Is selling on Amazon less profitable than selling direct?

Per unit, almost always yes — marketplace referral fees, fulfillment fees, and storage typically consume a larger share of revenue than the equivalent DTC costs of payment processing, shipping, and pick-and-pack. But per-unit margin is the wrong comparison, because DTC units also carry customer acquisition cost that marketplace units largely do not. Once you load paid media into the DTC side, the two channels often land much closer than the headline fee percentages suggest, and in some categories the marketplace unit is the more profitable one.

How do you account for cannibalization when launching on a marketplace?

Assume some share of marketplace sales are customers who would have bought on your own site anyway. Those units are not incremental revenue — they are existing revenue transferred to a lower-margin channel, so their true cost is the contribution difference between the two channels. Estimate the cannibalized share by watching whether total blended revenue rises by the amount of the marketplace revenue or by less, and by monitoring branded search and direct traffic to your own site after launch. Model a range rather than a point estimate, and treat a scenario where a meaningful share is cannibalized as the base case, not the pessimistic case.

What costs do brands forget when modeling marketplace economics?

The most commonly omitted costs are marketplace advertising, which is effectively mandatory for visibility and often runs 8 to 15 percent of marketplace revenue; long-term and peak-season storage fees on slow-moving units; higher return rates driven by frictionless marketplace return policies; the cost of separate inventory pools that cannot be shared with direct fulfillment; and the operational overhead of listing management, case handling, and account health monitoring. Individually each looks small. Together they frequently account for the entire difference between a modeled margin and the actual one.

Amazon vs. DTC: Channel Contribution Math

The comparison everyone makes is the wrong one

The marketplace conversation in most DTC companies runs the same way. Someone proposes launching on Amazon. Someone else points out that between the referral fee and fulfillment, roughly a third of revenue disappears before you've paid for the product. The room agrees that margin is better on the owned site, and the conversation ends.

That comparison is missing a line. On your own site you pay for every customer — paid media, agency fees, creative production, the whole acquisition stack. On a marketplace, a meaningful share of demand arrives from search behavior you didn't pay to generate. Comparing gross margin on one channel to gross margin on the other, without loading acquisition onto the DTC side, is comparing a number that includes its customer cost to one that doesn't.

The right unit of comparison is contribution per unit after all channel-specific variable costs, including acquisition. That's the only figure that answers the question you're actually asking: for the next unit of inventory I have, which channel returns more? Everything else is a proxy that will mislead you in at least one direction.

Marketplace fees are visible and acquisition costs are diffuse. That asymmetry, not the underlying economics, is why most brands over-estimate how much better their own channel is.

Two channel P&Ls, worked

Take a $60 retail product with a $16 landed unit cost, and build both sides honestly.

DTC channel, per unit sold:

Revenue $60.00
Landed COGS −$16.00
Payment processing (~2.9% + $0.30) −$2.04
Outbound shipping, net of what the customer pays −$6.50
Pick, pack, and packaging −$3.00
Returns and refund cost at ~8% −$3.20
Customer acquisition cost, blended, allocated per order −$18.00

Contribution: $11.26 per unit, or 18.8% of revenue.

Marketplace channel, per unit sold:

Revenue $60.00
Landed COGS −$16.00
Referral fee at 15% −$9.00
Fulfillment fee −$6.50
Storage and inventory placement −$0.80
Returns and disposal at ~12% (marketplace return rates run higher) −$4.80
Marketplace advertising at ~11% of revenue −$6.60

Contribution: $16.30 per unit, or 27.2% of revenue.

The channel with visibly worse "margin" produces about 45% more contribution per unit. Not because marketplaces are generous, but because the $18 of acquisition cost on the DTC side is larger than the fee stack on the marketplace side — and it was invisible in the version of this comparison that everyone runs in their head.

Change the inputs and the answer flips. At a $9 blended CAC the DTC unit wins decisively. At a 20% referral fee category with heavier fulfillment, marketplace collapses. The point isn't that marketplaces win; it's that the answer is entirely input-dependent and cannot be reasoned to from fee percentages. Build the table for your own numbers, using the same contribution margin discipline you'd apply to any SKU decision.

Build the channel P&L side by side — free, no login

The Ecommerce P&L Calculator models order economics in real time. Run it twice — once with marketplace fees, once with your blended CAC — and compare contribution per unit rather than headline margin.

Open the P&L Calculator →

The fees that don't show up in the fee schedule

The published fee schedule is the honest part of marketplace economics. The costs that break models are the ones that aren't on it.

Advertising is not optional. Organic marketplace placement for a new listing is effectively zero. Sponsored placement is how you get into the consideration set at all, and the 8–15% of revenue it consumes is a structural cost of the channel, not a growth investment you can switch off once you're established. Brands that model marketplace economics without an ad line are modeling a channel that doesn't exist.

