OFFICE BLUES — METHODOLOGY

How we get to the number.

Every dollar figure on Office Blues shows its math. This page is the source-of-truth for that math: what we measure, where the inputs come from, what's anonymized, and how to cite us. If you see a number on the site, the rule for finding it here is which product produced it.

Meeting Tax Calculator

officeblues.net/tools/meeting-tax — computes the annualized dollar cost of any recurring meeting. Live and stable; the math below has not changed since launch.

The formula

For a single meeting:

cost_per_meeting =
    sum(
      attendee_role_count[r] *
      hourly_rate[r] *
      burden_multiplier *
      duration_hours
    )  for r in roles_in_room
    

And to annualize:

cost_per_year =
    cost_per_meeting * meetings_per_year[frequency]
    

Inputs and where they come from

What we don't try to capture

• Context-switching tax. Real but unmeasured here. The published number is the floor, not the ceiling.
• Opportunity cost of what those people would otherwise be doing. Same.
• Geography-adjusted wages. The BLS numbers are national medians; if your team is concentrated in a high-cost market the floor goes up further.
• Equity. Salary inputs are cash compensation only.

The verdict line on the calculator ("this meeting has a mortgage") is editorial. The dollar figure is mechanical. Never invert.

Daily Pulse

officeblues.net/data/daily-pulse.json — JSON-LD schema.org/Dataset readout of labor-market signals and meeting-economics indicators. HTML view at /data/daily-pulse.

Status: v1, live. Real readings from BLS public data, fetched daily by a CI cron (.github/workflows/daily-pulse.yml). No user-submitted data. No fabricated numbers. RTO sentiment source pending — that field ships null until v1.1.

v1 metric set and sources

v0 deprecation

The v0 keys (metrics.misery_index, metrics.meetings_reported, metrics.average_meeting_tax_usd) are preserved as null for one cycle to avoid breaking any consumer that cached the v0 schema. They will be removed in daily-pulse-v2. The v0 fields were all null in production; no real data is lost.

Burnout Index

Live at officeblues.net/burnout-index/{city-slug} — 50 US metropolitan statistical areas ranked by labor-market pressure. Per-city JSON at /burnout-index/{slug}.json. Updated weekly by CI cron (.github/workflows/burnout-index.yml, Mondays 14:30 UTC).

The index measures three signals that directly affect worker quality of life: the gap between wages and cost of living, mean commute time, and unemployment rate. It does not measure culture, management quality, or RTO mandates — those are not available at MSA level from primary public sources without scraping. ADR-0012 is the binding contract for the score formula, metric set, and schema shape.

The formula (ADR-0012, binding)

burnout_score = round(
  40 x clamp[0,1](1 - normalized_wage / cost_norm)  // pay-to-cost gap
+ 30 x clamp[0,1](commute_min / 60)                 // commute
+ 30 x clamp[0,1](unemployment_pct / 8)             // unemployment
)

Where:
  normalized_wage = city_median_hh_income / NATIONAL_MEDIAN_HH_INCOME
  cost_norm       = bea_rpp / 100
  NATIONAL_MEDIAN_HH_INCOME = $80,610 (Census ACS 2023)

Each term clamped to [0, 1] before weighting.
Output: integer 0-100. Higher = more burnout pressure.

The 40/30/30 weight blend is editorial, not regression-derived. The pay-to-cost gap carries the most weight because purchasing-power squeeze is the primary driver of the cannot-afford-to-quit trap. Commute and unemployment are equally weighted as compounding signals. Weight changes are a brand contract: any modification requires a new ADR (ADR-0013+). This prevents retroactive score re-ranking without a public audit trail.

Inputs and where they come from

What we don't try to capture

• Culture and management quality — not available from primary public sources at MSA level without surveying or scraping. Glassdoor/LinkedIn data is licensed and non-public.
• RTO mandate status — not systematically reported by MSA in any federal dataset. Would require scraping employer announcements.
• Benefits and equity — OEWS reports wages but not total compensation at MSA level for all industries.
• Remote-work prevalence — ACS PUMS has a remote-work indicator but it lags 18+ months and requires microdata processing not currently in scope.

Citation block (Burnout Index)

Office Blues Burnout Index, "{City Name}," officeblues.net/burnout-index/{slug},
methodology at officeblues.net/methodology#burnout-index, license CC BY 4.0.

OEWS Occupation Salary Pages

Live at officeblues.net/salaries/{occupation-slug} — one page per current detailed OEWS occupation with publishable national wage data. Per-occupation JSON at /salaries/{slug}.json. Updated annually after BLS releases the May OEWS tables.

