Ad group models, optimization, predictive KPIs, and bidding
10 min
this article explains the following ad group settings models docid\ c1wuun2rd3rthba9u8rof optimization automation docid\ c1wuun2rd3rthba9u8rof predictive kpis docid\ c1wuun2rd3rthba9u8rof bidding docid\ c1wuun2rd3rthba9u8rof intraday pacing docid\ c1wuun2rd3rthba9u8rof what's a model? a model is a network of mathematical functions that takes the information from a bid request and scores it the score represents the relative likelihood that winning that bid and serving that impression will help achieve the model's goal the score can also be used to decide whether to bid on that request key points to know about models models are created by cadent's data science team, either manually or automatically models are re trained daily to continually improve their performance there are several types of models, and they can be available across all accounts or limited to a specific advertiser (e g , a custom cpa ad group) you can assign a primary model and as many as 12 additional models to an ad group, but in reality most ad groups only have two or three you can also opt for not using models at all the probability models (currently labelled "ctr" in ad manager) predict the likelihood an impression will deliver the results you want custom models, on the other hand, aren't predictive—they just give a relative score from 1 to 100 custom models are used with cpa campaigns optimization automation optimization automation uses ai to dynamically adjust bids and/or predictive model scores to balance performance and delivery there are three optimization automation options performance and price optimizer docid\ c1wuun2rd3rthba9u8rof performance optimizer docid\ c1wuun2rd3rthba9u8rof price optimizer docid\ c1wuun2rd3rthba9u8rof once you choose an optimization automation, you'll then select an optimizer performance and price performance and price optimization automation balances the two approaches simultaneously this is the most commonly used optimization automation performance the performance optimization automation will update an ad group's min model score to maximize performance, while ensuring the ad group still delivers its daily goal for example your primary kpi is viewability and your ad group is not meeting your viewability delivery goal you set it at 95% or above and it's coming in at 87% the performance optimization automation can raise the minimum viewability score that the ad group will bid on meaning, it becomes more selective, going after impressions that are more likely to be viewable price optimizer the price optimization automation dynamically adjusts an ad group's bid price to reach the primary kpi goal for example your primary kpi is viewability and your ad group is not meeting your viewability delivery goal you set it at 90% or above and it's coming in at 82% the price optimization automation can raise the bid price to get more viewable, better quality impressions predictive kpis a predictive kpi is a metric used to establish and evaluate a goal for an ad group key points to know about predictive kpis all ad manager kpis are commonly used metrics, and each one is defined by a simple formula for example, cost per click = \[total spend]/\[clicks] some of the kpis, like click through rate, are tailored to specific channels, but most can be applied to all ad group channels kpis are paired with one or more models to optimize towards the kpi goal (see screenshot below) an ad group can have up to three predictive kpis the first is the primary kpi and the other two are secondary the optimizer affect only the primary kpi \<font color="#a855f7"> \</font> bidding this is where you'll fine tune your bidding parameters bidding settings are contextual based on the predictive kpi you selected if you select a cpm or cpu predictive kpi, the bidding type will automatically match the setting (see screenshot below) if you select "custom or n/a" or "none," you can manually set the bidding type bid di ng type cpm cpm represents a fixed bid price that can be adjusted by the optimizer here's an overview of the settings related to cpm cpm bid price a user is willing to pay for 1,000 impressions cpm bid floor minimum price a user is willing to pay for a thousand impressions this setting is only required when the price optimizer is enabled cpm bid ceiling maximum price a user is willing to pay for a thousand impressions this setting is only required when price optimizer is enabled ecpm, ecpc, or ecpcv price ceiling maximum allowable effective pricing unit to ensure spend stays within budget constraints bidding type cpu cpu multiplies the target by the model score to determine the bid price target desired cost per unit goal when cpc or cpcv optimizer is applied target cpu floor minimum acceptable cost per unit, ensuring bids are competitive while maintaining efficiency target cpu ceiling highest acceptable cost per unit, to prevent overspending on performance goals ecpm, ecpc, or ecpcv price ceiling maximum allowable effective pricing unit to ensure spend stays within budget constraints min possible cpm bid minimum price a user is willing to pay for 1,000 impressions this setting is only required when the price optimizer is enabled max possible cpm bid maximum price a user is willing to pay for 1,000 impressions this setting is only required when the price optimizer is enabled intraday pacing the rate at which the platform meters out impression buying/spend throughout the day even we spread the daily goal evenly across every hour and then every minute of a given day asap we try to deliver the daily goal as soon as possible, making the minute goal the daily goal

