Amazon Now Lets You Bid for Two-Hour Delivery. At Most Margins, Breakeven Needs a 9 Percent Lift.

Amazon has begun inviting FBA sellers to bid a per-unit price to get additional products into its Sub Same Day network, the two-hour delivery service that runs out of dedicated fulfillment sites near 2,300 metro areas. EcommerceBytes reported the invitation on August 29, 2026, and Business Insider covered it the following day.

The terms, in Amazon’s own words to sellers: “You pay only for units that actually ship through Sub Same Day, at the per-unit price you bid. Participation is optional, and you’re never charged more than the price per unit you set.”

The recommendation, before the detail: for most sellers the answer is a narrow yes on a handful of SKUs and a firm no on the rest, and the number that decides it is your contribution margin per unit, not the bid. Cap your bid at roughly 8 percent of contribution margin, restrict it to products where delivery speed is the actual purchase trigger, and do not roll it out across the catalog in Q4.

What is actually changing

Amazon already places products into Sub Same Day on its own, selected by what it describes as customer demand and supply signals. That placement is free and Amazon has said it continues unchanged at no cost.

What is new is a second, paid lane. You nominate additional products and attach a per-unit bid. You are charged that amount on every unit that ships through Sub Same Day, and nothing on units that ship normally.

Structurally this is better than most FBA fee changes. It is pay-for-performance with a seller-set ceiling, not a flat charge applied to everyone. Give Amazon credit for the mechanism. The risk sits somewhere else.

The 12 percent figure is not your expected lift

Amazon’s seller-facing help page states that products placed in the Sub Same Day network have seen 12 percent higher sales than standard FBA delivery in areas where the program operates. That is Amazon’s own figure, not an independently audited one, and it is the number most sellers will anchor their bid to.

It is the wrong anchor, for a reason that has nothing to do with whether Amazon is being honest.

Amazon selects the free Sub Same Day placements using demand and supply signals. That means the products in the network today were chosen because they already had high velocity and dense local demand. Comparing that cohort to everything else and attributing the gap to delivery speed is a selection problem, not a measurement of speed.

The products you would be adding are, by definition, the ones Amazon’s own signals did not pick. Your prior on their lift should be lower than 12 percent, not equal to it. You are being invited to buy the treatment for the group the model already screened out.

That average also hides enormous spread. A phone charger, a furnace filter and a bottle of pet medication carry real speed sensitivity. A $180 jacket does not. Nobody comparison shops for four days and then needs it in two hours.

The breakeven formula

You pay the bid on every unit that ships Sub Same Day, including the units that would have sold anyway. You only earn margin on the incremental ones. That asymmetry is the whole decision.

The math works out to a clean rule:

Breakeven lift = bid divided by (contribution margin per unit minus bid)

Take a $34.99 item that nets you $9.00 per unit after referral fee, FBA fulfillment and landed COGS. Here is what different bids require:

• Bid $0.25 per unit, and you need a 2.9 percent lift to break even.
• Bid $0.50, and you need 5.9 percent.
• Bid $0.75, and you need 9.1 percent.
• Bid $1.00, and you need 12.5 percent.
• Bid $1.50, and you need 20.0 percent.

Read the one dollar line again. At a one dollar bid on a nine dollar margin, you need a 12.5 percent lift to break even, which is more than Amazon’s own marketing figure. You would be betting on beating the number Amazon uses to sell the program, on the SKUs Amazon declined to select for free.

Work it in units if that lands better. Say 1,000 of your units ship Sub Same Day in a month at a $0.75 bid. You pay $750. If the lift really is 12 percent, about 107 of those 1,000 units are incremental and the rest would have sold regardless. That is $7.00 spent per incremental unit against a $9.00 margin. You clear $1.99 a unit, and only because the 12 percent held.

Now assume the true lift on your unselected SKUs is 6 percent. Incremental units drop to roughly 57. Cost per incremental unit becomes $13.25 against the same $9.00 margin. You are now losing money on every genuinely new sale while also paying a toll on 943 units that never needed the help.

