LBL Sleep-Cognition Optimizer — Methodology & Validation

What this tool measures (and what it does not)

The LBL Sleep-Cognition Optimizer measures current behavioral sleep patterns — what your sleep actually does on workdays vs. free days — and sets out a personalized schedule with five levers scored against it.

From 12 LBL-original items, the Optimizer derives:

• Behavioral chronotype via the MSFsc methodology (Roenneberg 2003), expressed as a 0-100 morningness score.
• Social jetlag in minutes, in five bands that are the tool's own 30-minute steps.
• A regularity estimate on a 0-100 scale, the tool's own score for how steady your sleep looks across the week. It borrows its idea from the Sleep Regularity Index (Phillips 2017) but is not that index, and it has no target.
• Biology-behavior gap when paired with the LBL Chronotype Profile.
• Personalized 24-hour schedule with lights-out, wake time, an estimated peak cognitive window, exercise window, wind-down and caffeine cutoff.
• Lever alignment radar across Timing / Caffeine / Exercise / Wind-down / Cognitive, each scored 0-100.
• Five lever tiles with your answer, the recommendation and its reason, in the order your goal (Q12) sets.

The Optimizer does not measure:

• Your biological/trait chronotype directly — that is measured by the companion Chronotype Test.
• Sleep stage architecture or sleep efficiency — polysomnography is required for that.
• Light exposure quantitatively — light timing matters greatly but cannot be captured by self-report.
• Sleep disorders. Persistent symptoms warrant clinical evaluation.

The Optimizer measures what you are doing. The Chronotype Profile measures who you are biologically. Together they compute the biology-behavior gap, a comparison of the two scores rather than a clinical measure.

Why the tool looks at behavior, not only chronotype

Whether you are a morning or evening type is only part of the picture. The Optimizer also estimates two things your answers can show: how far your workday sleep sits from your free-day sleep, and how regular your sleep timing is.

Social jetlag. The construct was introduced by Wittmann et al. 2006. The Arab et al. 2024 meta-analysis (n=231,648 across 43 studies) found social jetlag linked with slightly higher BMI, body fat and waist circumference (for BMI, r = 0.12, varying widely between studies), a link its authors say needs further work to show cause.

Sleep regularity. The Sleep Regularity Index of Phillips et al. 2017 measures how often a person is asleep or awake at the same clock times 24 hours apart, from night-by-night sleep records; in their study of college students, less regular sleep went with later sleep timing and lower grades. Four typical times cannot give that index, so the Optimizer gives its own regularity estimate instead (Section 7). It has no target and cannot be compared with the published index's values.

Biology-behavior gap. When both the Chronotype Profile and Optimizer are taken, the gap compares the chronotype score your Chronotype Test answers give with the score your sleep times here give, on the same 0-100 scale. A 30-point gap means your current sleep times sit well away from the times your Chronotype Test answers describe; the labels are the tool's own conventions (Section 10), not a clinical measure.

This is why the Optimizer puts these behavioral figures first rather than treating chronotype as the headline: your sleep times are something you can change, and the figures show where they stand now.

Instrument structure

The Optimizer consists of 12 items in 5 sections.

Q12 (optimization goal) is conditional: "Reduce biology-behavior gap" is shown only when Profile data exists in localStorage, beside "Reduce social jetlag". Sleep regularity, a smaller gap and less social jetlag put sleep timing consistency first among the levers, peak cognitive performance puts cognitive task scheduling first, and better mood and energy puts first whichever lever scores lowest.

