Inventory, Cash Flow & Finance

Part of the AMZ Operator Series (companion to IDS) | Built July 2026 | Distilling: Chelsea Cohen (SoStocked/Carbon6) · Cyndi Thomason (Profit First for Ecommerce Sellers) · Tyler Jefcoat (Seller Accountant) · Kevin King · Yoni Mazor (GETIDA) · Scott Deetz (Northbound Group) · Juozas Kaziukėnas (Marketplace Pulse)

Amazon businesses rarely die of bad products. They die of cash: inventory bought with money that hadn't arrived yet, financed at rates the margin couldn't carry, taxed in April on profit sitting in a container. This is the finance layer under everything else in the series. Figures verified July 2026; fee mechanics live in AMZ 01, cross-referenced not repeated.

---

#1. INVENTORY MATH — FIVE FORMULAS YOU RUN

#1.1 Reorder point

Reorder point (units) = (avg daily sales × lead time in days) + safety stock
Lead time = production + freight + Amazon check-in (2–10 days; 2–4 wks in Q4)

Trigger the PO when available + inbound crosses the reorder point. Cohen's discipline: triggers, not calendar habit — "we order monthly" is how you end up with 9 months of one SKU and zero of another.

#1.2 Safety stock — the z-score version

Simple (demand noise only):   SS = z × σd × √LT
Full (demand + lead-time noise): SS = z × √( LT×σd² + d̄²×σLT² )

z: 1.28 = 90% service · 1.65 = 95% · 2.33 = 99%

Worked example — hero SKU:
d̄ = 40 units/day, σd = 12, LT = 45 days, σLT = 8 days, 95% service
Simple: 1.65 × 12 × √45          = 133 units
Full:   1.65 × √(45×144 + 1600×64) = 1.65 × √108,880 = 544 units
Reorder point = 40×45 + 544 = 2,344 units

Read the gap between 133 and 544: lead-time variability, not demand noise, drives your safety stock. An 8-day freight standard deviation quadruples the buffer — a reliable forwarder (AMZ 04) beats a better forecast. Cut σLT before you cut z.

#1.3 Weeks-of-cover targets by lead time

Post-May-2025 squeeze, Amazon wants FBA holding only what sells in ~30–45 days. Park the rest in AWD or a 3PL and drip-feed.

#1.4 Forecast layering

Forecast(week) = Baseline × Trend × Seasonality × Promo lift
Baseline:    trailing 8-wk avg daily sales, STOCKOUT WEEKS EXCLUDED
Trend:       (last 4-wk avg ÷ prior 4-wk avg), capped at 0.7–1.3
Seasonality: monthly index from your own last-year data or Keepa category curve
Promo lift:  planned deals only — Prime Day/BFCM ≈ 3–8× for the event days,
             then a −20–30% hangover week (pull-forward). Model both.

The #1 forecasting error: baselining off stockout periods — you forecast your inventory's ceiling, not demand. Second: letting one viral week set the trend (hence the cap).

#1.5 Stockout cost model — "we'll just reorder" is a lie

Direct loss    = days out × avg daily units × CM$/unit
Rank decay tax = honeymoon logic in reverse (AMZ 06): velocity signals reset,
                 organic rank slides in days; recovery needs boosted ads
                 ~2–4 weeks per week dark.
True cost      ≈ direct loss × 2–3

Example: 14 days out on the §1.2 SKU
Direct: 14 × 40 × $4.89 = $2,738
True:   $6–8K after relaunch-grade PPC + a month of depressed organic sales.

Cohen's rule: a stockout costs roughly 3× the naive lost-sales number — the number that justifies air-freighting a partial shipment (AMZ 04) and the FBM backup offer on every hero ASIN (AMZ 08 §1.2).

---

#2. THE FBA CAPACITY SYSTEM 2026 — IPI, LIMITS, AND THE BIDDING WAR

#2.1 IPI — the score that sets your ceiling

Four components, updated weekly: excess inventory %, sell-through rate, stranded inventory %, in-stock rate. Threshold: below 400 = storage caps, and since the 2025 tightening restrictions hit immediately, not at quarterly review. 400–550 is the vulnerable band; ≥500 is the operator target; 800+ can unlock effectively unlimited peak storage. Sell-through (90-day units shipped ÷ avg on hand) is the heaviest lever: ≥1.0 minimum, ≥2.0 healthy — i.e. ≤13 and ≤6.5 weeks of FBA cover.

#2.2 Capacity limits — how they're set

Since May 2025 Amazon caps you at roughly 5 months of projected sales (down from 6 — the cut that vaporized 40–75% of many sellers' storage overnight), adjusted by IPI, sales history, and network space. You see a confirmed limit for the current month plus estimates for the next two (the 3-month view) — plan POs against the estimates, because a Q4 estimate that shrinks in October is a container with nowhere to land. Overages incur a fee; the real damage is blocked inbound.

