Amazon FBA vs FBM Cost: Model Your Own Products With AI

Build a per-unit model for each product: selling price minus the Amazon referral fee, fulfilment or postage cost, storage, returns, packing time and product cost. Run it once for FBA and once for FBM, then compare profit per unit and the sales volume each needs to win. An AI assistant builds the sheet; you supply the fees from Amazon.

Generic FBA calculators answer "what would the fees be for this item", which is not the same question as "which method leaves me with more money". The second depends on things only you know: how long your stock sits, what postage you can really get, how much your own time is worth and whether FBA's Prime badge would sell more units.

Follow me on Instagram@sagnikteaches

This tutorial builds that comparison for two products with different shapes, a light fast seller and a bulky slow one. Every fee in the examples is an illustrative input, not Amazon's current rate. Replace them with the figures from Seller Central for your own items, because fee tables change and differ by size, weight and category.

Connect on LinkedInSagnik Bhattacharya

Where the AI helps and where it must not guess

An assistant like ChatGPT or Claude is good at turning a list of costs into a clean spreadsheet, writing the formulas, spotting a formula that double counts, and explaining a result in plain words. It is poor at recalling current Amazon fee tables, which is the place where a confident wrong number does the most harm. So the rule for this model is simple: every fee you type in comes from Amazon's own pages or the calculator in your Seller Central account, and the assistant only handles structure and arithmetic.

Subscribe on YouTube@sagnikteaches

That keeps the model honest and cheap. A $20 monthly chat plan is enough, and the free tier of either assistant can manage a small sheet. The assistant also helps with the less glamorous side of the work: checking that the formulas copy across rows correctly and that the total of costs equals the sum of the lines.

The inputs: one block per product

Create a single input block at the top of the sheet so that a fee change means editing one cell. These are the lines, in the order that works best.

The two rows people leave out are the most revealing: your own packing time and the storage cost of slow stock. An FBM comparison that treats your labour as free will always flatter FBM. An FBA comparison that ignores how long units sit will always flatter FBA.

Ask the AI to build the sheet

Build a per-unit profit model in a spreadsheet for ONE Amazon product.
Layout: an inputs block at the top (blue cells), then two columns,
FBA and FBM, with one line per cost, then profit per unit, margin %,
and monthly profit for a volume I type in.
Lines: selling price, product cost, referral fee (price x rate),
fulfilment fee (FBA), postage and packaging (FBM), inbound per unit (FBA),
storage per unit (cubic feet x monthly rate x months held),
packing labour (minutes x hourly rate), returns cost
(return rate x loss per return), customer service time.
Show every formula in cell references, no typed-in numbers in formulas.
Add a check row that sums the cost lines and compares with the total.
Do not fill in any Amazon fee yourself; leave those cells empty
and list which ones I need to look up in Seller Central.

The last instruction is the one that keeps the model reliable. If the assistant fills in a fee, delete it and replace it with your own source. If you work in Excel, the guide to using ChatGPT with Excel covers pasting formulas back and fixing the ones that do not copy cleanly.

Worked product one: a light, fast seller

Say a silicone baking mat set sells at $24.99 and costs $6.50 landed. The inputs below are illustrative.

FBA leaves $2.91 more per unit. At 150 units a month that is $437 a month, and here FBM does not win unless it sold about 38% more units than FBA, which is unlikely for a $25 kitchen item where the Prime badge matters. For this product the model says FBA, and the answer survives most reasonable changes to the inputs.

To test that, change one input at a time and see whether the winner flips. If postage drops from $4.90 to $3.80, FBM profit becomes $8.66 and FBA still leads by $1.81. If the FBA fulfilment fee rises by 50 cents, FBA's lead shrinks to $2.41. A conclusion that holds through those changes is a conclusion you can act on.

Worked product two: a bulky, slow seller

Now a ceramic planter at $59.00, landed cost $21.00, weighing 3.2 kg in a double box, selling about 20 units a month. Illustrative inputs again.

Here FBM leaves $3.00 more per unit. Storage is the culprit on the FBA side: four months on the shelf at a bulky size eats $2.88 per unit. It is an easy line to underestimate: the fulfilment fee is visible and argued about, while storage on slow stock is a quiet drain.

Break-even, two ways

There are two useful break-even questions, and they answer different worries.

Fixed cost break-even

Suppose fulfilling the planter yourself needs extra shelving, a label printer and shipping software that cost $45 a month and exist only because of this product line. FBM's extra profit per unit is $3.00, so FBM covers that fixed cost at $45 / $3.00 = 15 units a month. At 20 units, FBM earns $60 minus $45, so $15 more a month than FBA. That is a thin win, which tells you the decision is close and the packing workload may not be worth $15.

