Contractor Safety Rating: Unified Criteria for On-Site Decisions and Procurement

Managing contractor safety at large industrial enterprises often runs into three barriers: fragmented data, delayed response, and the detachment of operational personnel. Field evaluations, incident statistics, and audit results are tracked in separate silos, while line managers view contractor oversight as the sole responsibility of the occupational safety department. At Norilsk Nickel, a reassessment of this approach began after analyzing more than 30 accidents involving contractor employees over a single year: data on precursors existed, but failed to come together as a unified risk signal.

Viktoriya Perestyuk, Senior Methodologist, Predictive Analytics at Norilsk Nickel, examines in detail the design and pilot results of the contractor safety rating—a model that embeds safety assessment into the entire counterparty lifecycle: from prequalification and site access to quarterly monitoring and subsequent procurement.

Rating Architecture: Three Data Sources and 12 Criteria

The model consolidates three data streams into a single scale from −100 to +100 points, recalculated quarterly. First, section managers, their deputies, and chief engineers conduct on-site evaluations of work supervisors once every one to two weeks using a four-block checklist (taking 30 minutes each for the inspection and documentation). Second, events are recorded continuously: violations of cardinal safety rules, work stoppages, accidents, and self-reported risks. Third, once a quarter, 15 HSE specialists conduct a seven-block audit to verify the contractor's ability to systematically manage safety.

The first pilot at the Nadezhda Metallurgical Plant used 24 criteria, but testing revealed them to be redundant and subjective. In the updated version deployed at the Oktyabrsky Mine and other sites, 12 objective indicators (6 positive and 6 negative) were retained:

• Rewarding transparency (+20 and +10 points): 20 points are awarded for each violation the contractor identifies independently among its own workers or subcontractors, for having no poor evaluations of work supervisors, and for voluntarily rectifying findings on time. Customer-approved best practices and operating without stoppages earn 10 points each.
• Cost of concealment and violations (−10 to −30 points): A severe or fatal accident deducts 30 points for 6 months. Concealing an injury or failing an audit costs −20 points each, while violating cardinal rules resulting in a work stoppage, a poor evaluation of a site foreman, or an unsettled fine deducts 10 points each.

At the end of the quarter, the contractor falls into one of four zones: high-risk (under 40 points), acceptable (40–69), reliable (70–89), or the Safety Leader category (90 and above).

On-Site Management Cycle, Flexible Fines, and Procurement

To prevent the scores from remaining a mere formality on a spreadsheet, a five-step cycle has been established: observation, rating, weekly review meetings with the site director and chief engineer, corrective action, and follow-up control. A three-tier escalation applies to work supervisors: the first poor evaluation leads to a one-on-one discussion between the section manager and the foreman, the second to a review with the chief engineers, and the third to the work supervisor's removal from the site.

Claims management has also changed. Initially, 100% fines are processed for all violations through the legal department. However, every month a committee led by five vice presidents reviews appeals: taking into account the rating position and the speed of rectifying findings, between 20% and 100% of the fine is waived for the contractor (around 6 million rubles in fines have already been forgiven for partners), while Safety Leaders are granted one to two months of self-monitoring without inspections. In procurement, falling into the red zone becomes a disqualifying factor even if the contractor offers the lowest bid in a tender.

Pilot Results and Inspector Verification

Preparing to scale up from 51 contractors (62 contracts) to 385 organizations (623 contracts) tested both the counterparties and the client. On the one hand, feedback proved the model's effectiveness: after scoring 0 points on the first audit and 26 on the second, one contractor requested assistance, hired an occupational safety specialist, and proactively initiated a third audit, scoring 87 points. At the same time, the speaker separates operational effects from statistics: four injury-free months at the pilot site do not yet prove a direct impact of the rating on LTIFR—this correlation will need to be confirmed over a longer sample.

On the other hand, the labor intensity of manual metric collection was eliminated by developing proprietary AI-powered IT software, while box-ticking by inspectors was addressed through a "mystery shopper" mechanism. If an inspector gives a contractor around 10 satisfactory evaluations in a row without real improvements on the section, a verification is triggered: if the data is not confirmed, the employee faces disciplinary action, whereas three confirmed inspections in a row lead to a reward.

What You Will Learn from This Presentation

• How to reduce the number of contractor rating criteria from 24 to 12 while eliminating subjectivity during on-site inspections?
• Why awarding +20 points for self-reporting violations and −20 points for concealment changes contractors' attitudes toward risk reporting?
• How the three-step removal of work supervisors and the mechanism for waiving up to 100% of fines through the contractor committee work?
• How the "mystery shopper" verification for internal inspectors is structured and when a re-inspection of their checklists is triggered?
• What IT and organizational challenges must be solved when scaling the assessment from a single plant to hundreds of contracts?