Frame the decision
Define reader, operating decision, geography, cohort, period, and what would change the answer.
Research draft
Noindex pending exact-claim approval.
Methodology · evidence governance
From a useful question to a bounded claim.
The research system keeps official records, first-party operating data, interviews, and directional web signals in their proper lanes. A polished page never upgrades weak evidence.
DENOMINATORS STAY ATTACHEDCURRENT ≠ HISTORICALRECORDS ≠ OUTCOMESEXACT-DRAFT APPROVAL
Collection sequence
Six gates before publication.
Evidence-backed review drafts move through the content contract and approval system. A finished local page is still not authorization to publish it.
Inventory evidence
Classify official records, first-party operating data, interviews, public web evidence, and directional signals separately.
Normalize and bound
Keep denominators, event definitions, exclusions, status, and source dates attached.
Build the visual argument
Plan tables, charts, maps, source captures, captions, mobile behavior, and accessible data before prose.
Review exact claims
Check evidence IDs, privacy, rights, limitations, local difference, and buyer value against exact candidate bytes.
Verify publication
After approval only: deploy, inspect the live route, rerun browser QA, and confirm the public artifact.
Different sources support different claims.
Triangulation improves confidence only when sources are genuinely independent and their coverage is understood.
ClassUseBoundary
Official recordsCounts, filings, licenses, rules
May lag, duplicate, or describe paperwork rather than outcomes.
First-party operating dataWorkflow timing and outcomes
Requires event definitions, privacy controls, and comparable cohorts.
Structured interviewsMechanisms and context
Recall and selection bias prevent unsupported prevalence claims.
Public web evidenceVisible implementation signals
A website flag is not proof of revenue loss or business behavior.
Directional signalsQuestions and vocabulary
Normalized interest is not demand volume, bookings, or causality.
What the current source classes can support.
The boundary travels with the reading. It is not buried in a generic disclaimer.
Source classCurrent readingHard boundary
BLS OEWS20,930 estimated jobs; $28.37 median hourly wage
Employee survey estimate. Excludes self-employed workers and employer burden.
Census BPS66,179 housing units authorized in 2025
Administrative authorizations. Not starts, electrical scope, awards, or revenue.
TDLR license file2,956 DFW-address Electrical Contractor rows
Active licenses and business addresses. Not unique brands or market share.
Workahead website audit105 unique hosts across 10 city lists
Directory-derived automated screen. Not a census or outcome study.
NOAA Storm Events1,105 records in four counties, 2023–partial 2026
Event reports can duplicate episodes and do not measure electrical calls.
ERCOT outlook92,211 MW forecast regional summer peak
ERCOT-wide forecast. Not DFW-specific service-upgrade demand.
Franchisor disclosuresSix-figure investment ranges and 6% fee structures
Official sales disclosures. Not DFW unit economics or profitability.
Google / FrontdoorRanking, auction, lead, and dispatch mechanics
Product mechanics. No DFW booked-job cost or margin.
Numbers the design roadmap does not make true.
These remain collection requirements, not placeholders to fill with model-generated estimates.
42-shop benchmark
No eligible 42-shop dataset exists.
90-day hiring gap / 32% fill rate
Requires dated role-level vacancy cohorts and stable definitions.
Quote speed → close rate
Requires opportunity-level timestamps, comparable jobs, and outcomes.
DFW pricing bands
Requires normalized completed invoices with scope and jurisdiction.
Owner admin hours / $27K bill
Requires time diaries and an explicit value-of-time policy.
Channel cost per booked job
Requires spend, credits, attribution, bookings, collected revenue, and crew capacity.
Franchise market share or margin
Requires a stable local universe and comparable outlet financials.
Client before/after case study
Requires a real engagement, preregistration, consent, and approved records.
Exact bytes are preserved before interpretation begins.
The manifest currently preserves 11 raw files with byte counts and SHA-256 hashes. A hash establishes byte identity—not source correctness, publication rights, or interpretive validity.
TDLR active Electrical Contractor CSV3,787,077 bytes
1da3e401…c10c2fb14,013 rows; 2,956 in frozen DFW frame; 19 blank business counties; no duplicate license numbers.
2025 Census BPS annual workbook294,912 bytes
adb0a2df…105d11c59Final annual CBSA file used for 66,179 authorized units.
NOAA Storm Events detail files2023–partial 2026
Manifested raw filesFiltered to Collin, Dallas, Denton, and Tarrant; rows are not unique storms or jobs.
A label is part of the evidence contract.
Structure only
Illustrative values define the intended page without masquerading as research.
Evidence-backed candidate
Claims and assets are bound to evidence, but public approval has not been granted.
Approved public artifact
Exact bytes were approved, deployed, and checked at the live URL.
Return to the map
See where each research page fits.
The hub connects the market pillar, flagship reports, field notes, quarterly brief, and this methodology.
Back to Research Hub →