Branded vs non-branded on Amazon: what the split actually costs
Branded clicks cost 29% less, convert 1.9x better, and run at a third of the ACoS of non-branded. That gap is why a single blanket ACoS target quietly moves budget onto your own brand name.
What does branded search cost compared with non-branded?
Non-branded search is where Amazon growth comes from, and it is priced accordingly. We looked at 15 US Amazon stores on the Laurence platform: about $103M in combined annualized Amazon revenue and $4.0M of Sponsored Products spend over the 90 days ending August 25, 2026.
For the median store, a branded click cost $1.63 and a non-branded click $2.48. Branded clicks converted 27.4% of the time; non-branded, 12.0%. Branded ran at a 21.7% ACoS, non-branded at 77.2%. Comparing within each store, branded clicks were 29% cheaper, converted 1.9x as often, and ran at roughly one third the ACoS.
| Metric | Branded | Non-branded | Within-store gap |
|---|---|---|---|
| Cost per click | $1.63 | $2.48 | Branded 29% cheaper |
| Conversion rate(orders per click) | 27.4% | 12.0% | Branded 1.9x higher |
| ACoS | 21.7% | 77.2% | Branded roughly one third |
The pattern held almost everywhere: branded ACoS was lower in 14 of 15 stores, branded conversion higher in 14 of 15, and branded cost per click lower in 12 of 15. The two exceptions are structural: a specialty retailer whose "brand terms" are the manufacturer brands it resells, and a store whose name shoppers almost never type.
One definition matters for everything that follows. We classify the shopper's search term, not the advertiser's keyword, because the same broad-match keyword wins both kinds of traffic. A search term is branded when it contains one of the store's brand terms (brand name, sub-brands, common misspellings) or targets a product page in the store's own catalog. Searches for competitor brands count as non-branded. Classifying at the keyword level instead understated branded spend by about 21% and overstated non-branded sales by about 56% against a hand-labeled sample. A branded-split number that does not state its rule at this level cannot be checked.
How much of a brand's ad budget sits on its own name?
This is the number a CMO can hold their own program against, and as far as we can tell nobody publishes it. The median store put 25.6% of its Sponsored Products spend on branded terms, and that quarter of the budget produced 53.6% of ad-attributed sales.
The 0.5% floor is a private-label store: shoppers do not search its name, so there is nothing to defend. The 77.3% ceiling is the reseller defending the manufacturer brands it carries. A brand with real name recognition selling its own products typically sits between 12% and 33%.
How the ACoS metric drifts away from incremental sales
Most brands manage Amazon ads to one number: a blended ACoS, ad spend as a share of ad-attributed sales, across the whole account. Anyone graded on that number, whether an agency, an in-house team, or a bidding tool, will find that branded spend is the cheapest way to improve it. Branded clicks are cheaper and convert twice as often, so every dollar moved from non-branded to branded pulls the blended figure down.
Here is what that looks like with the median store's numbers. Nothing about either segment changes: branded still runs at 21.7% ACoS and non-branded at 77.2%. The only thing that moves is how the budget is split between them.
The takeaway: a blended ACoS can improve substantially while incremental revenue stands still, because the number rewards moving budget onto your own name. Nobody has to act in bad faith for this to happen; it is what optimizing the stated number looks like. If your team or agency reports one ACoS for the account, you cannot tell efficiency gains from reallocation, which is why the two segments need to be reported and targeted separately.
What is a click in each segment actually worth?
The drift follows from three pieces of arithmetic. Each points to what should replace a blended target.
1. Blended ROAS is just a weighted average
Blended return on ad spend is the two segments' ROAS weighted by their share of spend. At our medians (branded ROAS about 4.6, non-branded about 1.3), every ten points of spend moved to branded adds about 0.33 to blended ROAS with no change in demand. Any optimizer pointed at the blended number, human or software, climbs this slope first because it is the cheapest direction available.
