Effect of a brief, survey-delivered commitment to reduce pre-sleep smartphone use by ≥30 minutes on next-morning subjective rest among European adults

1. Executive summary

This report documents a pre-registered, randomized two-arm survey experiment conducted on StyleCodes and partner panels between 2025-11-01 and 2025-11-15 to estimate the causal effect of a brief, optional commitment prompt to reduce smartphone use in the hour before bed by at least 30 minutes on next-morning subjective restfulness among European adults aged 18-50. The study randomized consenting, eligible participants 1:1 to a Treatment cohort that received a single commitment prompt (n = 754) or to a Control cohort that received a neutral prompt (n = 754), yielding an overall effective sample of 1,508 respondents for the baseline and immediate post-randomization measures reported here.

Primary estimand (pre-registered): the intention-to-treat (ITT) difference in mean next-morning subjective restfulness (FOLLOWUP_RESTFULNESS, 1-5) between assignment to the commitment prompt and assignment to the neutral control. Secondary estimands include the complier average causal effect (CACE) using self-reported compliance (defined as a numeric reduction in pre-sleep minutes of ≥30), heterogeneity by baseline pre-sleep phone use, age group, and baseline sleep quality, and mediation via intention/confidence and perceived returns.

Immediate, post-randomization indicators demonstrate substantive uptake of the prompt and feasibility of the design. In the Treatment cohort 55% agreed to the commitment, and mean self-reported confidence and intention in the Treatment cohort were 3.60 (SD = 1.10) and 3.50 (SD = 1.20) on 1-5 scales respectively, indicating moderate-to-strong proximal intent among those who received the prompt. Baseline distributions indicate substantial pre-sleep smartphone use in the recruited sample: overall 35% report 31-60 minutes and 18% report more than 60 minutes of smartphone use in the 60 minutes before trying to fall asleep. Attention-check pass rates were high overall (Agree + Strongly agree ≈ 87%), and opt-in to the next-morning follow-up was substantial (overall 59%), positioning the study to estimate the pre-registered ITT and CACE effects once follow-up observations are collected and QC exclusions applied.

This report provides full documentation of the fieldwork, instruments, quality-control rules, immediate descriptive results from baseline and randomized-process measures, and a detailed analysis plan for the primary outcome (FOLLOWUP_RESTFULNESS). Follow-up data collection and pre-specified ITT/CACE analyses will be executed once the 24-48 hour follow-up window concludes and data are processed according to the pre-registered QC and inclusion rules described below.

2. Background and motivation

2.1. Problem statement and relevance

Excessive smartphone screen use immediately before bedtime is common and plausibly interferes with sleep onset and subjective rest. A brief, survey-delivered commitment to reduce pre-sleep smartphone use by a concrete amount (≥30 minutes) is a low-cost, easily scalable behavioral nudge that consumer platforms such as StyleCodes can deliver to readers. The present study asks whether such a single-night commitment, deployed in a consumer-site context and measured using a short next-morning follow-up, produces detectable improvements in subjective rest. The intervention is directly relevant to StyleCodes' audience - adults 18-50 in major European markets interested in wellbeing and daily routines - and could inform product features, messaging, and low-friction interventions on platform.

2.2. Gap and contribution

The study tests causal effects of a one-night, in-survey commitment in a mixed recruitment environment (site users and matched online panels). It addresses the gap between correlational findings linking reduced pre-sleep screen time to better next-morning rest and randomized tests of single-night behavioral pledges within consumer platforms. The immediate baseline distributions in this sample (35% report 31-60 minutes and 18% report >60 minutes of pre-sleep phone use) show the intervention is relevant to a large share of the recruited population.

3. Research questions and estimands

3.1. Primary research question and estimand

Primary question: Does assignment to the commitment prompt (Treatment) versus assignment to a neutral prompt (Control) increase next-morning subjective restfulness?

Primary estimand: ITT difference in mean FOLLOWUP_RESTFULNESS (1-5 scale) between Treatment and Control assignments. Analytical treatment: ordinary least squares with FOLLOWUP_RESTFULNESS treated as continuous (1-5), reporting mean differences and 95% confidence intervals; ordered-logit reported as sensitivity.