Scoring algorithm — full pseudocode

The exact algorithm running in your browser, and on the report's server since October 2026.

function midsleep(sleepMinutes, wakeMinutes):
    # Compute midpoint of sleep window, handling cross-midnight wraparound
    if sleepMinutes > wakeMinutes:
        adjustedSleep = sleepMinutes - 1440   # treat as previous day
        midpoint = (adjustedSleep + wakeMinutes) / 2
        return (midpoint + 1440) modulo 1440  # normalize to 0-1440 range
    else:
        return (sleepMinutes + wakeMinutes) / 2

function sleepDuration(sleepMinutes, wakeMinutes):
    if sleepMinutes > wakeMinutes:
        return ((1440 - sleepMinutes) + wakeMinutes) / 60   # hours
    else:
        return (wakeMinutes - sleepMinutes) / 60

function computeBehavioralChronotype(answers):
    wdMid = midsleep(answers.Q2, answers.Q3)        # workday midsleep
    fdMid = midsleep(answers.Q4, answers.Q5)        # free-day midsleep
    wdDur = sleepDuration(answers.Q2, answers.Q3)
    fdDur = sleepDuration(answers.Q4, answers.Q5)
    weeklyAvg = (5 * wdDur + 2 * fdDur) / 7

    # MSFsc: sleep-debt-corrected midsleep on free days (Roenneberg 2003)
    msfsc = fdMid
    if fdDur > weeklyAvg:                            # free-day oversleep
        excessMinutes = (fdDur - weeklyAvg) * 30     # half of excess hours
        msfsc = (fdMid - excessMinutes + 1440) modulo 1440

    # Linear mapping: MSFsc clock time → 0-100 morningness score
    # 2:00 AM → ~75 (morning); 4:00 AM → ~50; 6:00 AM → ~25; 8:00 AM → ~0
    msfscClock = msfsc
    if msfscClock > 720:
        msfscClock = msfscClock - 1440   # handle evening-hours edge case
    score = round(100 - ((msfscClock - 60) / 6))
    return clip(score, 0, 100)

function computeSocialJetlag(answers):
    wdMid = midsleep(answers.Q2, answers.Q3)
    fdMid = midsleep(answers.Q4, answers.Q5)
    diff = abs(fdMid - wdMid)
    if diff > 720: diff = 1440 - diff             # handle wraparound
    if diff < 30:  return diff, "Minimal"
    if diff < 60:  return diff, "Mild"
    if diff < 90:  return diff, "Moderate"
    if diff < 120: return diff, "High"
    return diff, "Severe"

function computeSRIProxy(answers, sjl):
    # The tool's own regularity estimate. The published Sleep Regularity
    # Index needs night-by-night records (Phillips 2017); this has no target.
    wdDur = sleepDuration(answers.Q2, answers.Q3)
    fdDur = sleepDuration(answers.Q4, answers.Q5)
    durationDelta = abs(wdDur - fdDur)   # hours

    sjlPenalty = min(40, sjl.minutes * 0.3)
    durationPenalty = min(20, durationDelta * 8)
    continuityPenalty = (answers.Q7 - 1) * 3

    sri = 100 - sjlPenalty - durationPenalty - continuityPenalty
    return clip(round(sri), 0, 100)

function computeBiologyBehaviorGap(behavioralScore, profile):
    if profile is null: return null
    bioScore = profile.biologicalChronotype.calibratedScore
    gap = abs(bioScore - behavioralScore)
    if gap < 10:  return gap, "Well-Aligned"
    if gap < 20:  return gap, "Mild Misalignment"
    if gap < 35:  return gap, "Moderate Misalignment"
    return gap, "Severe Misalignment"

function generateSchedule(answers, behavioral):
    chronotypeScore = behavioral.score   # Q2-Q5 are always answered
    targetDuration = 7.5   # hours, a planning default inside the 7-9 h advised for adults (Hirshkowitz 2015)
    latencyMinutes = [7, 15, 30, 65, 100][answers.Q6 - 1]   # counted once

    mustWake = parseTime(answers.Q10)
    sleepOnset = (mustWake - targetDuration*60 + 1440) modulo 1440
    bedtime = (sleepOnset - latencyMinutes + 1440) modulo 1440   # lights-out

    # Estimated peak cognitive window, shifted by the chronotype score
    peakStart = round(mustWake + (180 - (chronotypeScore - 50) * 1.5) + 1440) modulo 1440
    peakEnd   = (peakStart + 120) modulo 1440