Halving the lift did not halve the return. It inverted it.

The inventory cost is bigger than the bid

Sub Same Day sites are small. Business Insider reported that they stock on the order of 100,000 products against the millions held at a traditional fulfillment center. Treat that figure as reported rather than confirmed, but the direction is what matters: to be available across more metros you have to position units in more nodes.

That has three consequences your bid spreadsheet will not show.

You buy more inventory to cover the same demand. Splitting across more, smaller nodes means more safety stock in total, because each node needs its own buffer. Same sales, more capital tied up.

You are doing this at the top of the storage cost curve. Amazon’s peak monthly storage rates run October through December. Committing to a wider placement footprint in September means the inventory build lands squarely inside the most expensive storage window of the year. Pull your own rate card and price the delta before you nominate SKUs.

Whatever does not sell ages in January. More nodes means more opportunities for one location to sit long while another stocks out. Uneven sell-through across nodes is how sellers walk into aged inventory surcharges on a SKU that looked healthy in aggregate.

The bid is a few cents a unit. The inventory decision is thousands of dollars of working capital placed at peak rates. Sellers will spend an afternoon agonizing over the bid and thirty seconds on the placement footprint. That is backwards.

The part worth watching

Amazon’s language is careful. Free placement “continues unchanged, at no cost to you.” That is a description of today. It is not a commitment about next year.

We have seen this shape before on Amazon. A surface starts as free and algorithmically allocated, an optional paid enhancement appears alongside it, the paid lane fills, and the free allocation quietly thins because the inventory is finite and now has a price attached. Sponsored Products was optional once too.

That is pattern recognition, not a prediction, and Amazon has made no such statement. But it is the reason to treat your current free placements as an asset to monitor rather than a permanent entitlement. Note which of your ASINs are in Sub Same Day for free today, and check that list quarterly. If it starts shrinking as paid adoption grows, you will want to have noticed early rather than discovering it in a Q4 sales report.

What to actually do

Do not bid on anything Amazon already placed for free. You are not being asked to pay for those and a bid does not protect them. Start by identifying that list so you know what you already have.

Restrict candidates to speed-driven demand. Consumables and replenishment, urgent replacement items, low-consideration purchases under roughly $40, and anything whose real competitor is a shopper driving to a physical store. If a customer would reasonably spend two days deciding, delivery speed is not why they buy.

Set the bid from margin dollars, not a percentage of price. Cap it near 8 percent of contribution margin per unit, which keeps your breakeven lift under about 9 percent. That leaves a little room inside Amazon’s own claimed figure. Above that you are betting against the house line.

Build your own incrementality test, because Amazon will not give you one. Take matched SKU pairs of similar velocity and price. Bid on one of each pair, hold the other out. Run four weeks. Compare units and, more importantly, compare contribution margin dollars, not revenue. Revenue will look fine even when the program is costing you money.

Do not run the test in peak. Demand moves on its own in November and December and you will not be able to separate the program’s effect from the season. Either test now, in the first half of September, or wait until January.

Price the storage delta before you nominate. Model the additional units you will need in the network, multiply by peak monthly storage for October through December, and add that to the bid cost. If the combined number does not clear on a 6 percent lift assumption rather than 12, do not do it this quarter.

The honest summary

This is a reasonable mechanism attached to an unreliable number, landing at the worst possible moment in the storage calendar. The pay-only-for-what-ships structure is fair. The 12 percent lift figure is Amazon’s own, measured on a self-selected cohort, and is close to useless as a planning input for the SKUs you would be adding. And the real cost is not the bid at all, it is the working capital you commit to a wider network footprint five weeks before peak rates.

A narrow, tested yes on the right eight or ten SKUs is a defensible move. A catalog-wide opt-in because 12 percent sounded good is how sellers find out in February that they paid a toll on units they already owned.

Sizing the bid, picking the SKUs and building a holdout test that actually measures incremental margin is the kind of work we run for the brands we operate. If you want a read on whether this is worth it for your catalog before peak, book a strategy assessment and we will go through the numbers with you.