Last month I watched the same mess happen again: an off-Amazon order blew up, the product was sitting in the warehouse, and customer service still told the buyer to wait five days. We were not out of stock. We had just never separated the fulfillment lines.

Here is the blunt version. FBA exists for orders placed on Amazon. MCF lets you use that same FBA inventory to ship orders from your own store and other channels. One shared pool sounds efficient until peak season arrives, when your best-selling channel eats the stock first and the other channel takes the blame.

How I split MCF vs FBA now: orders first, inventory second

I learned this the hard way in Shenzhen in 2021, selling home storage. Amazon was about 70% of sales and my standalone store was just picking up, so I pushed everything into FBA and drew from one pool. During Black Friday week in November 2021, an influencer video sent more than 900 orders to my store in three days. Amazon orders drained the pool first, and over 400 store orders shipped six days late. Refunds hit 11%. That week ended the one-pool experiment.

Today I run two pools. The Amazon pool carries about 45 days of stock based on the last 30 days of daily sales. The off-Amazon pool carries about 30 days. There is a gate between them: if the off-Amazon pool drops below 10 days of cover, it may borrow from the Amazon pool, but the loan is logged the same day. The reverse is not allowed. Unlogged borrowing is how cycle counts stop reconciling, every single time.

One packaging detail matters more than people expect. MCF parcels once went out with Amazon packing slips inside. A customer who ordered from my own store opened an Amazon slip and assumed they had bought from the wrong shop. We switched everything to neutral packaging — brand name and order number only — and that class of complaint went to zero.

The math: where MCF and FBA actually differ

Do not ask which option is cheaper in the abstract. Run your own orders through three numbers. This table reflects a real month, September 2026, home storage, average selling price around $32, after removing outlier orders:

Item FBA order on Amazon MCF order off Amazon
Fulfillment cost Lower per order, stable speed Higher per unit than an on-Amazon order
Delivery speed Usually 1–2 days Typically 1–2 days slower than on-Amazon
Returns handling Standard Amazon returns flow Returns come back to FBA and need manual re-sorting
Best fit Amazon volume, peak-season orders Own store, crowdfunding, small wholesale orders

Fee schedules change, so check the latest rates on Amazon's official pages before you budget — do not price this year with last year's schedule. The three numbers I actually track are cost per fulfilled order, on-time ship rate, and hours spent on returned units. In that September 2026 month, across more than 900 off-Amazon orders, MCF shipped on time 94% of the time, versus 97% through a third-party warehouse I also used. MCF cost me three points of reliability but saved a separate warehouse lease and a system integration. At low volume, that trade is worth it. At high volume, it flips.

That flip is the counterintuitive part. MCF is not a permanent answer for a standalone store. Once off-Amazon volume holds above roughly 2,000 orders a month and the average order can absorb a third-party warehouse's fixed cost, your own warehouse or a 3PL usually wins. MCF shines at the start and in volatile periods: uneven volume, many SKUs, no appetite for running a second warehouse. Stay on it after you have outgrown it, and you are paying a flexibility premium forever.

My own decision line is simple enough to steal. Under 800 off-Amazon orders a month, use MCF and do not build a second warehouse. Between 800 and 2,000, run MCF as the main line and move oversize and high-return SKUs to a 3PL. Above 2,000 for three straight months, price the 3PL properly: compare MCF's per-order cost against the 3PL's fixed rent plus per-order handling fee. The exact thresholds shift by category, but the direction does not.

There is also a cost nobody puts in the spreadsheet: support time. MCF exceptions — tracking questions, return status, damaged parcels — all route through Amazon's process first, and each exception costs support an extra five to eight minutes on average. At low volume that is invisible. At 1,500 store orders a month, I calculated in late 2025 that MCF exceptions were eating support hours worth roughly 40% of a 3PL's monthly fixed fee. Leave that out and "MCF is cheaper" is a fake number.

A second story makes the SKU point concrete. A friend in Guangzhou selling pet supplies, around 600 store orders a month, enabled MCF for all 120 SKUs in 2024 because someone told him it was hassle-free. More than 30 of those were slow movers that sold a few units in six months, and the twice-yearly cleanup of dead stock nearly broke him. He cut the list to 28 qualifying SKUs, on-time ship rate recovered from 88% to 96%, and support tickets fell by about 30%. More SKUs is not better. The MCF list should only hold SKUs that survive two screens: recent sell-through and return rate.

The pitfalls I actually hit

First, tracking numbers must flow back to your store automatically. For two weeks in 2023 I had no auto-sync, and "has my order shipped?" tickets doubled. Connect the API or use a sync tool. Never copy tracking numbers by hand.

Second, lock inventory 21 days before a major sale. Before Black Friday and Prime events, the borrowing gate closes in one direction: the off-Amazon pool may borrow from Amazon, never the other way. It feels rigid. It is the rule that saves peak season.

Third, do not enable MCF for every SKU. Exclude slow movers, oversize items, and fragile goods. My cutoff is concrete: SKUs with more than 200 units sold in the last 90 days and a return rate under 5% make the MCF list. Everything else goes through the 3PL.

Fourth, reconcile monthly. By the 5th of each month, match the MCF invoice against store orders line by line, watching for duplicate charges, overcharges, and returns that never came back into stock. In March 2026 I found 17 returns that had been signed for but never restocked; the claim took two weeks to resolve. Skip the reconciliation and that money is simply gone.

This approach works because the order is fixed: split the pools, run your own numbers, then enable MCF only for SKUs that earn it. Reverse that order and you are setting up the next peak-season failure. This trick I tested myself, and it holds.

By Ah K, E-commerce Seller (5 yrs)