This article will go over each of the Sizmek DSP Automation Services functions.
Function that controls daily budget on the line item level. Rather than pacing evenly, Autopacer uses the following to increase or decrease budget compared to the even pace baseline:
Adv Rfi Ratio: Metric attribution between Third Party ad server and Sizmek DSP. Represented as a percentage
Front Load Fraction: At the beginning of a flight, daily budget is increased and tapers down over the length of the flight. Represented as a percentage
Daily Modifier: Expected campaign goal performance. Min and max percent difference is set by the “Allowed daily pacing variance +/- from even” goal in DSP. Represented as a multiplier
Use Autopacer Change History (link) to verify information is accurate and to see how each of the three modifiers are affecting daily budget.
Function that increases or decreases the min and max daily cost goals to deliver budget in full and maximize performance. Will only affect Bonus and CPA based pricing lines
Function that controls which model and model versions are assigned to campaign objects. Model Manager assesses the campaign metadata (i.e. targeting settings, goals) and runs an A/B test to decide to switch model and/or model version assignments or to keep the current assignment.
Note: Model Experiment is a function sometimes appearing in Change History. This is not Model Manager. Model Experiment is a manual assignment of a model performed either by R&D (if testing a new model) or by a CSM in Apollo (should be uncommon as this is a legacy process).
Global Margin Regulator (GMR)
Function that adjusts Min and Max Margin based on delivery and performance to maximize net revenue while achieving performance goal. GMR will only affect Variable Margin campaign types and can be disabled by Analytics if there is a business case.
Function that optimizes delivery, performance, and margins. In general regulus does the following:
Regulus first optimizes delivery without violating min margin
If Regulus can achieve impression goal, it will then maximize performance goal
If Regulus can achieve impressions and performance goal, it will then maximize margin (net revenue)
Specific optimizations are determined by the pricing type of the line. More details documented here.
Note that Budget Optimizer is not a part of the new DSP UI and it is recommended to instead use tactic grouping. Tactic Grouping functions similarly, but is a better way to achieve the same goal as it is well tested and widely used.
Function that controls tactic allocations based on revenue (delivery), performance, and margin maximizations. Budget Optimizer needs to be enabled at the line item level. The data it uses to make decisions are:
Min Margin: System will not accept any margin less than this value set even if line / tactic(s) are not achieving performance goal
Min Revenue / Impressions: How much the line needs to deliver on a daily basis to deliver in full, regardless of if line / tactic(s) are achieving performance goal
Max Revenue / Impressions: The revenue potential of the tactic
If did not achieve daily budget previously, what tactic delivered is assumed to be the max
If did achieve daily budget previously, assumed that max revenue / impressions can be increased
Performance Goal: Campaign goal set
Expected performance: Projected performance based on historical performance data (Calculated mean + confidence intervals with 95% confidence)
Using the above information, Budget Optimizer will first run a goal maximization based on the min margin and revenue (pacing) settings. If it believes the line can achieve the goal, will rerun to maximize budget while still achieving performance goal. If we cannot achieve the goal, conversion maximization will rerun but now it sets min revenue / impressions to 50% of the daily pacing (in effect forcing delivery even though it knows tactic(s) are not expected to achieve performance goal). Output will always be the recommended budget allocations.
Note: In effect, setting all tactics to 100% should work equivalently to Budget Optimizer. The difference is that with Budget Optimizer enabled, allocation decisions are displayed on the front end in DSP.