#2.3 Capacity Manager — the bidding mechanics

Need more than your limit? You bid for it:

1. Request extra cubic feet, naming the maximum reservation fee per cu ft you'll pay (no upfront payment).
2. Amazon grants requests highest bid first until space runs out.
3. You earn a $0.15 performance credit per $1 of sales from the added capacity — offsets up to 100% of the fee. Sell through and the space was free; fail and you pay the full bid.
4. Typical bids: $0.10–$0.50/cu ft, materially hotter before Q4 (fees rose 30%+ within months in recent peaks).

Bid math: extra 500 cu ft at $0.35 = $175 at risk.
Fully offset at $175 ÷ 0.15 = $1,167 in sales — if the space holds ~800 units
at $25 ASP, that's 47 units sold.
Rule: bid aggressively for proven velocity; never bid to warehouse a hope.

#2.4 Stranded & excess workflows (weekly, 15 minutes)

• Stranded (in FC, no active offer): Fix Stranded Inventory page, auto-relist rules ON, weekly sweep — stranded units pay storage, earn nothing, bleed IPI.
• Excess (Amazon's flag: >90 days of supply): work the per-SKU recommendation — price cut, deal, remove — or it compounds into the aged ladder.

#2.5 The aged-inventory avoidance calendar

The surcharge ladder starts day 181 and climbs to $7.90/cu ft (full table AMZ 01 §2.4). Run the calendar so no unit ever meets it:

#2.6 AWD as the upstream buffer

The structural fix for capacity limits and aged fees alike: container → AWD ($0.48/cu ft/mo flat, no Q4 spike, no aged surcharge, no placement fee on managed replenishment — AMZ 01 §2.10) → auto-replenishment drips FBA at 30–45 days of cover. Auto-replenished SKUs are LIL-fee exempt, and AWD stock doesn't count against your capacity limit. Trade-off: per-box processing/transport fees and 3–10 days reaction time — keep hero SKUs' FBA buffer at the top of the 4–8 week band. For importers this is the default 2026 architecture.

---

#3. THE CASH CONVERSION CYCLE — WHERE GROWTH EATS CASH

#3.1 The worked timeline (memorize the shape, then compute yours)

Day 0     PO placed, 30% deposit paid            −$9,000  (on $30K PO)
Day 30    Production done, 70% balance paid      −$21,000
Day 30–65 Ocean freight + drayage (35d)          −$4,500 freight/duty at entry
Day 65–72 Amazon check-in (7d)
Day 72–132 Sell-through (60 days at plan)         revenue accrues to Amazon
Day 132+  Final units' payout lands ~14 days later (settlement + DD+7 reserve)
          ≈ DAY 146: the last dollar of this PO's cash is finally home.

Capital locked ~146 days out-to-in. First cash IN ~day 86 (first sales +
settlement) — fully out of pocket nearly 3 months, whole in nearly 5.

CCC = DIO + DSO − DPO
DIO ≈ production + transit + shelf days · DSO ≈ 7–21 on Amazon
DPO = supplier terms (AMZ 04) — every day of net terms is someone else's
money funding growth; 30/70 balance-at-ship beats 50/50-at-order by weeks.

#3.2 Growth eats cash — the table nobody shows beginners

At CCC = 90 days and COGS = 30% of revenue, you permanently hold ~0.9× monthly revenue as pipeline capital. Doubling revenue means funding that amount again — from profit. Months of 100% profit reinvestment per doubling:

At 10% margin and a 120-day cycle, doubling takes a year of every profit dollar reinvested — pay yourself anything and you stall or stock out. This is "profitable and insolvent" in math form: self-funded max growth ≈ margin ÷ pipeline ratio. Grow faster only with §4.2-passing financing.

#3.3 Amazon payout mechanics 2026 (verify in Payments → Disbursements)

• Settlement: 14-day cycles, paid 3–5 business days after close.
• DD+7 (Delivery Date Based Reserve): funds release 7 calendar days after confirmed delivery. Europe migrated September 2025; US/Canada March 12, 2026 — the change that ended legacy zero-reserve and shipment-date accounts. Order-to-bank ≈ 14–27 days FBA, 20–35 FBM.
• Account-level reserves: separate animal — tied to account health, A-to-z volume, or velocity spikes (AMZ 10); can freeze 100% of a payout with no warning, which is why the runway KPI (§6) assumes one skipped settlement.
• Ops: Request transfer mid-cycle once eligible funds exist; reconcile every settlement (§5.2) — deposits are not revenue.

---

#4. FINANCING 2026 — WHAT MONEY COSTS AND WHEN TO TAKE IT

Amazon shut in-house lending March 6, 2024; "Amazon Lending" is now a marketplace of third-party partners inside Seller Central (Parafin, SellersFi, Wayflyer, CrediLinq among them), invite-based on sales data.

