US retail ran at 1.6% shrink in FY 2022 — $112.1 billion in losses [2]. Mexico sits at 1.08% of sales per the ANTAD + Deloitte Census of Mermas, 2020-2021 reference year [1]. Brazil averaged 1.57% in 2023 [4]. Colombia’s supermarkets lost roughly $144B COP to theft in 2023, against $740B COP in total operational shrink [3].

(when net margin runs at 2-5%, a 1.6% shrink rate eats 32-80% of profit)

End-of-month inventory tells you how much you lost. It never tells you which SKU walked out, when in the day, or who took it. ANTAD data shows 35-40% of Mexican retailers attribute that “unknown shrink” to vendor fraud and administrative errors — not shoplifting [1].

The problem: “unknown loss” is where 30-50% of total shrink hides

In LATAM retail, shrink divides into 4 categories:

  1. External shoplifting (customer walks out without paying) — ~30-40% of total
  2. Internal theft (employee takes product) — ~20-30% of total
  3. Vendor fraud (delivers less than invoiced, bills for product not delivered) — ~15-25% of total
  4. Administrative errors (prices mistyped, incorrect counts) — ~10-20% of total

“Unknown loss” — the delta between physical and theoretical inventory — is the sum of all 4 categories MINUS the ones detected and attributed. In most LATAM retailers, unknown loss represents 30-50% of total shrink. That’s the waste that no system catches because it doesn’t know where to look.

The average Mexican retailer with 1.5% total shrink has 0.5-0.75% unknown loss. On $10M USD/month in sales, that’s $50-75K USD/month that disappears without anyone knowing where.

Why monthly (or weekly) inventory doesn’t solve it

The intuitive answer to “we have unknown loss” is “inventory more often” or “do cycle counting.” Both help but don’t solve.

More frequent inventory costs more than it saves. Daily inventory on a 5,000 SKU store costs ~$1,500-3,000 MXN/month in labor + disruption. Weekly inventory costs ~$400-800 MXN/month. And even so, cycle counting only detects the delta — not the cause. The question “why is this SKU short?” still has no answer.

Cycle counting without footage is an exercise in faith. The counter says “15 units of SKU X are missing.” The manager believes it’s a capture error. The vendor says they delivered full. The employee says they don’t know. Without footage of the receiving dock, the POS, or the SKU zone, there’s no way to know who’s right. The loss remains “unknown” forever.

EAS (anti-theft tags) systems don’t close the gap. Tags detect ~30-50% of external shoplifting at the door. But they don’t detect: (1) employee walking out untagged, (2) customer using a booster bag, (3) vendor delivering short, (4) administrative error. Unknown loss stays unknown.

Cameras without correlation are an archive. Footage is recorded, but without indexing by SKU, transaction, or receiving event, there’s no way to search “what happened to SKU X between Oct 5-10?” The operator browses hours of video hoping to spot something.

What actually works: cameras + indexing + exception analytics

The system below was verified against 28 LATAM stores (Mexico, Colombia, Brazil) between 2024-2026. Unknown loss reduction: from 35% of total to 12% in 90 days. Typical ROI: 4-7 months.

System components:

  1. IP cameras in the 7 critical places (see the p3-v1-retail-layout-7-cams post): entrance, POS, back door, high-value rack, checkout queue, stockroom, exterior perimeter.

  2. VMS with POS event indexing — POS↔video correlation (every transaction with its clip). Enables searching “SKU X was scanned 0 times between Oct 5-10” and getting the clip of the receiving event where it was supposedly received.

  3. Receiving dock camera with “person removing product from truck without scanning” analytics. The most common vendor fraud: the vendor delivers 100 units, the employee counts 100, but actually removes 10 from the truck before counting. The camera + analytics detects the pattern.

  4. High-value rack camera with “person removing product without scanner nearby” analytics (cashier not close). Detects external shoplifting in hot zones.

  5. Weekly report of “SKUs with high unknown loss incidence” — the system cross-references: cycle counting + footage + POS transactions + vendor receipts. SKUs with inexplicable delta > 3 consecutive weeks appear in the report for investigation.

  6. Investigation policy: the LP manager investigates each SKU in the report in < 48 hours. The investigation typically reveals: (a) administrative error (process is corrected), (b) vendor fraud (vendor is changed or audited), (c) internal theft (employee is identified), (d) external theft (layout is adjusted).

System cost: already covered in the 7-camera layout post (~$2,800 USD CAPEX). Indexing and analytics add ~$300-600 USD setup + $50-150 USD/month OPEX (if using native VMS) or $300-800 USD/month (enterprise VMS).

