eBay Promoted Listings: Set Ad Rates With AI and a Margin Floor

Work out the most each item can afford to pay in ad fees (its margin before advertising, less the profit you insist on keeping), turn that into a percentage of the total sale amount, and never set a cost-per-sale (general strategy) rate above it. Use AI for the arithmetic and test one category at a time.

Most sellers set ad rates by copying eBay's suggestion or by guessing. The suggestion tells you what others are paying, which has nothing to do with whether you would still make money after paying it. A margin floor flips the question: instead of asking what rate wins clicks, you ask what rate you can survive, then test inside that limit.

Follow me on Instagram@sagnikteaches

General and priority: what each strategy charges for

eBay's seller pages now call the two Promoted Listings strategies general and priority. Many sellers, and most third-party guides, still know them as Promoted Listings Standard and Advanced. The general strategy is cost-per-sale: you set an ad rate, a percentage, and you pay only when a promoted item sells after a click on its ad. The priority strategy is cost-per-click: eBay's help page says you will always be charged for valid clicks, whether or not a sale follows, and that general fees still apply even if the ad later received a priority click. eBay describes its model as last-click attribution.

Connect on LinkedInSagnik Bhattacharya

One rule matters more than the names. Since January 2026, eBay's attribution page says a general-strategy sale counts when any buyer, not necessarily the one who bought, clicked the general ad for that item in the previous 30 days. The item must be promoted both at the time of the click and at the time of the sale, and the fee uses the ad rate in force when it sells. On a listing that gets steady clicks, that means nearly every sale of a promoted item carries the ad fee, which is why the margin floor below matters so much.

Subscribe on YouTube@sagnikteaches

eBay's help page also says the ad rate is a percentage of the item's total sale amount, including item price, postage, taxes and any other applicable fees, so postage the buyer pays is part of the fee base. eBay's general campaign page gives the allowed range as 2% to 100% of that total. You can set a fixed rate, which stays where you put it, or a dynamic rate, which eBay updates daily to its suggested rate; for any item whose ceiling sits below the suggestion, a dynamic rate will carry you straight past your floor, so use a fixed one. Where a figure matters below, I use an illustrative number and tell you where to swap in your own.

For a small seller the difference matters like this. With the general strategy, a bad rate costs you margin only on items that sell. With priority, a bad bid costs you money on clicks that never turn into sales, so it needs a budget and watching. Many sellers should master general campaigns first; priority earns its place once you know which items convert and want to bid for position. If you are unsure, start with a general campaign and a small group of items.

The margin floor, step by step

The floor has three ingredients: what an item sells for, what it costs you to deliver that sale, and the profit you refuse to go below. Everything else follows.

1. Selling price. What the buyer pays that eBay counts for fees. Use the figure on your fee statement.
2. Cost of goods. What you paid for the item, including inbound shipping or repair.
3. Delivery and packing cost. What you pay the carrier, net of what the buyer pays you, plus the box and tape.
4. eBay's selling fees. The fees on a sale for your category and account, from your statement. Do not use someone else's percentage.
5. Profit floor. The least you will accept per sale after everything. Pick a money amount or a percentage of the price.
6. Allowance for returns and losses. A small percentage for items that come back or arrive damaged. Use your own history.

The maximum ad spend per sale is selling price minus every cost above, including the floor. Divide that by the sale amount on which the ad fee is charged, and you have the highest rate you can afford. In a spreadsheet:

margin_before_ads = price - cost_of_goods - delivery_cost - ebay_fees - returns_allowance
max_ad_spend      = margin_before_ads - profit_floor
max_ad_rate       = max_ad_spend / fee_base      (fee_base = total sale amount: price, postage, taxes)
set_rate          = min(max_ad_rate, rate_you_want_to_test)

If max_ad_rate comes out at zero or below, the item cannot be promoted at a profit. That is a finding, not a failure; it tells you to price it differently or leave it alone.

The main worked example

All figures are illustrative; swap in your own. Say an accountancy practice's office-clear-out side business sells refurbished label printers, and one model sells for $60 including postage paid by the buyer.

If the order also carries sales tax, that tax is part of the fee base too, so the real ceiling is slightly lower; use the totals on your own sales report. Ignoring tax, 15.7% is the ceiling for this printer, and you would never go above 15%, perhaps starting at 8% and moving up. If eBay's suggested rate for the category were 12%, you would know you could match it and still clear $10. If the suggestion were 18%, you could not follow it without dropping below your floor, and the right move is not to bid for position on this item.

