Amazon Marketing Cloud · Audiences
AMC Audience Recipes: Seed SQL, Activation, Minimum Sizes
Rule-based AMC audiences are SQL queries whose result is a list of user_id values. The recipes below only use tables and fields from the AMC documentation and can be submitted directly as an audience query.
Rules that apply to every recipe
- The final SELECT contains only user_id.
- Only tables ending in _for_audiences (your own uploaded tables keep their names).
- Rule-based audiences need at least 2,000 records; Amazon recommends 2,500 or more, otherwise creation can fail.
- Lookalike seeds need at least 500 and at most 500,000 users.
- Check the size first with a normal analytics query (COUNT(DISTINCT user_id) over the same logic).
- Audiences land in DSP or sponsored ads: as targeting in DSP and Sponsored Display, as a bid boost in Sponsored Products and Sponsored Brands.
Retargeting
Cart abandoners
Why: Added an advertised product to the cart but did not buy: the group closest to purchase.
Activation: DSP targeting or Sponsored Display targeting; in SP/SB as a bid boost.
WITH carted AS (
SELECT user_id
FROM amazon_attributed_events_by_conversion_time_for_audiences
WHERE add_to_cart > 0
GROUP BY user_id
),
purchased AS (
SELECT user_id
FROM amazon_attributed_events_by_conversion_time_for_audiences
WHERE purchases > 0
GROUP BY user_id
)
SELECT c.user_id
FROM carted c
LEFT JOIN purchased p ON c.user_id = p.user_id
WHERE p.user_id IS NULLRetargeting
Detail page viewers without purchase
Why: Viewed the product detail page after an ad exposure but did not buy.
Activation: DSP targeting or Sponsored Display targeting; in SP/SB as a bid boost.
WITH viewers AS (
SELECT user_id
FROM amazon_attributed_events_by_conversion_time_for_audiences
WHERE detail_page_view > 0
GROUP BY user_id
),
purchased AS (
SELECT user_id
FROM amazon_attributed_events_by_conversion_time_for_audiences
WHERE purchases > 0
GROUP BY user_id
)
SELECT v.user_id
FROM viewers v
LEFT JOIN purchased p ON v.user_id = p.user_id
WHERE p.user_id IS NULLRetargeting
Sponsored ads clickers without purchase
Why: Clicked a sponsored ad and did not buy afterwards.
Activation: DSP targeting or Sponsored Display targeting.
WITH clickers AS (
SELECT user_id
FROM sponsored_ads_traffic_for_audiences
WHERE clicks > 0
GROUP BY user_id
),
purchased AS (
SELECT user_id
FROM amazon_attributed_events_by_conversion_time_for_audiences
WHERE purchases > 0
GROUP BY user_id
)
SELECT c.user_id
FROM clickers c
LEFT JOIN purchased p ON c.user_id = p.user_id
WHERE p.user_id IS NULLRetargeting
DSP-exposed non-buyers
Why: Saw a DSP impression but did not buy: for the second wave of a campaign.
Activation: DSP targeting.
WITH exposed AS (
SELECT user_id
FROM dsp_impressions_for_audiences
GROUP BY user_id
),
purchased AS (
SELECT user_id
FROM amazon_attributed_events_by_conversion_time_for_audiences
WHERE purchases > 0
GROUP BY user_id
)
SELECT e.user_id
FROM exposed e
LEFT JOIN purchased p ON e.user_id = p.user_id
WHERE p.user_id IS NULLExclusion
Purchasers to exclude
Why: People who just bought do not need to be paid for again in acquisition campaigns.
Activation: As an exclusion audience in DSP campaigns.
SELECT user_id
FROM amazon_attributed_events_by_conversion_time_for_audiences
WHERE purchases > 0
GROUP BY user_idLoyalty
New-to-brand buyers for a second purchase
Why: Bought from the brand for the first time in 12 months: the goal is the second purchase.
Activation: DSP targeting with a cross-sell creative.
SELECT user_id
FROM amazon_attributed_events_by_conversion_time_for_audiences
WHERE purchases > 0
AND new_to_brand = true
GROUP BY user_idLookalike seed
Repeat buyers as a lookalike seed
Why: At least two purchases: a small, valuable group from which Amazon models similar users.
Activation: As the seed of a lookalike audience (500 to 500,000 users).
SELECT user_id
FROM amazon_attributed_events_by_conversion_time_for_audiences
WHERE purchases > 0
GROUP BY user_id
HAVING SUM(purchases) >= 2