How to catch gas station losses before month-end.

Why most station losses get caught at month-end

The default review cadence at an independent gas station is monthly. The bookkeeper closes the month, the owner gets a P&L, and any variance shows up as a category-level number, fuel margin down 1.2¢, inventory shrink 2.3%, cash over/short $-184 for the month.

The numbers are correct. They're also useless for investigation. The 1.2¢ fuel margin gap could be from any of twenty days; the $184 cash variance could be from any of nine shifts; the 2.3% inventory shrink could be three SKUs or thirty. By the time the owner sees the aggregate, the shift schedule has rolled three times and attribution is gone.

Daily vs. month-end loss detection

Both cadences exist, but they answer different questions.

The three cadences are complementary, not substitutes. Skipping daily review because "we close the month" is the single biggest reason operators discover losses two months late.

The six loss categories

1. Cash variance

The drawer-level gap between expected cash (from POS sales, drops, refunds) and counted cash. Attribution window: hours. Detection: every shift close. Real-time alert: SMS to owner on over-tolerance variance.

Median independent runs $20–$60/week of unattributed cash variance per store. Most resolves as process error; the residual is the actionable signal. More on cash variance.

2. Lottery shrink

Missing scratcher serials, settlement mismatches, pack-activation errors. Attribution window: days. Detection: daily reconciliation against state-commission report. Pattern signal: same clerk, repeated gaps.

Median independent runs $50–$150/week of lottery shrink per store. Serial-level tracking surfaces it; pack-level tracking doesn't. More on lottery shrink causes.

3. Fuel margin compression

Realized margin per grade vs. target. Attribution window: weeks (because the cause, pump miscalibration, wholesale tracking, wet-stock, itself compounds over weeks). Detection: daily wet-stock reconciliation + weekly realized-vs-target review.

A 1¢/gal margin gap across 30,000 gal/month = $300/month. A sustained 3¢ gap = $900. Catching the drift before it compounds is what daily ATG-vs-dispensed reconciliation does. More on fuel margin.

4. Inventory shrink

SKU-level gap between expected on-hand (receipts in, sales out) and actual counted. Attribution window: weeks. Detection: rolling mobile cycle counts (30–50 SKUs/day) plus invoice-OCR ingestion for receiving accuracy.

NACS industry average c-store shrink: 1.5–2.5% of gross sales. Tobacco is the highest-dollar category at most c-stores. More on inventory shrink.

5. Expense drift

Recurring expenses (utilities, repairs, supplies, vendor bills) creeping up without anyone noticing. Attribution window: weeks to months. Detection: weekly category-level expense review with prior-month comparison.

A $50/month water-bill increase that nobody catches until year-end is $600 of unrecovered exposure. Detection is the same whether the cause is rate increase, leak, or vendor error. More on expense tracking.

6. Deposit mismatch

Bank deposit credits short against expected (from prior closeout cash). Attribution window: days (deposit lag + reconciliation latency). Detection: bank-feed-driven reconciliation via Plaid or similar.

The leakiest category for operators who don't reconcile daily. A missed deposit on the 1st of the month doesn't surface until the bank statement on the 28th, by then, attribution is gone. More on deposit reconciliation.

The daily review checklist

Ten minutes when nothing is anomalous. The owner runs this in the morning brief; the manager runs the same routine before opening.

• Last night's EOD variance per store, with reason codes.
• Lottery exceptions, missing serials, settlement mismatches.
• Fuel margin band per grade, realized vs. target.
• Cycle-count alerts on flagged SKUs (tobacco prioritized).
• Yesterday's vendor invoice queue, approvals pending.
• Deposit reconciliation status, credits posted, pending, missing.
• Voids and refunds per clerk vs. baseline.
• Open expense items requiring approval.
• Unresolved manager notes from prior shift.
• Cross-signal pattern flags (Loss Radar exceptions).

Anything flagged here should result in a same-day conversation. Not a Friday roundup, not a month-end aggregate.

Example: the same $400 leak, three detection cadences

Same operational pattern. Same overnight clerk skimming an average of $20/shift in cash by ringing $13 tobacco packs as $1.30. Pattern runs 20 shifts across the month = $400 of exposure.

Month-end review only

End of month financials:
  Cash variance for the month:    $-420
  Tobacco shrink:                  −3.1% (above 2.5% benchmark)

Action available:
  "Something's wrong with tobacco." Investigation needs to
  span 20 shifts; clerk schedules have rolled; cameras are
  on a 14-day retention so 6+ shifts of evidence is already
  gone. Probably absorbed.

Weekly review

Week 1 review (Monday morning):
  Cash variance for the week:     $-105
  Tobacco shrink:                  −2.8% (slight elevation)

Action available:
  "Pattern is forming." Camera review possible for week 1.
  Conversation with overnight clerks. Pattern recurs week 2,
  caught earlier. Half the exposure prevented, but week 1
  was already gone before the review.

Daily review (with cross-signal Loss Radar)

Day 3 morning brief:
  EOD variance overnight shift:   $-20 (within warning)
  Voids overnight shift:           5 (clerk baseline: 1)
  Tobacco SKU velocity drop:       Marlboro Box -12% (3-day)
  Loss Radar cross-signal flag:    "Repeated overnight variance +
                                    void burst + tobacco velocity
                                    drop. Same clerk all three days."

Action available:
  Camera review of three specific moments (under 5 min each).
  Conversation with clerk before her Tuesday shift.
  Pattern prevented after day 3. Total exposure: $60.

Same source data. Three very different outcomes, driven entirely by detection cadence, not by detection sophistication.

How real-time alerts change behavior

Detection cadence alone isn't enough, the cadence has to result in a conversation. Real-time alerts on out-of-band events produce that conversation before the next shift, which is the threshold for behavioral change.

The four alert types that actually change behavior at independent stations:

• Cash variance > configured warning band. SMS to owner at shift close.
• Missed EOD close. SMS to owner 30 min after configured close time.
• Missing scratcher serial, flagged in the morning brief with shift attribution.
• Cross-signal pattern. Loss Radar flag when two or more signals correlate on the same shift.