A second item, a bundle of ten spare cartridges at $18, has a margin before ads of $2.50 and a floor of $2.00. Maximum ad spend is $0.50, which is 2.8% of the price. At that rate you will rarely win a promoted slot. The sensible answer is to leave the cartridges unpromoted and let them sit under the printer listing as organic add-ons.

Let AI do the sums, from a table you control

You do not need a clever prompt for the arithmetic; you need clean input. Export your sold items from Seller Hub for the last three to six months, add a cost column from your own records, and keep a column for the fees you were charged. Remove buyer names and addresses before you paste anything into a chat tool; only item data is needed.

Below is a table of my eBay sold items: item title, category, sale price, my cost, postage cost, eBay fees charged, return flag (Y/N).
For each category, calculate:
1. average margin before ads (price - cost - postage - fees), as a money amount and as a percentage of price
2. return rate
3. the maximum ad rate if I want to keep a profit floor of [X] per sale, with the fee base equal to the total sale amount including postage
Show the working in a table. Use only the numbers in my table. If a category has fewer than 10 sales, mark it "too few sales to trust".

[table]

Illustrative output: a table with printers at 15.7%, cartridges at 2.8%, label rolls at 9.1% and a category flagged "too few sales" because there were only four. Then check two rows by hand with a calculator. Language models do arithmetic well on small tables and badly when a column has odd formats, such as currency symbols mixed with text. If a row looks off, the cause is usually a cell the model misread.

What to fix in what comes back

• A margin that is higher than the sale price: the model mixed up cost and price columns.
• A category with a very high ad ceiling and a handful of sales: small samples. Keep your rate low until more data arrives.
• A uniform percentage for everything: it has averaged across categories. Ask again by item group.

The same table can feed a spreadsheet formula directly, which is more reliable for ongoing use. For help with the spreadsheet side of this, prompts for Google Sheets covers formulas AI can write.

Test by category, one change at a time

A ceiling tells you what is safe, not what works. To find what works, test inside the ceiling, a category at a time, and keep the test clean.

1. Pick one category with enough listings and sales to judge, say 20 or more live items.
2. Split the items into two matched groups by price and age. Promote one group, leave the other alone. This hold-out is the part most sellers skip, and it is how you learn whether ads created sales or merely took credit for sales that would have happened anyway.
3. Choose a starting rate well under the ceiling, such as half of it. Do not change anything else (titles, prices, photos) during the test.
4. Run it for a fixed period. For fast-selling items that may be two weeks; for slow items, four to six. Write the end date down before you begin.
5. Compare, between the two groups, the number of sales, the revenue, and the profit after ad fees.
6. If the promoted group earned more profit after ad fees, raise the rate a step and repeat. If not, lower it or stop.

Illustrative test result for 40 printers and accessories, 20 promoted at 8% and 20 not: promoted group made 11 sales at an average price of $44 and ad fees of $38.70, the control made 7 sales. The extra four sales earned about $19 each in margin before ads ($76), and the ad fees of $38.70 are charged on all 11 sales, not just the extra four. Profit difference: extra margin of $76 less fees of $38.70 gives about $37 better off. A decent result on 40 listings, and one that justifies another round at 10%. Had the control made 10 sales, the same fees would have bought one extra sale for $38.70, and the right answer would be to stop.

The detail to notice in that sum is the line "ad fees are charged on all promoted sales". Some of those sales would have happened without the ad, and you paid for them anyway; under the any-click rule for general campaigns that is the normal case, not an edge case. That is the hidden cost of promotion. It is why the margin floor protects you even when the ad works.

General versus priority for a small seller

For priority, the margin floor still applies but needs one more number: the share of clicks that end in a sale. If an item converts at one sale in 40 clicks and a click costs $0.30, each sale costs $12 in ad fees, which is more than the printer's $9.40 maximum. The same item might be fine on a general campaign at 8%, which costs $4.80 a sale. Do the sum before you open a priority campaign, using your own conversion count rather than a quoted average.