    # The levers' limits
    caffeineCutoff = (sleepOnset - 420 + 1440) modulo 1440   # 7 h before sleep (Drake 2013)
    exerciseStart  = (sleepOnset - 300 + 1440) modulo 1440
    exerciseEnd    = (sleepOnset - 180 + 1440) modulo 1440   # 3 h before sleep (Stutz 2019)
    windownStart   = (bedtime - 60 + 1440) modulo 1440

    return {bedtime, sleepOnset, mustWake, peakStart, peakEnd,
            caffeineCutoff, exerciseStart, exerciseEnd, windownStart}

function lateness(t, limit, wake):
    # Minutes a time runs past a limit in the day from waking to the planned
    # sleep; a time inside the planned 7.5 h night counts from its nearer end.
    pos = (t - wake + 1440) modulo 1440
    lim = (limit - wake + 1440) modulo 1440
    night = 1440 - 450
    if pos <= lim: return 0
    if pos <= night or pos - night <= 1440 - pos: return pos - lim
    return 0

function computeAlignmentScores(answers, schedule, regularity):
    timing = regularity.score
    # Q8 counts at the latest time its answer allows (after 6 PM as 8 PM)
    caffeine = 100 if answers.Q8 == "none" else 40 if answers.Q8 == "varies"
               else max(0, round(100 - lateness(latest(answers.Q8), schedule.caffeineCutoff, schedule.mustWake) / 60 * 15))
    exercise = 50 if answers.Q9 == "none" else 45 if answers.Q9 == "varies"
               else max(0, round(100 - lateness(answers.Q9, schedule.exerciseEnd, schedule.mustWake) / 60 * 12))
    windDown = [100, 95, 80, 55, 30][answers.Q6 - 1]
    # Demand windows: before 9 AM, 9-noon, noon-3 PM, 3-6 PM, after 6 PM
    cognitive = 50 if answers.Q11 == "variable"
                else max(0, 100 - 25 * abs(window(answers.Q11) - window(schedule.peakStart + 60)))
    return {timing, caffeine, exercise, windDown, cognitive}

function leverOrder(goal, scores):
    levers = [caffeine, exercise, windDown, cognitive, consistency]   # consistency scores timing
    first = {regularity: consistency, cognition: cognitive, gap: consistency, sjl: consistency}[goal]
    if goal == "mood": first = the lowest-scoring lever, the earlier one on a tie
    return [first] + the other levers in order
    

The free result is worked out in your browser. Your answers leave it only if you buy The Bedtime Bridge: your 12 answers, and your Chronotype Test score when you have one, go to checkout so the report's server can work out the same figures and write your report.

MSFsc derivation — using the open scientific method

MSFsc (sleep-debt-corrected midsleep on free days) is the chronotype metric introduced by Roenneberg, Wirz-Justice & Merrow in 2003. The key insight: people's sleep timing on workdays is biased by alarm clocks and obligations; their sleep timing on free days reflects biological preference; but free-day timing is itself biased by recovery from accumulated weekday sleep debt. MSFsc corrects for that recovery oversleep.

Crucially, the MSFsc methodology is open published science. What is copyrighted is the MCTQ questionnaire — the specific item wording and response options used in published MCTQ studies. We do not administer the MCTQ. We collect sleep onset and wake times using our own LBL-original item wording (Q2-Q5), then apply the published MSFsc computation to those data.

The formula used:

• MSF = midsleep on free days = (free-day-sleep-onset + free-day-wake) / 2, handling cross-midnight wraparound.
• Weekly average sleep duration = (5 × workday-duration + 2 × free-day-duration) / 7.
• If free-day duration > weekly average (oversleep), subtract half of the excess hours from MSF to get MSFsc.
• If free-day duration ≤ weekly average (no oversleep), MSFsc = MSF.

Mapping MSFsc to the 0-100 morningness score is a linear approximation: MSFsc 2:00 AM → ~75 (morning); 4:00 AM → ~50; 6:00 AM → ~25; 8:00 AM → ~0. The true relationship is non-linear at extremes but linear is adequate for the educational use case.