Alpha: two-sided 0.05 for the primary ITT test; secondary and subgroup p-values adjusted for multiple testing via Benjamini-Hochberg false discovery rate at 5% and reported unadjusted alongside adjusted.

Variance estimation: robust (HC3) standard errors clustered at the country level for the primary ITT and subgroup analyses; cluster-robust SEs and non-clustered SEs reported in appendix.

3.2. Secondary and exploratory estimands

• CACE (complier average causal effect) using assignment as an instrument for self-reported compliance, where compliance is defined as a numeric reduction in pre-sleep minutes of ≥30 between baseline and follow-up (partial reductions of 15-29 min are recorded but not counted as compliers in the primary CACE definition).
• Heterogeneity of ITT and CACE by baseline pre-sleep phone-use category (BASE_MINUTES_CAT: 0; 1-15; 16-30; 31-60; >60), age groups (18-29, 30-39, 40-50), and baseline sleep quality (BASE_SLEEP_QUALITY).
• Mediation: the extent to which effects operate via immediate intention/confidence (TREAT_INTENTION/TREAT_CONFIDENCE) and perceived returns (PERCEIVED_RETURN_1-3).

Pre-specified covariate adjustment includes baseline minutes (numeric) or categorical baseline minutes, baseline sleep quality, age group, gender, and country strata to improve precision.

4. Fieldwork, recruitment, and sample

4.1. Fieldwork window and recruitment channels

Fieldwork occurred 2025-11-01 to 2025-11-15 on StyleCodes and via partner panels to broaden representativeness. Recruitment used on-site banners, email invitations to registered users, and targeted panel invites. Incentives were brief site-centric rewards (site points or content unlocks), and participants could opt into a one-time follow-up with a small site-point incentive.

4.2. Eligibility, screening, and quotas

Inclusion: age 18-50, smartphone owner/regular user, reports smartphone use in the 1-2 hours before bed. Exclusion: diagnosed severe sleep disorders under active clinical treatment and regular night-shift workers. Recruitment stratified by country × age group (18-29, 30-39, 40-50) × gender to match site audience; post-stratification weights were planned to align sample margins to target population margins.

4.3. Sample targets and realized sample

Pre-registered targets: required N = 600 per arm (1,200 overall); suggested effective N target = 720 per arm (1,440 overall) allowing QC-exclusion buffer. Realized effective sample: 754 respondents per randomized arm, for an overall effective N = 1,508. This exceeds the required sample and the suggested effective target, increasing power for primary and subgroup comparisons.

5. Questionnaire and measures

The instruments used are provided in full in the appendices (Instruments A and B). Key measures and their roles are summarized below.

• Screening and consent: S1-S4 and CONSENT.
• Core profile: age numeric, gender, country, employment role, work schedule, phone operating system.
• Baseline: BASE_MINUTES_CAT and BASE_MINUTES_NUMERIC (minutes used in the 60 minutes before trying to fall asleep), BASE_SLEEP_QUALITY (1-5), and brief difficulty checklist (DIFFICULTY_ONSET, DIFFICULTY_MAINTAIN, DAY_SLEEPINESS).
• Randomized items:
  • Treatment: T1 (commitment: "Tonight, I will reduce my smartphone use in the hour before I go to bed by at least 30 minutes"), T2 (TREAT_CONFIDENCE), T3 (TREAT_INTENTION).
  • Control: C1 (CONTROL_INTENTION) and C2 (CONTROL_CONFIDENCE), phrased without a commitment.
• Mechanisms: PERCEIVED_EFFORT_1-3, PERCEIVED_RETURN_1-3.
• Ancillary: WTP_EUR (one-night blocking feature), CONSTRAINTS_MULTI (multi-select), ATTN_CHECK (attention trap), OPT_IN_FOLLOWUP.
• Follow-up (24-48 hours after intended bedtime): FOLLOWUP_RESTFULNESS (primary outcome, 1-5), FOLLOWUP_MINUTES_NUMERIC, FOLLOWUP_COMPLIANCE (Yes/No/Partially), self-reported sleep duration and onset latency.

Quality-control and privacy procedures included attention checks, minimum completion time thresholds, deduplication by IP and fingerprinting, and storage of email separately and encrypted; email was deleted after follow-up collection.