Returns behave differently. Frictionless return policies and a customer relationship that belongs to the platform rather than to you produce return rates several points above DTC in most categories. Worse, returned marketplace units frequently can't be resold at full price, so the cost is the whole unit rather than a restocking expense. That's a compounding effect on the same margin leak that returns and discounts already create on the direct side.

Inventory splits. Marketplace fulfillment usually requires a dedicated inventory pool that can't serve your direct orders. You now hold safety stock twice, forecast twice, and can be stocked out in one channel while sitting on excess in the other. That's a real working-capital cost that never appears in a per-unit fee table.

Operational overhead. Listing management, content compliance, case handling, account health, counterfeit policing. It's rarely a full headcount at first and reliably becomes one — and it's a fixed cost that a per-unit model quietly omits.

Cannibalization is the whole decision

Everything above assumes marketplace units are incremental. Many aren't. A customer who searches your brand name on a marketplace and buys there is a customer who might well have arrived at your own site — and that unit didn't add revenue, it moved revenue from a channel to a different channel.

The economics of a cannibalized unit are entirely different from an incremental one. If DTC contribution is $11.26 and marketplace is $16.30, cannibalization is fine in this example — you're actually better off. But run the same exercise with a $9 CAC, where DTC contribution is $20.26, and every cannibalized unit costs you $4 of contribution. At 40% cannibalization on meaningful volume, that's a channel expansion that grows top-line revenue while shrinking profit, which is exactly the outcome that looks like success on a dashboard for two quarters.

Measure it rather than assume it. Watch whether total blended revenue rises by the full amount of marketplace revenue or by less. Track branded search volume and direct traffic to your own site in the months following launch. Segment new-to-brand rates within the marketplace itself — most platforms report this, and it's the single most useful number available for sizing incrementality. Model a base case where a substantial share is cannibalized, not an optimistic case where none is.

Cash timing runs the other way

One point genuinely in the marketplace's favor: settlement is usually predictable and reasonably fast, and platform-fulfilled inventory often turns faster than direct inventory because demand is more consistent. Faster turns mean less working capital tied up per revenue dollar.

Set against that, you're pre-positioning inventory in someone else's warehouse on their timeline, paying storage on units that haven't sold, and absorbing peak-season storage surcharges precisely when your inventory balance is highest. Whether the channel is net cash-favorable depends on your turn rates in each channel — which is the same working-capital question that governs how fast the whole business can grow, and it deserves the same explicit calculation rather than an assumption.

What you're actually giving up

The strategic cost of marketplace volume is that you don't own the customer. No email address, no purchase history you can segment, no ability to run a retention program against that cohort. Every marketplace order is a first order forever, from your CRM's point of view.

You can price this. If a DTC customer produces, say, $40 of downstream contribution from repeat purchases over 18 months, and a marketplace customer produces effectively none that you can influence, then the per-unit comparison above understates DTC by $40 times your repeat rate. In the worked example, a 30% repeat rate adds roughly $12 of expected downstream value to the DTC unit — enough to close most of the gap. In a category with genuinely low repeat rates, it adds almost nothing and the marketplace answer stands.

That's the honest way to hold this argument: not as a principle about owning the customer, but as a number that either changes the decision or doesn't. In high-repeat categories it usually does. In one-and-done categories, "we need to own the relationship" is a preference, not a P&L argument.

A decision framework

Build both channel P&Ls fully loaded, with acquisition cost on the DTC side and advertising, storage, and elevated returns on the marketplace side.
Add downstream repeat value to the DTC unit, sized off your actual repeat rate — not an aspirational one.
Estimate cannibalization as a range and compute contribution at the pessimistic end. If the plan only works at zero cannibalization, it doesn't work.
Check the inventory constraint. If you're supply-limited, the channels compete for the same units and the decision is purely which one returns more per unit — cannibalization is irrelevant because you'd have sold out anyway.
Price the overhead. Assume the operational load becomes a real role within a year and include it.
Set a review date with a defined kill criterion before launch, while you're still capable of being objective about it.

Done properly, this is a two-hour exercise in a spreadsheet, and it replaces a debate that most teams have been having on instinct for years.

Model your own channel mix

The failure mode we see most often isn't choosing the wrong channel. It's choosing on a per-unit margin comparison that never loaded acquisition cost onto the direct side, then discovering eighteen months in that revenue grew, contribution didn't, and a third of the new volume was customers who were already yours.

Want a channel P&L that accounts for cannibalization and downstream value?

We build fully loaded channel contribution models for DTC brands weighing marketplace, wholesale, and direct — including incrementality testing and the repeat-value math that decides most of these calls. Starts with a free diagnostic call.

Request a free audit →