Sales uplift break-even

The second question is whether FBA's Prime badge would lift sales enough to pay for its higher costs. FBM earns 20 x $13.91, minus $45, which is $233.20 a month. FBA needs to match that at $10.91 per unit, which means 21.4 units, about 7% more than 20. If you believe the Prime badge would add more than 7% to the planter's sales, FBA wins; if not, FBM does. For the baking mats, the same sum runs the other way: at 150 FBM units, FBM profit is $1,134 and FBA matches it at 108 units, so FBA can sell 28% fewer and still tie.

Write the break-even into the sheet as two cells so it recalculates. The formulas are plain: fixed cost divided by the per-unit difference, and FBM monthly profit divided by FBA profit per unit. The assistant can add them in a minute, and you can ask it to explain each in a sentence so you remember what they mean later.

What the assistant should say about the result

Once the sheet works, paste the two result columns back into the chat and ask for a plain-English reading. This is where an assistant earns its place, as long as you tell it to stay inside the numbers.

Here are my FBA and FBM columns for one product, with monthly volume 20.
Explain in under 150 words which method leaves more profit per unit and why.
Name the two cost lines that drive the difference. Then list three inputs
that, if wrong by 20%, would change the winner, with the break-even value
for each. Use only the numbers I gave you.

For the planter, an illustrative reply: "FBM leaves $3.00 more per unit. The gap comes mainly from storage ($2.88 against $0.30) and the fulfilment fee ($12.80 against postage of $9.90 plus packaging). The winner would change if FBM postage rose above about $12.90, if the FBA fulfilment fee fell below about $9.80, or if FBA sold at least 28% more units than FBM. Storage alone cannot flip it: even at zero months held, FBA would trail by 12 cents." The flip points are the useful part: they tell you which inputs to verify first. Postage at $12.90 is a quote you can request this week, so check that before deciding.

Notice what to fix in that reply. The 28% uplift ignores the $45 a month of FBM-only fixed costs from the break-even section, because the prompt never mentioned them; with that cost included, the uplift FBA needs drops to about 7%. An assistant told to use only your numbers will do exactly that, so the prompt has to carry every number that matters. Add a line such as "FBM also carries $45 a month of fixed costs" and ask again.

The fourth-quarter trap and seasonal stock

Amazon charges monthly storage per cubic foot of the space your stock takes up, and its storage fee page lists a peak rate for October to December that, for standard-size items, is roughly three times the January to September rate. Its FBA page also lists an aged-inventory charge for items held in a fulfilment centre for more than 181 days. Rather than copy rates into this tutorial, which change, build the effect into your model through the months-held input. Add a second storage line for stock you expect to hold through the peak months and use Amazon's current storage table for each period. A seller of seasonal gifts may find that a unit bought in August and sold in December is fine, while the same unit still on the shelf in March costs much more than the sheet shows.

A sensible habit is to run each product twice: once with your normal days in stock, and once with double that. If a product only wins with fast turnover, you have learnt that it is a fast-seller strategy, and a slow month should trigger a decision, not hope. Pair that with a rule such as "if units on hand exceed 90 days of sales, stop reordering and consider removal or a price cut".

Ranking a whole catalogue in one pass

When you have 20 products, building 20 models by hand is slow. Instead, build one wide table with a row per product and columns for each input, then have the assistant fill the formula columns. Keep the inputs on the left, the FBA and FBM profit columns in the middle and a "winner" column on the right with a formula such as the larger profit.

I have one row per product with columns: price, landed cost, referral %,
FBA fee, postage, packaging, inbound, cubic feet, months held, storage rate,
packing minutes, hourly rate, return rate, loss per return, monthly units.
Add columns for FBA profit per unit, FBM profit per unit, difference,
winner, and monthly gain from choosing the winner.
Sort by monthly gain, largest first. Do not change my input cells.

Sorting by monthly gain matters more than sorting by percentage. A product that gains $3 per unit at 400 units a month matters far more than one that gains $12 per unit at 5 units. Start by acting on the top three rows and leave the rest.

A short example of mixing the two

An online clothing store that also sells on Amazon might have a light T-shirt that sells 300 a month, a heavy jacket that sells 25 and a bulky gift hamper that sells 6. The model might point to FBA for the T-shirt (small, fast), FBM for the hamper (bulky, slow, fragile) and a close call for the jacket. Holding the jacket in two places, Amazon and your own storeroom, is a legitimate choice if the model says it is close, but it adds a stock-sync job. If that is you, the tutorial on syncing stock across Shopify, Amazon and eBay covers the plumbing. Otherwise keep each product on one method and review it when the numbers move.

Checking the AI-built sheet

A model that looks tidy can still be wrong. Four checks take ten minutes.

1. Hand-calculate one column. Add the lines for the FBA column on a calculator and compare with the sheet's total.
2. Push extremes. Set the price to $0 and confirm profit is negative by the sum of the costs. Set the return rate to 100% and confirm returns cost equals the loss per return.
3. Look for typed numbers inside formulas. A 0.15 hiding in a formula will not change when Amazon's rate does. Ask the assistant to list any constants inside formulas.
4. Compare with Amazon's calculator. Amazon's Revenue Calculator estimates FBA and your own fulfilment side by side for a product. Put the same product into it. The FBA fee total should match yours within a few cents; if not, a line is missing.