6. Randomization and intervention details

Randomization was implemented in-survey after baseline measures with a 1:1 allocation to Treatment and Control. Treatment text and commitment wording were exactly as pre-registered (T1). The commitment is optional and administered the night of intended bedtime; fidelity measures include immediate TREAT_CONFIDENCE and TREAT_INTENTION recorded after presentation of the prompt. The Control cohort received neutral wording and the same immediate intention/confidence items, phrased hypothetically to permit comparable measurement without eliciting an explicit commitment.

Randomization seed and reproducibility: the randomization seed was stored (seed = 20251101) and included in archived analysis scripts to ensure reproducibility of assignment.

7. Data quality, fraud controls, and exclusions

Quality-control procedures and pre-registered exclusion rules were applied as follows.

• Attention check: ATTN_CHECK was embedded; the pre-registered rule defines pass as selecting "Agree" or "Strongly agree." Respondents failing an attention check were flagged.
• Minimum completion time: respondents completing below 50% of the median completion time were flagged as speeders; a default threshold of 6 minutes was used if pilot median was unavailable.
• Deduplication: duplicate fingerprints or identical IP + fingerprint hashes were flagged and duplicates excluded.
• Exclusion rule: respondents meeting any two of (attention-check failure, speeders, duplicate fingerprint/IP) were excluded from the primary analysis. Single-flag cases were retained in the primary sample but examined in sensitivity analyses.
• Post-stratification weights: weights were computed to match country × age group × gender margins; weights were normalized to the sample size.
• Follow-up inclusion rule: primary analysis includes follow-up completions made within 24 hours after reported bedtime, with acceptances up to 48 hours; responses outside 48 hours were excluded from the primary ITT but included in sensitivity analyses.
• Follow-up reminders and incentives: two automated reminders were sent (12 hours and 36 hours after the initial follow-up invitation). The first reminder contained a 1.00 EUR-equivalent site-point incentive credited upon completed follow-up.

Observed attention-check results indicate high overall attentiveness: in the Treatment cohort 45% selected "Agree" and 45% selected "Strongly agree" (90% combined); Control cohort reported 47% "Agree" and 40% "Strongly agree" (87% combined); overall combined "Agree" + "Strongly agree" ≈ 87%.

A log of QC exclusions will be provided in the methods appendix; sensitivity analyses will present results including and excluding flagged respondents.

8. Descriptive results: baseline characteristics and balance

This section reports baseline characteristics and tests of balance across randomization arms for key variables. All percentages are reported to one decimal place; means and SDs are reported to two decimals.

8.1. Baseline pre-sleep smartphone use

Overall baseline distribution of reported minutes of smartphone use in the 60 minutes before trying to fall asleep was: 0 minutes 6%; 1-15 minutes 15%; 16-30 minutes 26%; 31-60 minutes 35%; more than 60 minutes 18%.

Balance across arms for BASE_MINUTES_CAT is near-identical. Treatment cohort shares were 0 minutes 5%; 1-15 minutes 15%; 16-30 minutes 25%; 31-60 minutes 35%; more than 60 minutes 20%. Control cohort shares were 0 minutes 6%; 1-15 minutes 14%; 16-30 minutes 26%; 31-60 minutes 34%; more than 60 minutes 20%.

These near-identical distributions support the internal validity of the randomized comparison for outcomes associated with baseline phone use.

8.2. Baseline sleep quality and sleep difficulty

Overall distribution of baseline sleep quality (BASE_SLEEP_QUALITY) was: Very poor (1) 9%; Poor (2) 21%; Fair (3) 33%; Good (4) 27%; Very good (5) 10%.

Baseline sleep-onset difficulty (DIFFICULTY_ONSET) overall: Not during the past month 29%; Less than once a week 35%; Once or twice a week 26%; Three or more times a week 10%. Treatment and Control cohorts showed closely similar distributions (Treatment: 30% / 35% / 25% / 10%; Control: 28% / 36% / 26% / 10%).

Balance across arms on these baseline metrics is sufficient for the randomized comparison and for covariate adjustment to gain precision in primary ITT estimation.

9. Immediate post-randomization responses (mechanisms and uptake)

This section documents uptake, proximal intentions, confidence, perceived effort, perceived returns, constraints, WTP, and opt-in to the follow-up.

