How Advertising DSP Optimizes Your Campaigns
This page outlines how the DSP optimization engine, which is powered by Adobe Sensei, optimizes the packages in your campaigns. For tips and tricks on how to manually optimize your campaigns, contact your Adobe Account Team.
Package optimization goals operate at two levels:
-
For each package: DSP allocates budget to each placement within the package based on the placement’s performance against the selected KPI.
-
For each placement/auction in the package: DSP calculates the real-time economic KPI value for each auction per placement, and then uses this value to determine the bid.
note note NOTE The economic value can be heavily weighted based on how well a placement is spending. If a placement is behind its spend goal, then it will be allowed to buy lower-quality auctions. If a placement is easily meeting its spend goal, then it will focus on higher quality auctions.
Package Optimization
DSP can optimize your delivery in two fundamental ways, with 20 variations available to align with your specific performance goal. You can choose to:
-
Prioritize the performance rate
-
Prioritize balancing cost efficiency with performance rate
See Optimization Goals and How to Use Them to determine which optimization goal will help you achieve your KPI(s).
Packages That Prioritize the Performance Rate
For optimization goals that prioritize the performance rate, DSP predicts the performance of each auction and always bids at the Max Bid. Examples of applicable optimization goals include Highest Viewability Rate, Highest Clickthrough Rate, and so on.
This optimization mode works well if:
-
You already know the effective/acceptable CPM level (for example, a historical benchmark).
-
You’re willing to manually adjust the Max Bid for each placement if you experience challenges with scaling.
-
You’re prioritizing scale over efficiency.
Pacing logic pacing-logic-performance
-
If spending is on pace, bidding will become more selective, so that you’ll only bid on auctions predicted to have high performance rates.
-
If spending is behind pace, bidding will become less selective, so that you’ll bid on auctions predicted to have lower performance rates in order to catch up to the pacing goal.
Clearing Price/Bid Shading clearing-price-performance
After it executes the pacing logic, DSP runs the proposed bid through a clearing price prediction model. If the prediction indicates that the bid can be lowered with minimal decrease to the win rate, then the bid is decremented per the prediction.
Packages That Prioritize Balancing Cost Efficiency with Performance Rate
For some optimization goals, DSP predicts the performance of each auction and adjusts bid prices automatically, never exceeding a placement’s Max Bid. Examples of applicable optimization goals include Lowest CPM, Lowest CPA, Lowest Cost per View, Lowest Cost per Click, and so on.
Pacing Logic pacing-logic-balanced
-
If spending is on pace, then DSP becomes more price sensitive, bidding lower amounts to trade off win rate with the pacing plan.
-
If a performance metric is also being balanced (all goals except Lowest CPM), then the predicted KPI is blended into the amount that is bid. You therefore bid higher for auctions that are predicted to be more performant on a “cost per” basis.
-
If spending is behind pace, then DSP becomes less price sensitive and bids higher amounts, up to the Max Bid, to trade off win rate with the pacing plan.
Clearing Price/Bid Shading clearing-price-balanced
After it executes the pacing logic, DSP runs the proposed bid through a clearing price prediction model. If the prediction indicates that the bid can be lowered with minimal decrease to the win rate, then the bid is decremented per the prediction.
Placement Optimization
Placement pre-bid filters are the strictest way to ensure strong performance. DSP uses pre-bid filters strategically across different ad types to achieve performance goals across placements within each package. You can use pre-bid filters concurrently with package-level optimization or independently.
See Placement-level Pre-Bid Filters and How to Use Them to determine which pre-bid filter will help you achieve your KPI(s).