Stop shrinkage before it leaves the store

Cashier-level fraud is the most difficult category of retail loss to detect and the most expensive to ignore at scale. A single cashier who voids legitimate transactions, accepts goods without scanning, or applies unauthorised discounts may generate ₹5,000–₹20,000 in loss per shift — small enough to evade individual stock audits but substantial when multiplied across hundreds of counters over a full financial year. The structural gap in traditional LP is that POS exception reports produce data without context. A void on its own is ambiguous: it could be a genuine correction, a training error, or a coordinated fraud mechanism. Only the video of that transaction resolves the ambiguity.

Slinai's POS correlation layer integrates with Oracle Simphony, Posist, NCR Aloha, Petpooja, and custom-developed systems through API and webhook connectors. When a void, no-sale, refund, or high-discount transaction fires in the POS log, Slinai captures the terminal ID and timestamp, correlates the event to the matching billing-counter camera view, clips a 90-second segment centred on the transaction, and routes a combined alert — transaction record and clip — to the LP manager in a single investigation package. No manual timestamp lookup. The LP team sees both the anomalous transaction and the corresponding customer interaction simultaneously and resolves the event in minutes rather than after multi-day investigations.

Over time, the correlation layer builds behavioural baselines per terminal — average void frequency by daypart, discount rates by weekday, refund patterns by category — and surfaces deviations in weekly exception reports ranked by risk score. LP supervisors focus coaching and disciplinary conversations on specific evidenced events rather than suspicion and gut feel. For chains running 80 or more billing counters, this shift to evidence-based cashier management typically yields a 15–25% reduction in exception event rates within the first two financial quarters.

Warehouse and distribution-centre shrinkage in India is harder to detect than retail floor theft — it is more easily masked as administrative error and more rarely attributed to deliberate loss. Shrinkage events cluster around three phases: inward receiving, where short shipments are recorded as full acceptance; storage and picking, where high-value goods are diverted during low-supervision night shifts; and dispatch, where cases are sealed without full contents or riders substitute at exit points. In each phase, the camera evidence exists — but without a system correlating footage to GRN timestamps, WMS pick records, and dispatch manifests, that evidence sits passive in an archive.

Slinai monitors receiving docks with zone-calibrated cameras that capture unloading sequences alongside GRN timestamps — so variance between what arrived on the vehicle and what entered the WMS can be correlated with video evidence rather than attributed to vendor counting error by default. In pick zones, anomaly detection flags unusual dwell near high-value SKU locations during low-traffic periods, capturing the behavioural signature of pilferage before a stock audit reveals the gap. Dispatch gates benefit from automated vehicle and personnel tracking that confirms handoff sequences against WMS-generated order manifests when integrations are available.

For 3PL operators managing multiple warehouse clients from a shared facility, Slinai provides client-separated zone monitoring with access controls — so Client A's LP team cannot view footage of Client B's stock locations while both benefit from the same underlying camera infrastructure and LP event logging platform.

Dark stores — fulfilment hubs designed to deliver groceries and essentials within 10 to 30 minutes — operate at a pace that makes traditional loss prevention methods structurally incompatible with the business model. Pickers move at 800 to 1,200 items per hour. Orders are sealed, dispatched, and in customers' hands before any physical variance can be caught by on-shift supervision. The result is a category of loss that appears in the P&L as customer complaints, platform refunds, and inventory gaps — but cannot be attributed to specific pick events, specific operatives, or specific shift windows without video correlation to the WMS data already captured in the system.

Slinai monitors pick-zone aisles in dark stores with camera coverage calibrated to high-value SKU bays and fast-moving product sections, flagging unusual dwell at premium product locations, items placed into personal containers rather than order totes, and picking patterns that deviate significantly from the expected route for the active order type. When integrated with WMS data, pick-event timestamps correlate with specific order IDs and picker assignments — so a customer complaint about a missing ₹800 item resolves in a 90-second clip review rather than an entire shift's footage scan. The investigation that previously took two days completes in twelve minutes.

For quick-commerce operators running 40 to 200 dark stores across multiple cities, multi-site roll-up dashboards display pick-shrink incidents by city hub, operational time slot, and product category. Shrinkage patterns recurring across multiple hubs in the same SKU category often indicate a systematic receiving gap rather than a picking problem — a distinction that changes the remediation from people management to vendor management, with clip evidence to anchor both conversations.

Running a loss prevention programme across 50, 150, or 500 locations requires a fundamentally different toolset than managing LP at a single flagship store. The challenge is not that individual store incidents are hard to investigate — it is that incident data from 150 stores, generated by different camera hardware, interpreted by different store managers, and tracked in different formats across area-manager spreadsheets, cannot be aggregated into a coherent picture without a purpose-built multi-site platform. Without that aggregation, HQ LP directors are dependent on what area managers choose to escalate — creating systematic blind spots precisely where LP intervention is most needed.

Slinai's multi-site LP dashboard provides a hierarchical view from regional cluster down to individual zone — normalised across camera brands, NVR generations, and site formats. LP managers see incident counts by category — shrinkage anomaly, cashier exception, after-hours event, compliance gap — by region, and by time window. Outlier stores, those showing incident rates two or more standard deviations above their format cluster average, surface automatically, enabling targeted site visits, audit prioritisation, and format-specific control deployments based on data rather than manager intuition. The same dashboard surfaces camera health and zone coverage status per site, so LP managers can identify blind spots before those gaps are exploited.

For LP directors managing franchise networks, the multi-site view supports brand-to-franchisee reporting — sharing anonymised loss rate benchmarks and remediation guidance without exposing one franchisee's footage to another. API exports of incident data and site-level metrics allow LP event flows to join broader risk management frameworks, ERP audit trails, or insurance reporting requirements rather than remaining siloed in the camera system.