9.1. Commitment uptake in Treatment

Among respondents randomized to Treatment (n = 754), 55% explicitly agreed to the commitment to reduce smartphone use by ≥30 minutes tonight; 45% did not commit.

This 55% uptake is a central immediate-treatment indicator used in CACE analyses where self-reported compliance at follow-up is used to estimate treatment-on-the-treated effects.

9.2. Confidence and intention (Treatment vs Control)

Immediate confidence and intention measures taken after randomization indicate moderate levels of expected adherence.

• Treatment cohort TREAT_CONFIDENCE mean = 3.60 (SD = 1.10); shares across the 1-5 scale were 6%, 12%, 24%, 36%, 22% respectively.
• Treatment cohort TREAT_INTENTION mean = 3.50 (SD = 1.20); shares across the 1-5 scale were 8%, 14%, 22%, 34%, 22% respectively.

Control cohort self-reported hypothetical readiness for a similar reduction shows slightly lower confidence and intention:

• CONTROL_CONFIDENCE mean = 3.40 (SD = 1.10); shares 8%, 14%, 30%, 32%, 16%.
• CONTROL_INTENTION distribution: 10% very unlikely, 18% unlikely, 30% neither, 28% likely, 14% very likely.

These immediate measures will be used as mediators in pre-specified mediation analysis and as covariates in adjusted ITT models.

9.3. Perceived effort and perceived return

Perceived effort items (E1-E3) and perceived return items (R1-R3) indicate average perceived difficulty and expected benefit.

• Perceived effort item means in Treatment: PERCEIVED_EFFORT_1 mean = 3.00 (SD = 1.10); PERCEIVED_EFFORT_2 mean = 3.10 (SD = 1.10); PERCEIVED_EFFORT_3 mean = 2.50 (SD = 1.00) (note: PERCEIVED_EFFORT_3 is reverse-coded in analysis plans).
• Perceived return item means in Treatment: PERCEIVED_RETURN_1 mean = 3.80 (SD = 1.00); PERCEIVED_RETURN_2 mean = 3.50 (SD = 1.10); PERCEIVED_RETURN_3 mean = 3.70 (SD = 1.00).

Overall perceived-return means are similarly high: PERCEIVED_RETURN_1 overall mean = 3.65 (SD = 1.05), indicating respondents generally expect benefits from reducing phone use before bed.

9.4. Reported constraints

Respondents selected multiple constraints that would make it difficult to reduce phone use before bed. The most frequently selected options overall were:

Average number of options selected per respondent overall = 2.75.

Habitual checking is the dominant barrier (59% overall; Treatment 60%; Control 58%), which suggests interventions addressing habit disrupting or alternative evening activities may be necessary complements to commitment prompts.

9.5. Willingness-to-pay for a one-night blocking feature

Respondents reported limited immediate willingness to pay for a one-night blocking and notification-locking feature offered as a one-night add-on. The distribution of maximum WTP selections is shown in the table below.

WTP distribution (EUR)

Median WTP = 0 EUR; IQR = 0 EUR-1.99 EUR.

Median zero willingness-to-pay indicates many respondents would not pay out of pocket for a single night of automated blocking, although a nontrivial minority express small positive WTP.

9.6. Attention and opt-in rates

Attention-check results indicate high levels of careful responding. Overall ATTN_CHECK responses: Strongly disagree 3%, Disagree 4%, Neither 6%, Agree 46%, Strongly agree 41% (Agree + Strongly agree = 87% overall). Treatment cohort reported 45% Agree and 45% Strongly agree (90% combined); Control cohort 47% Agree and 40% Strongly agree (87% combined).

Opt-in to the next-morning follow-up was substantial: Treatment 60% opted in, Control 58% opted in, overall opt-in = 59%.

These opt-in rates, combined with follow-up reminders and a 1.00 EUR-equivalent incentive on first reminder, position the study to obtain a sizeable follow-up sample for analysis of FOLLOWUP_RESTFULNESS and compliance.

10. Planned primary and secondary analyses

This section restates the pre-registered analysis plan and operational decisions made under the gaps defaults.