When a promoted listing is a poor idea

• Thin margins. If the ceiling is under 3%, a promoted slot is rarely winnable.
• Items you cannot restock. Selling one-off items at an ad cost is fine; promoting an item you run out of just spends money on interest.
• Weak listings. An ad sends more people to a listing. If the photos, title or description are poor, more traffic produces more failed visits. Fix the description layout and the item specifics before you pay for visits.
• Unreliable stock or shipping. A rush of sales you cannot ship on time will hurt your account, as the seller level audit explains.
• Prices that were never competitive. If similar items sell for less, a higher ad rate does not change the buyer's mind.

A weekly routine that takes twenty minutes

1. Export last week's promoted and non-promoted sales.
2. Paste the table into your sheet and let the formulas refresh margin before ads, ad fees and profit after ads for each item.
3. Ask AI for a one-paragraph summary of which categories are below the floor, using only the table, and read the numbers behind it.
4. Lower the rate on any category where profit after ads is below the floor.
5. Raise the rate by one step only on categories where profit after ads rose in the test period.
6. Log the change and the date.

After a month you will have a table of rates by category and a record of why each was set. That record is worth more than any suggested rate, because it comes from your own sales, your own costs and your own floor. If you sell the same stock on other marketplaces, compare the cost per sale on each; listing across marketplaces covers the differences in rules.

Reading eBay's suggested rate against your ceiling

The campaign screen shows a suggested rate for each item or group. eBay says the suggestion draws on factors such as item attributes, seasonality, past performance and current competition: where many sellers promote, the going rate rises, and a low rate may show your item less often. That is useful information, because it tells you what visibility costs in this market. Put the suggestion beside your ceiling and you get one of three cases:

• Suggestion below your ceiling. You can match it, and the test is about whether a lower rate would do. Start at the suggestion or a little under, and step down only if sales hold.
• Suggestion above your ceiling. You cannot compete for position on this item at a profit. Use a rate at or below your ceiling and accept less visibility, or improve the margin by lowering the cost or raising the price.
• No suggestion. Treat it as a new market for the item. Start low and test.

Say three categories look like this in one illustrative sheet: printers (ceiling 15.7%, suggestion 12%), label rolls (ceiling 9%, suggestion 11%) and cartridges (ceiling 2.8%, suggestion 7%). You would test printers at 6% and 9%, test label rolls at their ceiling of 9% knowing you may lose some position, and skip cartridges. One sheet, three different decisions, none of them taken from the suggestion alone.

Price, rate and visibility trade against each other

Sellers sometimes raise a price to fund a higher ad rate. That can work, or it can sink sales, because a higher price changes the buyer's comparison on every search. Test it the same way: change one thing at a time. If your ceiling is too low, there are four levers and each has a cost.

1. Raise the price. Gives room, but may reduce conversion.
2. Cut the cost of goods. Slow to do, but permanent.
3. Cut delivery cost. Cheaper packaging or a better carrier rate can add a point or two of margin.
4. Lower the profit floor. Only if you have a reason; it is the one lever that costs you directly.

Ask AI to compute the ceiling under each lever, as a table, from the same input you gave before. A simple prompt: "Using the table above, show the maximum ad rate for the printer if the price rises by $3, if postage cost falls by $1, and if the profit floor falls to $8. Show each as a separate row." That puts four numbers on the table instead of one argument, and you pick the lever you can live with.

A test plan written down before the test

Tests drift when the plan lives in your head. Write it in a few lines first, and ask AI to check it for holes without letting it change your numbers.

Here is my test plan for eBay Promoted Listings (general strategy, cost per sale). Look for problems: anything that would make the result unreliable, such as groups that are not comparable, a period that is too short, or a second change made at the same time. Do not suggest new rates and do not invent figures.

Category: [name]
Items promoted: [number], rate [x]%
Items not promoted: [number]
How groups were chosen: [description]
Period: [dates]
Other changes planned in this period: [list or none]
What counts as success: [profit after ads higher than control by at least $X]

A typical illustrative reply flags that the promoted group's items are all newer than the control's, so sales may reflect listing age, and that the period includes a seasonal peak. Both are fair; fix them before you start. The model cannot tell you whether the result will be good, only whether the experiment is fit to be believed.

A small seller walked through four weeks

Say a four-person insurance brokerage runs a side shop clearing refurbished office phones and headsets. Week 1: the owner exports six months of sales, adds cost and postage, and finds headsets have a ceiling of 11% while phones sit at 6%. Week 2: they promote 15 of 30 headsets at 5%. Week 3: they see 9 sales from the promoted group against 5 from the rest, check the profit after ads, and find a gain of about $22. Week 4: they raise the rate to 7% on the same group, and the sales do not rise. They return the rate to 5%, leave the phones alone, and write both findings in the log.

