A Compliance Safety and Health Officer (CSHO) rarely arrives at a site without a hypothesis. Long before a vehicle turns into the gate, a targeting decision has usually already been made - built from data the facility itself submitted months earlier. Recordkeeping filed under OSHA 1910.119 and submitted through the Injury Tracking Application (ITA) under 29 CFR 1904.41 no longer just satisfies a filing deadline. It becomes an input into where OSHA sends its limited inspectors. A rising incident rate, a spike in DART cases, or a gap between a PSM-covered facility's reported history and its API RP 754 Tier record are signals, not just numbers - and ignoring that shift rarely produces a smaller inspection. It produces a wider one, and often a costlier one.
The Data-Driven Shift Behind Every OSHA Inspection
For most of OSHA's history, inspection selection ran on programmed sweeps, complaints, and referrals. That model has narrowed. Electronic submission under 29 CFR 1904.41 gave OSHA a live, establishment-level dataset to run its Site-Specific Targeting (SST) program against, and National and Regional Emphasis Programs now build inspection lists from RMP data, industry classification, and injury history - not a rotating calendar. A facility no longer needs a visible mistake to get flagged. It can get flagged by what its own Form 300A already says relative to its industry peers.
Why the stakes justify the shift
~2.3 million deaths per year attributed globally to work-related accidents and disease - International Labour Organization (ILO).
A worker injured roughly every 7 seconds in the United States - National Safety Council, Injury Facts.
How OSHA Selects a Site Before Anyone Walks Through the Gate
Establishments with 250 or more employees, and those with 20 to 249 employees in designated high-hazard industries, must electronically submit Form 300A data annually; establishments with 100 or more employees in select high-hazard sectors must submit the full Form 300 log and Form 301 detail. That case-level data gives the SST program something to sort against, not just a rate to rank. National Emphasis Programs layer on top of it - the PSM Covered Chemical Facilities NEP, for example, builds its list from RMP data and injury history, so a facility with a clean recent record can still land on it because its process type and reporting profile match the pattern the program is designed to catch.
The Inspection Trigger Chain
Underneath every data-driven inspection sits the same four-link sequence. Most facilities don't fail at Capture or Submit - forms get filed on time. They fail at Correlate, the point where an inspector's data-driven question (“show me the corrective action tied to this recurring incident code”) meets a CAPA process that was never built to answer it quickly.
1 Capture Incident, near miss, or Tier 1/2 event logged at site level | → | 2 Submit Data filed via ITA, RMP, or state-plan equivalent | → | 3 Correlate OSHA cross-references submission against benchmarks & history | → | 4 Trigger Inspection scheduled, scope partly defined by the correlation |
The chain carries a feedback loop: findings from one visit - an overdue CAPA, an RCA that stops at “human error,” an MOC never formally closed - feed directly into the facility's profile for the next cycle. A weak link doesn't distort one inspection. It compounds into every inspection after it.
This is the exact failure point SOAPBOX.CLOUD™ was built to close.
Why Documentation Decides the Outcome
An inspection built on data doesn't stay confined to the numbers for long. Once the initial questions confirm the pattern the data suggested, the inspection shifts to documentation - and documentation is where CAPA and RCA discipline either holds up or doesn't.
Scenario: A refinery turnaround is running SIMOPS across three units. An MOC record for a process modification was approved on paper but never formally closed before the change went live. It has nothing to do with the incident the CSHO was sent to investigate - but once the inspector is inside the PSM program under 1910.119(l), an open MOC tied to a live process change is its own finding.
Scenario: A CSHO requests every CAPA linked to contractor-reported incidents in the last twelve months, and asks not just whether each is marked closed, but what evidence proves it was verified as effective. A CAPA closed on a signature alone, with no linked evidence, reads exactly the way it should: unverified.
What Mature Organizations Do Differently Before the CSHO Arrives
ISO 45001 Clause 9.1 calls for monitoring and evaluating system performance; Clause 10.2 calls for corrective action that is evaluated for effectiveness, not simply logged as complete. Organizations that treat those clauses as daily operating discipline - an RCA that names a defensible root cause, a CAPA that carries its verification evidence with it, an MOC log where “approved” and “closed” are tracked separately - can answer a data-driven inspector's questions in the time it takes to ask them. None of this is a response to the inspection. It's the operating condition the inspection is testing for.
The One Question Spreadsheets Cannot Answer
When a CSHO is in the lobby, the question that decides the rest of the day is simple: can this facility produce every CAPA, RCA, and MOC record tied to the flagged incident - with verification evidence - before the first question is finished?
Can you produce every CAPA, RCA, and MOC record tied to a flagged incident type - with verification evidence - before the CSHO finishes the first question?
A spreadsheet can hold that information. It cannot connect it on demand, because answering requires the records to already be linked - incident to RCA, RCA to CAPA, CAPA to verification evidence - before anyone asks. Every inspection that runs long, every scope that widens from one unit to an entire site, traces back to that same unanswered question.