2. ACoS prices attributed sales; decisions should price incremental sales
Some share of branded-attributed sales would have happened without the ad. Divide each segment's ACoS by the share of its sales that were truly incremental and you get the real cost of an incremental dollar. Assume non-branded is fully incremental, the assumption least favorable to branded. Then the break-even is 21.7% ÷ 77.2%, about 28%: if fewer than 28% of your branded-attributed sales are truly incremental, branded is the more expensive channel per incremental dollar, despite reporting one third the ACoS.
Nobody has measured branded incrementality on Amazon. The point is not that branded spend is bad. The point is that its case rests on a number most reporting silently sets to 100%.
3. A bid should come from what the click is worth
The most you should pay for a click is its expected conversion rate, times average order value, times the segment's target. Run one blended target and branded's willingness to pay is 2.3x non-branded's at identical order values (27.4% against 12.0%), so the blended optimizer outbids on branded in every auction. The full value of a click also includes repeat-purchase margin and the organic-rank value covered next, and the segments differ on every term: branded has high conversion, an incrementality discount, and no rank value; non-branded has lower, noisier conversion, near-full incrementality, and the only rank value there is.
Where the "99% cannibalized" number actually comes from
The case against defending branded terms usually arrives with one statistic: 99% of branded ad clicks would have happened anyway. The source is Blake, Nosko, and Tadelis, "Consumer Heterogeneity and Paid Search Effectiveness," Econometrica 83(1), 2015 (NBER w20171). eBay turned off paid search ads on its brand keywords on Yahoo and MSN, with Google as the control, and roughly 99.5% of the click traffic came back through the organic links sitting just below.
It is a great study, and it describes a specific situation: a household-name brand, on a general search engine, with its own organic result directly under the ad and nobody else bidding on its name. When your brand terms look like that, the finding probably applies to you too. On Amazon they usually do not. The slot your ad vacates is filled by a competitor's ad, not your organic listing, and bidding on rivals' brand names is standard practice. On a contested brand keyword, the shopper who does not click your ad may not scroll to your listing; they may buy the product sitting where your ad was.
Where competitors are bidding on your name, defending it is buying back a sale, not paying for one you already had.
Nobody has published an Amazon-native randomized branded shutoff, so branded cannibalization on Amazon is real but unmeasured. The practical rule: defend your brand terms where the auction is contested, hold back where it is not, and price the defense against what a conquesting competitor would take, not against a portfolio ACoS it will always beat.
How non-branded spend compounds into organic rank
Amazon's ranking rewards sales performance on the query, and it does not distinguish a paid conversion from an organic one. So every non-branded ad conversion does two jobs: it books margin today, and it raises your purchase share on that query, which is what organic rank responds to. The organic sales that follow never touch the ad console, so ACoS cannot see them.
It is zero on your own name. You already hold the top organic slots for your brand terms, so a branded ad conversion buys no rank.
It compounds. Rank drives organic sales, and organic sales drive rank. Cutting a non-branded program shows up in the P&L months later, in the organic line, where nobody attributes it to the cut.
It is largest where you convert well but are rarely seen. Incremental conversions on those queries move rank fastest.
We are deliberately not publishing an elasticity number; public claims about how hard paid sales move organic rank have no stated method, and the response is query- and account-specific. Even without a number, the sign is clear: judged on ACoS alone, non-branded is always undervalued, and branded never is.
How to set separate targets for the two segments
Branded and non-branded are different products that share an ad console. One defends demand your brand already created; the other buys demand from the category. Three changes fix the drift.
Report the two segments separately, always. A blended ACoS that improved means nothing until you know which segment moved. The split takes one classification rule and an afternoon.
Give each segment its own target. Judge branded against the cost of defense: what a conquesting competitor takes when you vacate the slot. Judge non-branded as customer acquisition: against contribution margin, repeat purchase, and the organic rank it is building.
Price bids from what each click is actually worth. This is how Laurence runs it: we model conversion rates per target, price the uncertainty, and bid to each segment's own economics every hour, so branded spend cannot masquerade as performance.
If your reporting shows one ACoS line for the whole account, the split above is the first thing worth asking for. It is also the first thing we show you: the branded versus non-branded breakdown of your own account is on the first report you see.
Get started
