Adaptive Optimization models are Criteo’s most advanced engine optimization models. They allow you to input a target, against which the Criteo Engine optimizes your CPC’s in real time to ensure you hit those targets. Performance is maximized by adjusting your bid levels every hour to ensure your target is reached with more precision than you would be able to achieve through manual CPC management.
This optimization, which is on 24/7, dynamically reacts to external factors such as levels of competition on placements and seasonality effects. This, in turn, results in an average 5 % increase in sales at a constant ROI.
Criteo allows you to set targets for Cost of Sale with our Adaptive Revenue Optimization (ARO) model and Cost per Order with our Adaptive Conversion Optimization (ACO) model.
Adaptive Optimisation model set up
Adaptive Optimization models are a new generation of business optimization models and you need to start the process by accepting a Terms & Conditions addendum in the Management Centre which reflects their new way of optimizing your campaign.
The addendum covers the dynamic update of bids to reach the relevant target (Cost of Sale or Cost per Order). If you are not prompted to agree to the terms & conditions addendum when logging in to the Management Centre, you need to ask your Account Strategist to activate it.
Once the Addendum has been accepted, you need to discuss your campaign’s starting target with your account strategist. This is the target the Adaptive Optimisation model will optimize against.
To ensure the smoothest transition possible, we recommend you keep a campaign running for about a week with a stable COS/CPO on the relevant standard optimization model before switching to one of our Adaptive Optimization models. We also recommend that the starting target is relatively close to the current campaign COS/CPO, otherwise, our Adaptive Optimization algorithm could need several days to learn and find the right bid level for that target.
Managing and evaluating performance
The Adaptive Optimisation models rely on your campaign sales to calculate the perfect bid level. As such, accuracy, target stability and performance will improve with more sales volume. We, therefore, recommend having an average of 5 daily attributed sales. To maximize performance through our Adaptive optimization models we suggest merging campaigns which have similar targets.
Smaller campaigns will typically take longer to reach stable performance.
For the engine to calculate the optimal bid level for a given target, it is important to keep the input target stable. Changes of more than 20 % should be avoided, as it can take up to one week for the engine to achieve a stable COS/CPO in line with the new target.
This new way of optimizing your campaign against a set target requires a slightly different approach to managing and evaluating your campaign performance.
Performance against the target should be measured and evaluated over a 7-day window for bigger campaigns which generate 20+ sales a day and 14 days for smaller campaigns.
This is due to a combination of things, such as the engine’s response to valuable traffic and the time between a click and a resultant sale.
You can be assured, however, that the engine is constantly working to achieve your target.
Adaptive Optimization models adapt dynamically to external conditions but might not have enough insights about significant events affecting your traffic and sales such as Black Friday where the conversion rate jumps significantly. If you anticipate such an event, make sure to contact your Account Strategist to implement the best strategy for your campaign.
If you are worried about exceeding your budget through this automated bidding management, you can, of course, revise your target at any time in the Management Centre to influence the rate of spend, as per the common practice when managing CPCs. Alternatively, you can contact your AS to discuss steps to control the trend of the spend of your campaigns.