#4.2 The two rules that keep debt from killing you

RULE 1 — the spread test. Finance a reorder only if, per inventory turn:
  Contribution margin after ads (%)  >  cost of capital for that turn (%)
  e.g. 6% flat fee, one 90-day turn → 6%/turn (~24%/yr)
       SKU at 22% CM after ads → +16-point spread → finance it
       SKU at 8% → the lender makes more than you → don't

RULE 2 — never finance ACoS-negative growth. If actual ACoS > breakeven ACoS
(AMZ 01 §4) on a mature SKU, borrowing to "scale" it is paying interest to
lose money faster. Debt amplifies the sign of the business.

Jefcoat's screen before any borrowing: a 12-week cash forecast with remittances layered onto DD+7 payouts — one negative week means the advance is too big, whatever the fee.

---

#5. BOOKKEEPING & TAX — THE UNGLAMOROUS MOAT

#5.1 Accrual vs cash — why cash-basis books lie

Cash basis expenses inventory when you buy it: a $60K Q3 container makes Q3 look like a disaster and Q4 like genius — both false. Run accrual for inventory/COGS (or the hybrid: cash opex, accrual inventory). Cash treatment may be allowed for tax under the ~$31M receipts threshold, but management books must recognize COGS at sale or pricing, ads, and financing decisions all run on fiction. Thomason's Profit First adaptation: inventory is not an expense account — fund it from a separate INVENTORY bank account and allocate every payout on receipt (profit %, owner %, tax %, opex %) so tax and profit money is physically gone before it can become stock.

#5.2 The stack (verified July 2026)

Non-negotiable: ledger reconciles to settlements, management P&L reconciles to ledger, monthly. Two sources of truth is zero.

#5.3 COGS discipline

Landed cost per unit per PO (unit + freight + duty + prep), recognized at sale. Update after every PO — tariff moves (AMZ 04) change COGS mid-year, and stale unit costs poison pricing and the §4.2 spread test. Mirror every change into Amazon's Manage Your Sourcing Cost portal — it's now your reimbursement basis (AMZ 01 §5.2).

#5.4 Sales tax after Wayfair — what's actually left on you

• Marketplace facilitator laws (all sales-tax states): Amazon collects and remits on marketplace orders — kills ~95% of the old FBA sales-tax nightmare.
• Still yours: home-state registration; states where you hold non-Amazon nexus; zero/informational returns some states require even when Amazon remitted everything; all DTC/Shopify/wholesale sales — economic nexus (typically $100K/state) is on you (TaxJar/Avalara).
• The FBA inventory-nexus debate: inventory in a state's FC = physical presence, and states keep winning (California's 2025 Diet Standards ruling found nexus on under $2,500 of inventory; CA also chases FBA sellers for its $800 LLC franchise fee, and inventory can create income-tax nexus too). Operator posture: comply fully at home, register where you have real non-marketplace exposure, buy one hour of a state-tax CPA before ignoring the rest — and document the decision.
• 1099-K: OBBBA restored the $20,000 AND 200-transaction threshold from tax year 2025 — irrelevant to what you owe; income is taxable with or without the form.

#5.5 Entity & the S-corp trigger

Single-member LLC from day one (liability wall, clean banking, Schedule C simplicity). Elect S-corp when net profit sustainably clears ~$80–100K: reasonable salary + distributions free of ~15.3% SE tax; costs ~$2–3K/yr in payroll and return prep, so below the threshold it isn't worth it. C-corp almost never (double tax; exception: some foreign-owner structures). Quarterly estimates from month one of profitability — see §9.

#5.6 Inventory write-offs

Dead stock is only deductible when disposed of — liquidated, donated, destroyed — with paper: liquidation remittance, donation receipt, destruction certificate. Write-downs without disposal require accrual method and defensible market value. Every December, force the §2.5 day-150 list through final disposition before Dec 31 so the loss lands in the current tax year.

---

#6. THE WEEKLY FINANCE DASHBOARD — 12 KPIs, EVERY MONDAY

Any KPI out of range two weeks running = a root-cause ticket with an owner and a date (AMZ 08 discipline).

---

#7. REIMBURSEMENTS & FEE AUDIT OPS

Full policy detail in AMZ 01 §5 — this is the operating cadence. The 2024–25 changes made this a monthly job: claim windows are now ~60 days for lost/damaged, 60–120 for returns, 15–75 for removals (down from 18 months), and pre-order losses pay out at your uploaded sourcing cost, not sales price.

