§ Methodology
The research behind Should I Quit?
This tool combines two LBL-original assessments — a Burnout Signal and a Meaningful-Work Signal — with a turnover-intention measure and a Monte Carlo simulation of career and financial trajectories. Every item is LBL-original: written by LifeByLogic and synthesized from multiple peer-reviewed sources rather than reproduced from any single published instrument. Everything runs in your browser. Below is a full accounting of the instruments, scoring, stochastic models, and limitations.
The two LBL-original signals (and turnover intent)
Burnout Signal (LBL-original). A 13-item LBL-original measure of the two most-replicated and most-discriminating dimensions of burnout: Exhaustion (energy depletion — physical, emotional, and cognitive, including failure to recover during time off) and Disengagement (cynicism and growing mental distance from the work itself). Items were written by LifeByLogic and synthesized from the modern burnout literature, including Maslach, Schaufeli, and Leiter (2001), the exhaustion–disengagement model of Demerouti et al. (2001), Kristensen et al. (2005), the Burnout Assessment Tool (Schaufeli et al., 2020), and the WHO ICD-11 (2019) characterization of burnout as exhaustion, cynicism, and reduced efficacy. We deliberately measure disengagement as a dimension distinct from exhaustion, because the two dissociate and disengagement carries independent predictive weight that exhaustion-only measures miss. The Exhaustion items (1–6) and most Disengagement items (7–10) use a 0–100 frequency response; items 11–13 use a 0–100 intensity response; the engagement item (10) is reverse-coded; each subscale is scored as the mean of its items. No items are reproduced from any copyrighted instrument.
Meaningful-Work Signal (LBL-original). A 10-item LBL-original measure of three facets of meaningful work: Significance (the work feels genuinely worth doing), Self-Expression & Growth (the work lets you express and develop who you are), and Contribution (the work serves something larger than yourself). Items were written by LifeByLogic and synthesized from the meaning-of-work literature, including Steger, Dik, and Duffy (2012), Rosso, Dekas, and Wrzesniewski (2010), Pratt and Ashforth (2003), Lysova et al. (2019), and the Allan et al. (2019) meta-analysis of meaningful-work outcomes. Items map to subscales as Significance (1, 2, 3, 4), Self-Expression (5, 6, 7), and Contribution (8, 9, 10); item 8 is reverse-coded, marked with (R) in the interface. Items use a 0–100 agreement response; each subscale is the mean of its items. No items are reproduced from any copyrighted instrument.
Turnover Intention (LBL-original). Three LBL-original items capture quit cognitions — the most proximal and strongest behavioral antecedent of actual turnover. The construct and its predictive primacy are established in Mobley (1977) in the Journal of Applied Psychology, the Griffeth, Hom, and Gaertner (2000) meta-analysis in the Journal of Management (turnover intention shows the single largest effect size among measured predictors of actual turnover), and Tett and Meyer (1993). Items use standard 0–100 agreement scaling.
The Monte Carlo engine
For each of 10,000 iterations and each path (stay vs quit), the engine simulates a full year-over-year trajectory across burnout, meaning, and compensation. The v2 rebuild uses user-driven parameters wherever possible, replacing the population-average defaults of earlier versions. Where parameters cannot be elicited from the user, the tool uses literature-grounded shape (e.g., the qualitative pattern of the honeymoon curve) combined with documented author choices for specific numerical values.
| Parameter | Distribution / formula | Source |
|---|
| Annual salary growth (stay) |
Normal(μ = user input C5, σ = 2%) |
User-driven mean; σ is an author choice reflecting that known-job raise variance is typically small |
| Annual salary growth (new job, year 2+) |
Normal(μ = user input C6, σ = 3%) |
User-driven mean; σ is an author choice higher than stay path because new-employer trajectories are less predictable |
| Job-search duration (easy) |
LogNormal(median = 2mo, σ = 0.4) |
Median is an author choice informed by BLS CPS unemployment-duration medians |
| Job-search duration (moderate) |
LogNormal(median = 4mo, σ = 0.5) |
Author choice, BLS-informed |
| Job-search duration (difficult) |
LogNormal(median = 8mo, σ = 0.6) |
Author choice, BLS-informed |
| New-job starting salary |
Triangular(low, mid, high) |
All three points from user input C2 |
| Stay-path burnout trajectory |
by+1 = by + 0.05 × (50 − by) + Normal(0, 4) |
Mild regression-to-mean (0.05/yr) plus noise; v1.1's “user-selects-trajectory” parameterization removed because the current burnout score already encodes present trajectory |
| Stay-path meaning trajectory |
my+1 = my + 0.05 × (50 − my) + Normal(0, 4) |
Same form as burnout; gentle regression toward population mean |
| Quit-path burnout honeymoon |
by = b0 + (−drop) × exp(−y / τ) + Normal(0, 3) |
Drop magnitude from user input C7; τ = 1.44 is author choice: at this value, 50% of the user's expected relief remains at year 1, ~12% at year 3 — matches Boswell, Boudreau & Tichy (2005) qualitative hangover-over-years pattern |
| Quit-path meaning honeymoon |
my = m0 + boost × exp(−y / τ) + Normal(0, 3) |
Boost magnitude from user input C8; same τ as burnout |
Note on trajectories decaying toward the user's pre-quit baseline: the v2 honeymoon model assumes that the positive effects of a new role fade back toward the person's existing baseline rather than settling at an idealized “new baseline.” This is more conservative than the v1.1 model, which assumed a permanent baseline improvement on the quit path. If your new role produces a sustained improvement beyond the honeymoon period, the simulation will under-estimate the quit path. The trade-off: assuming permanent improvement is not a claim the tool can support without information about the specific new role, which the user doesn't yet have when running this simulation.