How to find your own inputs in a spare hour

Most of the effort in this model is gathering numbers, and none of it needs special software. Work through the list once and note the source next to each figure, so that the next review is a refresh rather than a hunt.

• Landed cost: open your last supplier invoice, add freight and any duty, divide by the units received. If you reorder several times a year, use the average of the last two orders.
• Referral rate and fulfilment fee: enter the product into the revenue calculator in Seller Central, or read the fee preview on the listing. Screenshot it with the date.
• Storage rate and months held: take the rate from Amazon's current storage table and the months from your inventory age report. For FBM, the shelf cost is a share of your rent or shed divided by the units it holds.
• Postage: ask your carrier for the rate for the exact box size and weight, not the average parcel. Add packaging costs from your last order of boxes.
• Return rate: from your returns report over the last 90 days, split into resaleable and not resaleable. Loss per return is the cost of the unit you could not resell plus any return postage.
• Labour minutes: time ten parcels with a phone stopwatch, from picking the item to the label on the box.

Once you have those, the whole exercise is about twenty minutes per product. A first run on your top five sellers will probably tell you more than any generic comparison article, because the answer is specific to your parcel sizes, your carrier rate and how fast your stock moves.

Realistic mistakes and how they show up

• Average storage time guessed. A seller assumes 30 days, actual stock sits 110 days. The FBA column looks great and cash drains. Pull a real inventory-age report and use the average.
• FBM postage priced at retail. The quote used a post office counter price, not the carrier rate the seller actually gets. Use an invoice from last month.
• Returns ignored for fragile items. Breakage on ceramics can double the return cost. Count unsellable returns separately from restockable ones.
• Time at zero. Packing 20 parcels a month at 6 minutes is two hours. If you could use those hours elsewhere, that cost is real.
• One model for the whole catalogue. A size-tier boundary can swing the fee by dollars. Model each product, or at least each size class.

Turning the result into a decision

For each product, record four things on one line: the FBA profit per unit, the FBM profit per unit, the break-even volume or uplift, and a decision with the reason. Then add a review date. A simple rule is that you choose FBA when it wins by more than a dollar and its break-even uplift is below 10%, you choose FBM when it wins by more than a dollar after fixed costs, and you leave the product as it is when the gap is smaller, because switching has its own costs.

Two linked topics help once you have decided. If you pick FBA, restock planning with lead times and limits keeps storage costs down, which is the lever the planter example showed matters most. If you pick FBM, a repricer with a margin floor stops price cuts from eating the $13.91 you worked out.

A cost model is also the base for your accounts. When you later reconcile payouts, the fee lines you modelled should match what Amazon charged. The comparison of Xero and QuickBooks for Amazon sellers shows how the fee lines arrive in each. If you want help setting up the sheet with your own products, that is what a short AI implementation consultation can cover, but the model above is complete enough to build in an afternoon on your own.

FBA versus FBM modelling questions

Should I trust the fee numbers an AI gives me?

No. Take every fee from Amazon's own fee pages and the calculator in Seller Central for your exact product, size and category. Use the AI to build the spreadsheet, check the formulas and explain results, not to recall rates, which change and which it can get wrong.

Does the model include advertising?

Leave it out of the comparison unless you expect it to differ between the two methods. If you run the same ads either way, it shifts both profits by the same amount and clouds the comparison. Add it back when you work out the final profit per unit.

How often should I rerun the comparison?

Every time Amazon updates its fee tables, when your postage rates change, and before a big reorder. A quarterly review is a sensible minimum. Keep the inputs in one block so an update takes ten minutes rather than a rebuild.

What if I want to do both at once?

That is possible, and many sellers keep FBA stock for the fast lines and fulfil slow or bulky ones themselves. Model each product separately, then decide per product rather than for the whole catalogue.

Further reads

• Walmart Marketplace vs Amazon: Which Should a Small Seller Add Next? — Test the same cost sums on a second marketplace.
• Is AI Amazon PPC Software Worth It? Break-Even for Small Sellers — Work out whether AI ad software pays back at your volume.
• AI for Amazon Listing Images: Infographics Inside Amazon's Rules — Lift conversion on the products you decide to keep on FBA.
• Amazon Seller Virtual Assistant or AI: Which Tasks Go to Whom — A task-by-task split for Amazon sellers: what a virtual assistant should do, what AI should draft, what stays with you, and what each costs.
• AI Dropshipping Product Finder: Check Demand and Supplier Risk First — A product shortlist is only a start; use this evidence sheet to reject weak demand, thin margin and unreliable suppliers.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Amazon selling fees page, Fulfillment by Amazon page and Seller Central storage fee help page, Revenue Calculator (read your own figures there), ChatGPT and Claude pricing pages, checked November 2026. All fee inputs in the worked examples are illustrative.