Monthly checklist (first week — windows expire):

1. Sourcing costs current in the IDR portal for every ASIN changed this month.
2. Reconcile Inventory Adjustments / Lost & Found; file FC claims on discrepancies >30 days old.
3. Returns: refunds >60 days old with no physical return → claim.
4. Inbound: units received < shipped, with carton-level proof (BOL, packing list).
5. Fee audit per AMZ 01 §6.2 (classification, dimensions, surcharges).
6. Log recovered $ against KPI #8.

Self-file vs services: GETIDA and Seller Investigators (Carbon6) charge 25% of recovery (GETIDA waives first ~$400; enterprise negotiates 10–18%; some newer tools ~18–20%). Hybrid remains the standard (AMZ 01 §5.3): service for breadth, self-audit for inbound claims where a bad auto-filed claim can flag the account. With 60-day windows, an annual cleanup recovers nothing — the backlog no longer exists.

---

#8. SCENARIO PLANNING — THE THREE DRILLS

#8.1 The Q4 overorder trap

Forecast Dec: 3,000 units. Fear buys 4,500 "to be safe."
Reality: Dec sells 2,800; January velocity drops to ~0.7× baseline.
Leftover: 1,700 units ≈ 3+ months of Q1 cover bought with Q4 cash —
  ~$12K locked (at $7.25 landed) · aged clock running toward day 181 ·
  Q1 storage + utilization surcharge · liquidation at −10% margin hands
  December's profit back.
Kevin King's rule: plan to RUN DRY ~Jan 10–15. A Dec 20+ stockout costs
almost nothing (January velocity crashes market-wide, rank decay minimal);
a Q1 liquidation costs real margin. Order for sell-through, not safety.

#8.2 Tariff shock rerun (drill it before it happens — mechanics in AMZ 04)

A +20-point tariff on your HTS code moves the AMZ 01 §3.1 example from 19.6% net to ~14.5% — and the cash hit lands earlier: duty is paid at entry, 60–90 days before the revenue. Within 48h of an announcement: (1) new landed cost per SKU; (2) rerun the §4.2 spread test on financed inventory; (3) per-SKU call — eat it / reprice test / resource (AMZ 04 matrix) / pre-buy one cycle before the effective date; (4) update COGS everywhere (§5.3), IDR portal included.

#8.3 Fee-increase absorption playbook (Amazon re-rates every January)

New fee lands (+$0.50/unit, say):
1. CM% still ≥15%?             → absorb, note it, move on.
2. Re-pack out of the fee?     → AMZ 01 §2.2 band engineering: shave the
                                  half-inch / ounce; SIPP; lighter dunnage.
3. Reprice: +$1 test           → hold if conversion drop <10% and net $ rises;
                                  watch the $10/$50 price-band cliffs (AMZ 01).
4. Channel shift: FBM/3PL math → AMZ 08 §1 table; hybrid offer.
5. None of it works, CM% <10%  → kill the SKU AT THE REORDER POINT (sell
                                  down, don't re-up) — never fire-sale rank
                                  you could monetize while exiting slowly.

---

#9. HOW OPERATORS DIE HERE

---

#MODULE SUMMARY — THE TEN COMMANDMENTS OF AMAZON FINANCE

1. Reorder on triggers, not calendars — demand × lead time + z-score safety stock, and attack lead-time variance before demand variance.
2. A stockout costs 3× the lost sales (rank decay is the multiplier) — but a January stockout is nearly free; plan Q4 buys to run dry mid-January.
3. FBA at 4–8 weeks of cover, network at 8–12, the rest upstream in AWD — inventory placement is fee management.
4. Defend IPI ≥500 like account health — below 400 the capacity cliff is immediate, and Q4 Capacity Manager bids make you a forced buyer.
5. Day 150 is a decision gate, not a suggestion — no unit sees day 181 without a written disposition.
6. Know your CCC to the day — ~146 days PO-to-payout on ocean freight; growth is funded 3–5 months before it pays.
7. Finance only positive unit economics — contribution after ads must beat cost of capital per turn; never borrow against ACoS-negative "growth."
8. Accrual COGS or you're flying blind — A2X/Link My Books into a real ledger, Sellerboard for daily decisions, reconciled monthly.
9. Allocate every payout on arrival — profit, owner, TAX, opex, inventory as separate accounts (Thomason); April only surprises sellers who spent the government's money on stock.
10. Audit and claim monthly — the windows are 60 days now. Recovery is 1–3% of revenue (Mazor); money you don't claim is Amazon's tip.

---

Cross-refs: AMZ 01 (fee stack, LIL/utilization/aged ladders, AWD rates, reimbursement policy) · AMZ 04 (supplier terms, freight variance, tariffs) · AMZ 06 (honeymoon/rank velocity behind the stockout multiplier) · AMZ 08 (FBM backup offer, 3PL economics) · AMZ 10 (account-health reserves) · AMZ 11 (SBA acquisitions, exit-grade books).