10.1. Primary ITT analysis

• Outcome: FOLLOWUP_RESTFULNESS (1-5).
• Estimator: ordinary least squares regression estimating the mean difference in FOLLOWUP_RESTFULNESS between Treatment and Control assignment (binary indicator), with robust (HC3) standard errors clustered by country.
• Rounding: report mean differences to two decimals and 95% confidence intervals to two decimals.
• Covariate adjustment: pre-specified covariates (baseline minutes numeric or categorical, baseline sleep quality, age group, gender, country strata) included in adjusted models to improve precision; unadjusted ITT also reported.
• Inclusion: primary analysis includes follow-up completions within 24 hours after reported bedtime, accepting completions up to 48 hours; responses outside 48 hours are excluded from the primary ITT but included in sensitivity checks.
• Multiple testing: two-sided alpha = 0.05 for the primary ITT; secondary and subgroup tests adjusted with Benjamini-Hochberg FDR at 5%.

10.2. Secondary and robustness analyses

• CACE: two-stage least squares with assignment instrumenting self-reported compliance (binary, defined as reduction in follow-up pre-sleep minutes ≥30 relative to baseline). Partial compliance will be described but not counted as compliance for the main CACE.
• Heterogeneity: subgroup ITT estimates by BASE_MINUTES_CAT, age groups (18-29, 30-39, 40-50), and baseline sleep quality. Pre-specified interaction tests will be reported and FDR-adjusted.
• Mediation: path analysis and causal mediation analysis to assess the proportion of any observed ITT effect mediated by TREAT_INTENTION/TREAT_CONFIDENCE and PERCEIVED_RETURN_1-3.
• Sensitivities: results presented with and without QC exclusions, with survey weights applied, with cluster-robust SEs and without clustering; ordered-logit models for FOLLOWUP_RESTFULNESS as alternative specification.

10.3. Missing data and imputation

• Missing covariate data will be handled with multiple imputation where necessary; missing outcome data is handled implicitly in the ITT by retaining all randomized units for whom outcome is observed (primary ITT is based on observed outcomes within the 24-48 h window). Inverse-probability weighting using opt-in and completion models will be used as robustness checks to address selective follow-up.

11. Risks, limitations, and mitigation strategies

• Nonresponse to follow-up: mitigated by two automated reminders (12 h and 36 h) and a 1.00 EUR-equivalent incentive on first reminder; main analyses will present inverse-probability weighted estimates and sensitivity bounds.
• Self-report bias and measurement error: mitigated by numeric follow-up device-use checks (FOLLOWUP_MINUTES_NUMERIC) and explicit, clear instructions emphasizing anonymity.
• Selection and generalizability: the sampling frame combines on-site users and matched panels with post-stratification weighting to target margins; however, participants remain website visitors and panelists and may not represent all European adults 18-50.
• One-night intervention: inference is limited to single-night commitment effects; long-term behavioral change and clinical sleep outcomes are out of scope.

12. Deliverables and timeline

Deliverables include this final report, a detailed methods appendix with full instruments and QC logs, pre-registered analysis code and de-identified cleaned data files, SVG-ready figures and PNG fallbacks, and two site-ready copy blocks for StyleCodes summarizing consumer-facing messaging. Primary outcome analyses will commence after the close of the 48-hour follow-up window and completion of QC, at which point ITT and CACE estimates, subgroup plots, and mediation analyses will be produced.

13. Conclusion

This pre-registered randomized survey experiment achieved the planned balanced sample (754 per arm; overall n = 1,508) and demonstrates strong immediate engagement with the commitment prompt: 55% of those in Treatment agreed to the commitment, and mean TREAT_CONFIDENCE and TREAT_INTENTION were 3.60 and 3.50 respectively on 1-5 scales. Baseline distributions show appreciable pre-sleep smartphone use (35% 31-60 minutes; 18% >60 minutes), and perceived returns from reduction are generally positive (PERCEIVED_RETURN_1 mean ≈ 3.65 overall). Attention-check pass rates and opt-in rates are adequate for the planned primary ITT and secondary analyses. Follow-up data collection and QC processing will enable estimation of the registered ITT effect of the commitment prompt on next-morning subjective restfulness and complier effects among those who report a ≥30-minute reduction.