Social jetlag thresholds and rationale

The five social jetlag bands are the tool's own conventions in 30-minute steps, not health thresholds:

The Arab et al. 2024 meta-analysis pooled 43 studies (n=231,648) and found social jetlag linked with slightly higher BMI, body fat and waist circumference (for BMI, r = 0.12, varying widely between studies), a link its authors say needs further work to show cause. It set no cut-point, so the bands describe how far apart your workday and free-day sleep sit, not a level of health risk; the result marks whether yours is under an hour or an hour or more.

The regularity estimate and its disclosure

The Sleep Regularity Index (SRI) was developed by Phillips et al. 2017 and quantifies how consistently a person sleeps at the same times day-to-day. It is computed from night-by-night sleep records, such as a sleep diary or actigraphy, over days to weeks: the chance of being in the same state, asleep or awake, at two times 24 hours apart, scaled so that 100 means sleeping and waking at exactly the same times every day.

Four typical times cannot give that index. Instead, the Optimizer computes its own regularity estimate from your reported workday vs. free-day timing and night waking. The estimate:

• Starts at 100 (perfect regularity).
• Subtracts up to 40 points penalty for social jetlag magnitude (0.3 points per minute, capped).
• Subtracts up to 20 points penalty for sleep duration variance between workdays and free days (8 points per hour of difference, capped).
• Subtracts up to 12 points penalty for sleep continuity (Q7 awakening frequency, 3 points per scale step).

It has no target and cannot be compared with the published index's values: a score here and a published SRI score do not mean the same thing, so no cut-off from SRI studies applies to it.

We disclose that this is the tool's own estimate, not a measured Sleep Regularity Index. How closely it tracks the published index in the same people (Level 3 convergent validity) is not yet established. People who want the published index need several weeks of night-by-night records, such as a sleep diary or a research-grade actigraphy wearable.

Schedule generation algorithm

The recommended schedule starts from the user's hard constraint, the must-wake time (Q10), and the time falling asleep takes (Q6). The sleep window follows the same rule for everyone; the chronotype score moves only the estimated cognitive peak.

The schedule is a starting point, not a fixed prescription. Sleep adapts gradually, so the result suggests moving lights-out earlier in 15-minute steps, holding each step for a few nights.

The lever alignment radar

The radar shows how your current answers line up with the recommended schedule on the five levers. Each axis is scored 0-100; higher = better aligned.

The radar shape is informative: a regular pentagon-like polygon = well-aligned across all dimensions. Collapsed inward on specific axes = where the largest leverage exists for sleep optimization. Each lever tile below the radar shows your answer, the recommendation for that lever and its reason, and your goal (Q12) decides which tile comes first.

Biology-behavior gap analysis

When the user has taken the LBL Chronotype Profile (Profile data exists in localStorage under key 'lbl-chronotype-profile'), the Optimizer computes:

gap = |biology_chronotype - behavior_chronotype|

where biology_chronotype is the calibrated score from the Profile and behavior_chronotype is the 0-100 score derived from MSFsc in this Optimizer. The four labels are this tool's own conventions, not clinical thresholds.

The direction of the gap is also computed and reported: "behavior more evening than biology" vs. "behavior more morning than biology". The schedule does not change with the gap or its direction: lights-out comes from the must-wake time and Q6 alone, and choosing to reduce the gap (Q12) puts sleep timing consistency first among the levers.

Validation strategy

Users should treat the Optimizer as an educational tool for sleep self-assessment and schedule personalization. Schedule recommendations are evidence-grounded but not validated as superior to alternatives in randomized trials.

