Most security agencies have a rough, informal sense of which guards are reliable and which ones cause problems โ it lives in a supervisor's head, or in scattered notes. Guard Score turns that informal sense into a single number per guard per month, calculated the same way for every guard, so performance across dozens or hundreds of guards can actually be compared at a glance.
Why a 300-900 range
The range mirrors India's CIBIL credit score deliberately โ it's a scale agency owners, HR managers and even clients already understand intuitively. A score in the 750+ range reads as "Excellent" the same instinctive way a high CIBIL score does; a score down in the "Needs Improvement" band signals a guard who needs attention, without anyone needing to learn a new grading system.
The five inputs and their weights
Guard Score combines five sub-scores into one weighted total:
- Attendance โ 40%. The single biggest factor, and the only one that's fully automatic โ computed directly from the guard's actual punch records for the month, combining their presence rate with how consistently their punches fell inside the site's geofence.
- Behaviour โ 20%. A supervisor's rating of conduct on site โ professionalism, cooperation, incident-free performance.
- Uniform compliance โ 15%. Whether the guard consistently reports for duty in proper uniform.
- Work quality โ 15%. A supervisor's assessment of how well the guard actually performs their duties.
- Grooming โ 10%. Personal presentation standards.
The four rating-based factors default to a neutral 70 until a supervisor actually rates the guard, so a newly onboarded guard doesn't start with an artificially low or high score before anyone has observed them.
Why attendance is computed, not rated
Letting a supervisor "rate" attendance would just reintroduce the subjectivity Guard Score exists to remove. Instead, the attendance sub-score is derived mechanically: mostly from what fraction of scheduled days the guard was actually present, with a smaller weighting for how often those punches were geofence-verified rather than flagged. It's the one input in the whole score nobody can can accidentally (or deliberately) rate too generously.
The AWOL penalty
An unauthorized absence โ a scheduled day with no attendance record and no approved leave covering it โ isn't just averaged into the attendance sub-score quietly. It applies a flat 20% penalty to the whole month's Guard Score on top of that, and separately triggers a โน100/day payroll penalty. The intent is that AWOL should visibly hurt the score, not get diluted into a barely-noticeable dip across a month of otherwise-fine attendance.
Supervisor scores work the same way
Supervisors themselves get scored too, using the same underlying engine โ but since a supervisor doesn't punch attendance the way a guard does, their score is ratings-only (behaviour, work quality, etc., entered by HR or admin), with no attendance component and no AWOL exposure.
Scores update automatically, every month
A guard's score for the current month recomputes automatically as attendance accumulates through the month, and a scheduled daily job keeps every active guard's current-month score up to date without anyone needing to trigger it manually. Editing a rating recomputes that guard's score instantly rather than waiting for the next scheduled run.
Conclusion
Guard Score doesn't replace a supervisor's judgment โ behaviour, uniform, grooming and work quality are still human calls, entered by the person actually on site. What it adds is a consistent structure around those calls, combined with hard attendance data, so "which of my guards are performing well" stops being a matter of memory and becomes something visible in one number, month over month.