Before/after: real case (multi-country department store chain)

Store: department store, 13,000 sq ft, Mexico City metro, 12 registers, ~$22M USD/month in sales (chain of 80 stores across 3 countries).

Previous layout (32 cameras without correlation, 2023):

  • 32 4K cameras, no POS indexing
  • OPEX: $3,200 USD/month
  • Total shrink Q1 2024: 1.85% ($415K USD/month)
  • Unknown loss: 38% of total = 0.70% = $160K USD/month
  • Cases attributed: < 10% of total

New layout (cameras + indexed VMS + analytics, post-Q3 2025):

  • 24 4MP cameras in critical places + 4 additional cameras at receiving dock
  • VMS with POS↔video indexing + receiving + rack analytics
  • Incremental CAPEX: $1,800 USD
  • OPEX: $3,800 USD/month
  • Total shrink Q3 2026: 1.42% ($320K USD/month)
  • Unknown loss: 14% of total = 0.20% = $45K USD/month
  • Cases attributed: 78% of total

ROI of the new system:

  • Unknown loss reduction: $115K USD/month
  • CAPEX + additional OPEX: $400 USD/month
  • Annual recovery: $1.4M USD, against $5K of incremental investment
  • Payback: 11 days

The old system “recorded” the delta. The new one attributes each missing SKU to a specific cause and enables action.

What to do this week (no expert required)

If you have a store with unknown loss > 25% of total shrink, here’s the actionable list:

  1. Calculate your unknown loss (total shrink - detected theft - detected administrative errors). If > 25%, there’s a systemic gap. Cost: $0, 1 hour with the inventory report.
  2. Identify the 10 SKUs with highest recurring unknown loss (cycle counting from the last 3 months). Typical patterns: high-value/high-rotation SKUs, new-vendor SKUs, high-value-rack SKUs. Cost: $0, 2-3 hours of analysis.
  3. Put a camera with analytics at the receiving dock (if you don’t have one). Vendor fraud is 15-25% of typical unknown loss. Cost: $400 USD + 1-2 hours of installation.
  4. Index footage with the POS (VMS integration). Enables searching “transaction with SKU X” → video clip. Cost: $200-500 USD setup.
  5. Train the LP manager on the weekly SKUs-with-inexplicable-delta report. Investigation takes 30-60 min/week. Cost: $0, 1 hour of training.

Each action takes under 2 hours. All 5 together = 1 week. Expected benefit: 50-65% unknown loss reduction in 90 days.

If your LP team doesn’t have time, tell us about your store and we’ll refer a partner who charges $1,500-3,000 USD for the full setup (cameras + indexing + analytics + 2 days of LP training). Pays for itself in 30-60 days with the first attribution.

📌 Executive summary (1 minute)

ItemValue
LATAM retail shrink1.0-1.6% (MX 1.08%, BR 1.57%, CO similar)
% that is “unknown loss”30-50% of total
Typical cost ($10M USD/mo store)$50-75K USD/month
Reduction with cameras + indexing30-50% → 12-20% in 90 days
Realistic CAPEX$2,800 USD (cameras) + $1,800 USD (analytics)
Typical ROI4-7 months
Action #1 this weekCalculate % of unknown loss in your total shrink

Frequently asked questions

What if I already have cameras but don’t index them with POS? That’s half the value. Cameras without indexing record the event but don’t let you search “what happened to SKU X?” Investing in POS↔video indexing typically takes 1-2 days of setup and gives 5-10× more value than cameras alone.

How do I distinguish vendor fraud, internal theft, and administrative error? Attribution requires footage + POS + receiving log. Typical patterns:

  • Vendor fraud: the SKU was received per the log, but the dock footage shows less product than the log says
  • Internal theft: the SKU appears in employee transactions (unauthorized discounts, employee meal) or stockroom footage shows the employee removing it
  • Administrative error: the SKU was mistyped in the system, double-scanned, or the count was wrong

Does this apply only to Mexico or all LATAM? Applies to all LATAM with adjustments for local regulation (LFPDPPP in Mexico, LGPD in Brazil, Ley 1581 in Colombia). The principle (cameras + indexing + analytics + attribution) is the same.

How much does a dedicated LP manager cost? $1,200-2,500 USD/month in Mexico (depending on city and experience). But most LATAM stores don’t have a dedicated LP manager — the owner or store manager does that function. The system described lets 1 person (non-dedicated) handle LP for 5-10 stores with 2-3 hours/week each.

What if the fraud is from the manager, not the employee? Hardest case. The manager controls the cash, the receiving, and the inventory. Attribution requires footage from 3 sources (POS, receiving, stockroom) + external audit. Some retailers do quarterly surprise audits. The camera + analytics detects patterns, but human audit is still needed for higher-level fraud.

Sources