Edge cases and handling

• Shift workers: The Optimizer assumes typical workday/free-day patterns. Rotating shift workers should consult specialized clinical guidance; the standard recommendations do not apply, and The Bedtime Bridge report is sold only to adults on regular day schedules.
• Cross-midnight sleep: Sleep onset after midnight (e.g., 1 AM) and wake before noon is handled by the midsleep wraparound logic. Sleep onset before midnight (e.g., 9 PM) is treated as previous-day for midpoint computation.
• Oversleep correction: If a free-day duration exceeds weekly average, MSFsc subtracts half the excess hours. This is the Roenneberg 2003 correction. For users without weekday sleep debt, MSFsc = MSF (unchanged).
• No regular caffeine / no regular exercise: Q8 and Q9 have "I don't drink caffeine" and "I don't exercise regularly" options. No caffeine scores 100 on the caffeine lever, and no regular exercise scores a neutral 50 rather than zero.
• Varies widely: Q8 and Q9 have "Varies widely" options. These score 40 (caffeine) and 45 (exercise), because a time that varies cannot be checked against its cutoff.
• No Profile data: The biology-behavior gap is not computed and the goal of reducing it is not offered; everything else is the same, since the schedule always uses the chronotype the sleep times give. Q1's self-selected chronotype is shown beside it, and the user is encouraged to take the Profile for full analysis.
• Hard wake-time conflict: Lights-out comes from the must-wake time (Q10) and Q6 whatever the chronotype, so an evening type with an early start may find the recommended bedtime implausibly early. The other levers (caffeine timing, exercise and wind-down) still apply, and the result suggests moving lights-out earlier in small steps.
• Long time to fall asleep or night waking: If falling asleep takes 45 minutes or more (Q6) or the user often struggles to fall back asleep (Q7), the result asks them to talk to a doctor before moving their bedtime earlier, since an earlier lights-out can mean more time lying awake, and the page offers no report.

Limitations

The Optimizer inherits limitations common to all self-report behavioral sleep assessments, plus several v1.0-specific limitations.

Self-report limitations:

• Recall bias for typical bed/wake times; people overestimate sleep duration and underestimate latency.
• "Typical" answer assumes consistent pattern; users with high variability give noisy answers.
• Self-reported caffeine and exercise timing simplify what is often a varied pattern.

Implementation-specific limitations of v1.0:

• The regularity estimate is self-report only; how closely it tracks the published index is untested.
• Light timing is not measured (cited as future direction with wearable integration).
• Schedule recommendations not validated in RCT.
• Cross-cultural validity of self-report items untested.
• The Optimizer is designed for typical sleep schedules; rotating shift workers, polyphasic sleepers, and those with sleep disorders require clinical guidance.

What this Optimizer is NOT:

• Not a clinical diagnostic instrument for sleep disorders.
• Not a substitute for polysomnography or actigraphy.
• Not validated in RCT for effectiveness; recommendations are evidence-grounded but not proven superior.
• Not appropriate for users with suspected DSPD, ASPD, Non-24, narcolepsy, sleep apnea, RLS, or Shift Work Disorder — these require clinical evaluation.

Independent review

The Optimizer v1.0 instrument design, scoring algorithm, schedule generation logic, and validation framework were reviewed by:

• Abiot Y. Derbie, PhD — Cognitive neuroscientist specializing in neurodevelopmental neuroimaging. Reviewed MSFsc derivation correctness, SJL formula, SRI proxy disclosure framing, and schedule generation logic.
• Eskezeia Y. Dessie, PhD — Researcher in behavioral and brain sciences. Reviewed item wording, edge-case handling, prescription alignment scoring, and clinical safety framing.

Their review covers v1.0 as released on May 13, 2026; the repair of October 14, 2026 is set out in the version log. Neither reviewer is a clinical sleep medicine specialist. Users with suspected sleep disorders should consult a physician board-certified in sleep medicine.

Version log

Next scheduled review: November 13, 2026. Unscheduled review triggers: publication of new MSFsc or SRI validation studies; new evidence on social jetlag or sleep regularity; methodology errors identified via corrections; user-reported edge cases.

Methodology FAQ

Last reviewed

This methodology page was last reviewed on October 27, 2026. Next scheduled review: November 13, 2026.

How to cite this methodology

Related pages