conversation and brief cognitive tasks while the model analyzed 30+ voice features: speech rate, pause patterns, pitch variability, articulation precision and more.

study design

Baseline. Stress. Recovery.

The three-phase design tested whether voice biomarkers could detect meaningful wellness change during both decline and recovery — the real-world fluctuation quarterly surveys completely miss.

01

Baseline · days 1–3

Normal routines established individual wellness baselines across seven dimensions.

02

Controlled stress · days 4–7

Sleep deprivation, physical exertion, cognitive challenges and emotional stimuli. Energy declined 35%, mental fatigue increased 52%, stress rose 48%.

03

Recovery · days 8–10

Calming music, breathing exercises and pleasant imagery enabled an 85–95% return to baseline, demonstrating sensitivity to improvement as well as decline.

finding 01

Conversational format drives exceptional engagement

Despite 30 total interactions per participant, the study achieved 89% adherence — dramatically exceeding typical wellness survey completion.

Traditional survey (low)40%
Traditional survey (high)60%
Voice biomarkers89%

Participants described the assessments as checking in with a colleague rather than taking a test. Natural speaking requires less cognitive effort than reading and typing.

The conversational AI made it feel like a wellness coach checking in, not an assessment being administered. That naturalness is why adherence remained high throughout the full 10 days.
Vijay Ravi, Research Lead, Amplifier Health
finding 02

Voice captures wellness change in real time

During the stress phase, voice analysis captured concurrent declines across multiple dimensions — then tracked recovery as conditions improved.

DimensionBaselineStress peakChangeAgreement
Energy7.1 / 104.6 / 10↓ 35%89%
Mental fatigue3.8 / 105.8 / 10↑ 52%91%
Focus7.5 / 105.4 / 10↓ 28%88%
Stress3.2 / 104.7 / 10↑ 48%82%
Mood7.4 / 104.4 / 10↓ 41%79%
finding 03

Specific voice features predict specific states

Correlation coefficients between acoustic features and wellness dimensions, all at p < 0.001.

Cognitive tasks → focus0.73
Speech rate → energy0.71
Articulation → energy0.69
Voice tremor → stress0.68
Pause patterns → fatigue0.67
Pitch variability → mood0.65

Overall agreement between model predictions and self-reported scores across all seven dimensions was r = 0.84.

Why it matters operationally

focus · r = 0.73

Concentration you can act on

Cognitive task performance predicts concentration ability — directly tied to productivity and decision quality.

energy · r = 0.71

Fatigue before burnout

Speech rate and articulation precision indicate energy levels, making early fatigue detection possible.

stress · r = 0.68

Workload adjustment in time

Voice tremor and pitch variability signal stress, enabling intervention before a crisis develops.

the detection insight

Real-time measurement enables proactive support

Quarterly surveys provide delayed snapshots — an employee might struggle for weeks before the next one lands. Voice biomarkers capture wellness states the same day, so organizations can offer support when it is needed rather than weeks after the fact. When wellness monitoring feels like care rather than surveillance, employees actually show up.

Wellness measurement your workforce will actually complete

If you run an employee wellness, EAP or performance program, we can pilot conversational voice check-ins against your existing instrument and compare adherence and agreement directly.

Internal validation study, n = 7, 10 days. Outputs are wellness signals, not clinical findings or diagnoses.