Ad Load, Slotting, and Filtering

The quantity and positioning of sponsored listings are as important as their content. This gives way to two important optimizations:

• Ad load. The number of slots made available to show ads to the user. The fill rate is the rate at which ads are actually inserted into those potential slots.
• Ad slotting. The positioning of the ads slots among the organic results.

Improperly implemented ads techniques may disrupt the user experience and compromise long-term revenue. Clients can choose to dynamically communicate the ad load and ad slots on the delivery request, or Promoted can optimize this positioning within its own Blender rules. Promoted will work with the marketplace to determine ideal ad load and slotting that ensures a positive user experiences

Prefiltering and Ads Insertions

Ad slotting dictates the positions in which ads may be inserted; it does not guarantee ads will be inserted into those positions. Promoted makes a model-informed decision to help decide which ads are low quality. There are many reasons why an ad may be eliminated from consideration for lacking quality:

• Prefiltering on the client's end (usually for minimum relevancy or eligibility thresholds) before sending the ads candidates to Promoted
• Prefiltering on Promoted's end, which may include enforcing minimum p(click) or p(conversion|click) values and additional relevancy or eligibility threshold
• Adjustments to the Complementary Bid, a factor in the insertion bid that controls for the user experience. See Ads Auctions and Bids for more information.

In some cases, ad slots may be repositioned or replaced with organic results, especially for queries with limited retrieval candidates (e.g., a specific query with an exact text match). There is an upper bound of revenue that can be pulled from the marketplace, especially when factoring in long-term considerations. For example, bombarding users with ads may increase short-term revenue but decrease user experience and long-term user retention.

Ads Slotting in Blender

Promoted incorporates ads slotting information into the Boosted Blender rules. Promoted also offers probabilistic, weighted, or situation-dependent ad slotting. For example, a sponsored listing could appear in a random position within the top three results. This may benefit both the marketplace and the advertiser by limiting users that may simply scroll past sponsored listings in fixed positions.

Boosted (sponsored) listings

When native ads are implemented with organic search and feed, Promoted calls them Boosted listings. In this example, ads are inserted into slots 1, 3, and 7 of the results.

// New boosted value and score
"boosted = IF(FEATURE(\"adsEnabled=true\") == 1, IF(FEATURE(\"boosted_profile_config.is_boosted\") == 1, 1, 0), 0), 0), 0);",
"boosted_score = IF(boosted == 1, p_click * FEATURE(\"boosted_profile_config.bid_value\"), 0);",

"SELECT<BOOSTED> WHERE(boosted > 0) SORT(clean_long_quote_match, title_query_substring, boosted_score, score) ABSOLUTE_POSITION(0, 2, 6);",

After filtering out ineligible and low-quality listings, the SELECT<BOOSTED> rule uses a boosted_score that combines model-generated probabilities with ad-specific features like the Insertion Bid to determine the particular ads that will be inserted.