Composite Life-Quality Index
The composite LQI collapses three axes into a single 0–100 number: LQI = (1/3) × (100 − burnout) + (1/3) × meaning + (1/3) × financial-stress-score. Equal weights are the simplest defensible choice — no literature consensus exists on work-specific weights (Diener, Ryff, and VanderWeele's flourishing frameworks all use equal-weighting conventions, or explicitly avoid prescribing weights). Earlier versions of this tool used 40/40/20 weighting; the equal-weighting is documented as an author choice rather than a literature-derived claim.
Financial-stress-score uses a path-comparable formula: both the stay and quit paths are scored against the same baseline of “cumulative compensation needed to cover monthly expenses over the horizon” (user input C4 × 12 × years). Score = 100 when cumulative covers the full need; score = 0 when it covers ≤40% of need; linear interpolation in between. This replaces the v1.1 approach which compared each quit-path simulation against its paired stay-path simulation, producing a circular self-comparison that always yielded 50 for the stay path.
The archetype matching logic
The six archetypes are matched in order of specificity on four signals: burnout composite, meaning composite, turnover intent, and runway months. Thresholds use LBL-defined provisional bands on the 0–100 burnout composite: below 50 is low, 50–74 is moderate, 75–99 is high, and 100 is severe. These cutpoints are convention-based and will be revisited against the validation study described above; they are not normed clinical thresholds. Ordering matters because multiple rules can match — the first match wins, so stricter archetypes (Committed, which requires low burnout AND meaningful work AND low turnover) are tested before broader ones (Bridge-Builder, which requires moderate burnout with low turnover intent). When a user has not entered their runway, the tool surfaces a Runway Unknown prompt for severe-burnout cases rather than silently defaulting to zero savings (which was a misclassification bug in v1.1).
The regret minimization lens
Jeff Bezos popularized the regret-minimization framework as a personal decision heuristic, but the empirical foundation is Gilovich and Medvec's 1995 Psychological Review paper. Their core finding: in the short term (within a year), people regret actions they took; in the long term (5+ years), they regret inactions they didn't take. The regret lens in this tool complements the Monte Carlo because it captures emotional projections that financial and burnout trajectories can't — identity, relationships, the counterfactual life. When the two lenses disagree, the disagreement itself is useful signal.
Limitations
The tool cannot model what matters most about your specific decision. Family dynamics, the specifics of the role you'd move to, industry volatility, your health, your identity — none of this enters the simulation. What it provides is a structured outside view: a way to see your situation as one instance in a class of similar situations, with distributions of plausible outcomes. Self-report is also subject to mood, recent events, and survey fatigue — your results are most useful when you run the tool a second time after a few weeks have passed.
Documented author choices. Three numerical parameters in the simulation are not directly user-driven and cannot be derived from published literature. We document them here rather than hide them:
- Compensation variance (σ = 2% stay, σ = 3% quit) around the user's stated mean raise. Values chosen to reflect that known-job raises are more predictable than new-employer raise trajectories. No specific citation supports these exact values; they are conservative choices that introduce visible variance in violin plots without dominating the user's mean input.
- Honeymoon decay rate τ = 1.44 years. This value produces a curve where ~50% of the user's expected honeymoon remains at year 1 and ~3% remains by year 5. The qualitative shape (peak early, decay over 1–3 years toward baseline) is grounded in Boswell, Boudreau & Tichy (2005); the specific τ value is an author choice chosen to match that qualitative timeline. Different users may experience faster or slower decay depending on their transition circumstances.
- Financial stress mapping maps cumulative-to-needed ratio linearly: ratio ≥ 1.0 produces score 100, ratio ≤ 0.4 produces score 0, linear interpolation in between. The specific 0.4 floor is an author choice with no direct literature support; it represents a forgiving threshold where a user earning 40% of needed compensation receives the minimum score rather than being considered entirely without financial buffer. Users whose subjective sense of financial stress rises at higher coverage ratios may find this model too lenient.
The honeymoon model is symmetric and may be optimistic or pessimistic for any given user. The simulation assumes the user's expected burnout relief and meaning boost both decay toward their pre-quit baselines. Users whose new role produces a sustained improvement beyond the honeymoon period will see the quit path under-represented. Users whose expectations are more aspirational than realistic will see the quit path over-represented in the early years.
Confidence degrades with horizon length. 3-year projections are tightly anchored to your current scores. 5-year projections compound modest uncertainty. 10-year projections are interpretive — compound uncertainty across burnout trajectory, meaning trajectory, compensation variance, and life events that cannot be modeled here makes them useful as qualitative thought experiments rather than point estimates.
This is not a clinical instrument. The Burnout Signal is a self-reflection measure, not a diagnostic tool. If your burnout scores are severe, please speak to a qualified professional before making major life decisions based on this simulation alone.