Cannibalisation

Cannibalisation in Amazon advertising is the phenomenon where paid ad placements capture sales that would have occurred organically anyway. The ad is "credited" with the sale in standard reporting, but the incremental contribution of the ad to revenue is small or zero. The advertiser is paying for traffic they already owned for free.

Where cannibalisation lives

The single largest cannibalisation surface in most accounts is branded keyword spend:

• A shopper types your brand name into Amazon.
• Your branded keyword ad appears at TOS, your organic listing appears immediately below.
• The shopper clicks the ad. The sale is attributed to paid.
• Counterfactual: had the ad not appeared, the shopper would have clicked the organic listing two positions below. Same sale, no ad spend.

Other cannibalisation surfaces:

• Defensive Sponsored Display on your own PDP — your own visitor sees your ad for your own product.
• Highly-ranked organic keywords where your ad is one position above your organic.
• Cross-format overlap — a Sponsored Brand banner + a Sponsored Product ad both winning the same shopper at different funnel stages.

Why brands still defend branded keywords

If branded spend is largely cannibalistic, why run it? Three structural reasons:

1. Competitor conquesting. If you don't bid your brand, a competitor will. Their ad appears above your organic, and the shopper considers their product instead.
2. Real estate dominance. Brand + Sponsored Brand banner + organic = the entire top of the SERP. Even if individual placements cannibalise, the combined SERP dominance shapes purchase decisions.
3. NTB attribution. Branded keywords often surface in NTB data — new customers searching the brand for the first time.

The right framing is not "stop spending on branded keywords" — it's "spend less on branded than uncontested, and treat the spend as defensive overhead, not direct response."

Measuring cannibalisation: iACOS and AMC holdouts

The two analytical tools that quantify cannibalisation:

• Incremental ACOS computed against an organic-only baseline.
• Amazon Marketing Cloud holdout studies — pause spend in geographic or temporal cells and measure the lift the spend was actually producing.

Both routinely show that branded keyword spend has a 60–85% cannibalisation rate. The true incremental ACOS on branded is typically 3–6× the reported ACOS.

Common mistakes

• Reading branded keyword performance from the standard ACOS report and declaring it "best campaign in the account." It's the most-cannibalised campaign in the account.
• Eliminating branded entirely. The defensive value is real even if the direct-response value is small.
• No holdout testing. Without AMC or equivalent, cannibalisation is unmeasured — and what isn't measured isn't managed.