Auto Optimization

Use SpringServe's auto optimization to reprioritize your waterfall. In the Optimization tab of a supply tag, enable optimization and select your parameters. Optimization runs every 20 minutes and uses the parameters you specify to reprioritize your demand tags.
Optimization parameters can be set on multiple levels:
  • account - basic settings
  • supply partner - settings tab
  • supply tag - optimization tab

Optimization Metric:

  • Opportunity Fill: This metric prioritizes demand tags based on imps/opportunities. It allows a tag that is lower on the waterfall to climb back up.
  • Request Fill: This is going to prioritize demand tags based on imps/requests. This tends to favor publisher fill.
  • Score: This prioritizes using the formula of Net Revenue per 1,000,000 requests. This takes in to consideration PPM, Request Fill, and fees.  This will favor your profitability.
  • Response Time: Prioritizes demand tags based on average response time. This works well in a variety of situations: short VPAID window, trafficking directly to a publisher's player, trafficking to an adserver that is strict on response time.
  • Fill Speed: This prioritizes demand tags based on Request Fill/Response Time. This metric tends to favor publisher fill but also adds importance to demand tags that return quickly.
  • Score Speed: Prioritizes demand tags based on Score/Response Time. This metric tends to favor profitability but also adds importance to demand tags that return quickly.
  • Opportunity Score: This prioritizes using the formula of Net Revenue per 1,000,000 opportunities. This takes in to consideration PPM, Opportunity Fill, and fees. This will favor your profitability while taking into account the fact that tags lower on the waterfall will get fewer opportunities.
  • Opp Fill Speed: This prioritizes demand tags based on Opportunity Fill/Response Time. This metric tends to favor publisher fill but also adds importance to demand tags that return quickly, while allowing tags a chance to climb up the waterfall.
  • RPM: This prioritizes the demand tags based on revenue * 1000 / impressions. This should be used in CTV without pods or with custom pods, and ensures tags with a higher RPM get priority in a tier.
  • RPS: This prioritizes the demand based on the RPM divided by the duration of the ad which allows for selection of the demand that will give you the highest revenue per second, meaning more ads can be served and more revenue generated in a single broadcast call. This is recommended for use with a dynamic pod set up.

Lookback Minutes: the number of minutes the adserver looks back to make an optimization decision

  • A shorter window produces more impulsive and market driven results
  • A longer window produces less re-prioritization (allows an advertiser to remain prioritized even if they have a hiccup in fill)

Optimization Type:

For Direct Connect supply tags, there is an option to set Optimization Type. The choices are 'SpringServe' and 'Custom'. The option 'SpringServe' allows SpringServe's optimization to select the best demand from your waterfall for your supply partner. Setting Optimization Type to 'Custom' requires that SpringServe's optimization adheres to the tiers of your waterfall when selecting demand for your supply partner. The Optimization Type 'Springserve' is the default as it will likely yield the best performance for your supply partner.

Notes on How to Select your Optimization Parameters

A poorly performing tag that is low in your waterfall has trouble climbing back up.  It must compete against the tag in the next lowest priority and produce better results, which is rare. Much of your decisioning about how to use these settings depends on a number of factors:
  • How much request volume does the supply tag get? The more request volume, the more actionable optimization can be.
  • Where is the tag being trafficked?  If its going into another ad server in to another waterfall, does that waterfall also auto opti? Does it prioritize you based on fill?  If so, fill based metric might be best.  If it is going to a player, response time or score could be better
  • How often to you plan on monitoring performance?  If you are going to watch it closely, you can use shorter windows and reprioritize your waterfall manually if need be. If you want it more on 'auto pilot' a longer window is better.  Keep in mind, when you use quick stats for optimization and you have a short window, it is easy to think the adserver is making the 'wrong' decision, but the ad server is looking at a snapshot, and in quickstats you are looking at a full hour, day, yesterday, or 7 days.
  • What is the most important metric to you: Latency? Profitability? Positioning in your publishers waterfall?  Depending on these factors - Response Time, Score, and Opp/Request Fill respectively. If you want to take into account multiple aspects, select a speed metric.

Additional thoughts:

  • An important strategy to utilize in spring is the process of taking a demand tag, duplicating it, and assigning it to one supply tag.  This will allow you to target a specific demand tag specific to a supply tag. By doing this, individual tags specific to supply will have higher fill, consume less opportunities, and will remain higher in the waterfall.