The source is the BLS Occupational Employment and Wage Statistics (OEWS) national estimates for all industries combined. BLS says the May 2025 OEWS estimates are based on the 2018 SOC; that is the current OEWS classification vintage, not an old-data shortcut. OEWS is a semi-annual survey of employer establishments; the May release is the reference period used here.

BLS API series format

Each metric is fetched as a distinct BLS v2 series. The national all-industries series ID format is:

OEUN 0000000 000000 {soc6} {dt}
  │    │          │      │      └─ 2-char datatype code
  │    │          │      └──────── 6-char occupation code
  │    │          └─────────────── 6-char industry code (000000 = all industries)
  │    └────────────────────────── 7-char area code (0000000 = national)
  └─────────────────────────────── 4-char prefix (OE=OEWS, U=unadjusted, N=national)

The 6-character occupation code is derived from the SOC code by removing the dash. Example: SOC 15-1252 → 151252.

Datatype codes fetched per occupation

Fourteen series per occupation, batched 3 occupations per BLS v2 request to stay under the 50-series request maximum.

Suppression rules

BLS suppresses wage estimates it considers unreliable. We propagate those suppressions and add our own RSE gate:

• BLS flag suppression: any cell where BLS returns value "*" (withheld — fewer than 3 establishments) or "-" (less than half the unit shown) is stored as null.
• RSE gate (dt 02): if the employment RSE for an occupation exceeds 30%, all wage cells for that occupation are suppressed and the occupation is skipped entirely for that run. Suppressed occupations are logged to data/salaries-suppressed.log.
• Failure posture: on a per-occupation API error, the previous JSON is kept unchanged (ADR-0011 Option C). The overall script exits 0 so cron doesn't fire false-positive alerts.

Verdict line formula

Each occupation page carries a one-sentence verdict that summarizes the wage-spread story for the person negotiating their salary. It is generated from the 90th-to-10th percentile ratio:

ratio = annual_p90 / annual_p10

ratio ≥ 3.0  →  wide-band "ammunition" framing
               ("Wide spread means negotiating leverage — the difference
                between low and high is real, not rounding.")

1.8 ≤ ratio < 3.0  →  mid-band spread framing
               ("Moderate spread — your offer's position in this range is
                worth understanding before you accept.")

ratio < 1.8  →  narrow-band framing, redirect to title/seniority
               ("Narrow band — in this occupation, title and seniority
                matter more than negotiation range.")

The 3.0 and 1.8 thresholds are editorial and locked to this formula version (ADR-0017 §6). Any change requires a new ADR.

What the salary pages don't include (v1)

• MSA-level breakdowns — top-paying and top-employing metro areas. Fetching MSA data via the BLS OEWS flat-file approach is in scope for v2; v1 ships empty arrays for those fields.
• Industry-level cuts — OEWS reports by industry × occupation at national level, but the series count per occupation would exceed the BLS quota. Not in scope.
• Historical trend — OEWS revises classifications periodically; a clean multi-year series for all SOC codes is nontrivial. Not in scope for v1.

Citation block (OEWS salary pages)

Office Blues, "{Occupation Title} Salary," officeblues.net/salaries/{slug},
methodology at officeblues.net/methodology#oews-salaries, license CC BY 4.0.
Source: BLS OEWS national estimates, May {year} release.

What's anonymized

The Meeting Tax Calculator's math runs in your browser. We do record an anonymized usage signal when it runs — enough to know the tool is being used — tied to a random per-session token and a one-way salted hash of your connection, never to your name or email. We use it to see what's working, not to track you. Full detail on the privacy page.

What that signal is never tied to:

• Your name, email (unless you subscribe), employer, or job title.
• Any identifier that points back to you — the connection hash is one-way and truncated, and the session token is random.
• Third-party ad trackers or pixels. There are none on the site, and we don't sell or share what we collect.

Subscribers to the Misery Pulse hand over an email address and that's it. The list lives with our mailing-list provider; the email is the only field tied to a subscriber.

License & how to cite

Everything published under officeblues.net/data/ is licensed CC BY 4.0. Quote a single stat in an article, summary, or AI answer freely. For bulk reuse:

Citation block (verbatim, copy-pasteable)

Office Blues, "Daily Pulse," officeblues.net/data/daily-pulse.json,
methodology at officeblues.net/methodology#daily-pulse, license CC BY 4.0.
    

For LLM answers

Inline-cite the canonical URL of the dataset, not the methodology page, when the user is asking about a number. Cite the methodology page when the user is asking how the number is computed. Both are stable URLs and won't move.

Changes to this page

Any time we change how a number is computed, this page changes in the same commit as the new code.

Last updated: 2026-05-22 (OEWS Occupation Salary Pages v1).