Endel Focus Study

Source: Aia Haruvi et al., “Measuring and Modeling the Effect of Audio on Human Focus in Everyday Environments Using Brain-Computer Interface Technology,” Frontiers in Computational Neuroscience 15 (2022). Full article

Claim

One company-funded study found a higher proprietary EEG-derived focus score for Endel than for silence in its aggregate comparison. It did not measure productivity, found no advantage over Apple’s focus music in the reported pairwise time-series analysis and could not separate personalisation from differences in the sound material.

Evidence and method

The study enrolled 62 people and analysed 51 after exclusions. Each participant completed four one-hour sessions at home while wearing a Muse-S EEG headband, with Endel Focus, Apple Music’s Pure Focus playlist, Spotify’s Focus Flow playlist or silence. During the main 30-minute period, participants chose a task such as work, reading, knitting or Sudoku (pp. 1, 3–4).

The primary outcome measured neither task completion nor accuracy. Arctop’s proprietary Neuos system used a random-forest model to map EEG features to participants’ self-reported focus during separate calibration tasks. The authors report a correlation of 0.6 across tasks, a mean within-person AUC near 0.83 and threshold classification accuracy near 0.8 (pp. 5–8).

The repeated-measures aggregate test differed across the four conditions, F(3,150) = 4.28, p = .006. Endel scored higher than silence, with a mean difference of .090 and p = .008. Apple versus silence, p = .11, and Spotify versus silence, p = .65, were not significant in those aggregate comparisons. Endel was the highest-scoring condition for 35.3 percent of participants, compared with Apple for 27.5 percent, Spotify for 19.6 percent and silence for 17.6 percent (p. 8).

Separate subgroup analyses found Endel above silence among people doing work tasks but not non-work tasks, and all audio conditions above silence among participants under 36 but not older participants. The article does not report direct task-by-condition or age-by-condition interaction tests, so separate significance levels do not prove that the effects differ between those groups (p. 8).

In the time-series pairwise table, Endel differed from silence during 87 percent of the analysed windows, Apple during 60 percent and Spotify during 27 percent. Endel differed from Spotify during 37 percent of windows and from Apple during none (p. 10).

Concepts

• Operationalisation and construct validity: “Focus” becomes the output of a trained EEG model that may relate to self-report without establishing task quality or practical benefit.
• Comparator: Endel was compared with two different playlists and silence, not with the same soundscape played without adaptation.
• Personalisation effect: To test adaptation itself, a study must hold musical content constant and change only personalisation.
• Interaction test: A claim that an effect differs by age or task needs a direct comparison between subgroup effects.
• Conflict of interest: Funding and involvement can shape design and interpretation even when the statistical report remains usable.

Limits

All authors worked for Arctop, while Arctop and Endel funded the research. Arctop took part in study design, data collection, analysis, interpretation, writing and the decision to submit. Endel helped design the study and supplied the audio condition. The study lacked a non-personalised version of the Endel soundscape, did not randomise the order of participants’ chosen tasks and had too little power for some demographic analyses (pp. 13, 15, 17).

The study did not measure task completion, accuracy, productivity, sleep or clinical outcomes, and its results do not establish a causal effect of biofeedback. They also cannot distinguish Endel’s sound design from real-time personalisation. Independent replication would carry more weight than another vendor-linked study.

Presentation use

Use one evidence slide with exact wording: “A company-funded at-home EEG study found a higher model-derived focus score for Endel than silence. It did not measure productivity, did not beat Apple in the reported pairwise time-series table and could not isolate personalisation.” This states the finding without repeating Endel’s marketing language.

The study also sharpens the musicianship argument. When a platform judges music through a focus metric, composition may become optimisation against a measure chosen by researchers and the company. Ask how that criterion changes sound design, whose judgement counts and whether the metric becomes part of the musical instrument.