What they did not do is notable: they did not extend the test to phones because headsets worked, did not change prices mid-test, and did not accept the suggestion because it was higher. With a month's effort they had one item group that paid for promotion, one that did not, and a rate that had a reason behind it.

Checking the numbers before you trust them

Ad data can mislead in a few ways. Attribution is generous to the ad: under the general strategy a sale counts if anyone clicked the ad in the previous 30 days, so a buyer who found your item through ordinary search can still produce an ad fee and an ad sale in your report. That inflates ad performance in your sheet. Fees can change with your account or category, so refresh the fee column from your most recent statement. Postage and taxes the buyer pays are part of the fee base, so use the total sale amount, not the item price, in the fee_base column. And tiny samples create convincing patterns: three sales out of eight clicks looks like a 37% conversion until the next week brings none. Keep the "too few sales" rule from the prompt, and wait for a sample before you act on a percentage.

Finally, remember what the margin floor is not. It does not tell you the ad is working, only that you will not lose money if it does not. Whether it works is a question for your hold-out group, your log and a few weeks of patience.

A mistake that eats margin quietly

Picture, as an illustration, a seller who sets a blanket rate of 10% across every item because the suggested rate for most of the store is near that. The cheapest items have ceilings near 4%. For three months the cheap items sell briskly, which looks like success, and monthly profit falls. The cause only becomes visible once the numbers are split by category: the 10% on $15 items costs $1.50 a sale against a margin before ads of $2.20. A small sample of one week's data would have shown it. The fix is one line in the sheet: rate = the lower of the ceiling and the test rate. That single formula is why this tutorial uses a ceiling rather than a target.

Keep the habit: every rate has a number next to it explaining why it is that high and no higher. If you cannot say, lower it.

Questions about ad rates and margins

Does the general (Standard) strategy charge me if nothing sells?

No. It is cost-per-sale: the fee applies only when the promoted item sells and someone clicked its general ad in the previous 30 days. On eBay's current rules that click need not come from the buyer, so most sales of a well-viewed promoted item carry the fee. Priority (formerly Advanced) is cost-per-click, so clicks cost money whether or not a sale follows.

Should I just use eBay's suggested ad rate?

Treat it as a starting point and a market signal, not your limit. The suggestion reflects what other sellers pay, not what your item can afford. Compare it with your own highest affordable rate and never go above that.

Can AI tell me what rate will win sales?

No. It can do the arithmetic on your margins and lay out a test, but only your own results show what a rate buys. Be wary of any answer that quotes a conversion rate or a rate that works for most sellers.

How long should I test one rate?

Long enough to see enough views and sales to compare, which depends on your volume. For slow items that may be several weeks. Change one thing at a time, so a rate change is not mixed up with a price or title change.

Further reads

• eBay Bulk Listing Tool: Magical Bulk Listing vs Third-Party Apps — Listing tools change how fast you can build a promoted range.
• eBay Listing Removed? Read the Notice, Fix It and Appeal With AI — A removed promoted listing stops spending, and costs sales.
• AI for eBay Descriptions: A Condition, Flaws and Measurements Template — Honest condition wording lowers returns, which protects margin.
• eBay Bulk Listing for Sports Cards: An AI Error Checklist — Margins on cards are thin; avoid errors that cost returns.
• DeepSeek or ChatGPT for eBay Sellers: Which Writes Better Listings? — Which model to ask for spreadsheet help.
• Can You Use AI Photos on eBay? What the Picture Policy Says — eBay's picture policy has no AI-specific rule, but six clauses still decide what an AI edit may do. Each one, and one edit put through them.
• eBay Store Automation With AI: What Is Allowed and What Isn't — Which eBay store jobs AI can safely draft, where the bot and agent rules bite, and how API-based tools differ from scrapers.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: eBay seller centre Promoted Listings page (general and priority strategies); eBay's advertising attribution page (30-day any-click attribution for general campaigns from January 2026); eBay's general campaign strategy page (2% to 100% range, fixed and dynamic rates); eBay's Promoted Listings help page (ad rate as a percentage of the total sale amount, last-click model, priority clicks always charged); third-party seller guides only to confirm the older Standard and Advanced names.