Pulse Distribution Engine

0 Kudos
Started ‎08-30-2016 by
Modified ‎10-27-2016 by

Pulse Distribution Engine

by Ooyalan ‎08-30-2016 01:25 AM - edited ‎10-27-2016 04:29 AM (781 Views)

Delivery Pacing

 

The Pulse Distribution Engine uses a delivery algorithm which takes numerous pieces of data into account (for example, goal type, targeting and prioritisation, site traffic, insertion policy settings, ad weights, and clash protection) when deciding which goal to pick and which ad to display. The main goals of the Pulse Distribution Engine are to:

  • make sure all goals reach their configured delivery goal,
  • distribute the delivery evenly across campaigns and time,
  • avoid under- and over-delivery,
  • distribute ads from different goals according to the Distribution and Prioritisation Rules,
  • maximise your revenues and deliver on your advertiser commitments.

Pulse's strength lies in its strategy to deliver campaigns dynamically over the campaign period by using real-time data which is recalculated in every ad call.

 

As a result, Pulse:

  • accounts for changes in traffic patterns and delivers a bit more on days with high traffic and a bit less on days with low traffic,
  • maximises fill rates and ensures that inventory is not wasted by trying to follow a strict daily cap,
  • ensures campaigns deliver a bit less on days where other campaigns need more impressions, and then increase again a few days later to catch up.

Note: There are also a number of settings in Pulse which you can manually adjust to optimise your campaigns. For more information, refer to Campaign Optimisation.

 

Ideal Delivery

 

All impression goals with an end date in Pulse should deliver according to an ideal delivery curve to ensure an even distribution during the entire goal period and at the same time safeguard against over- and under-delivery. The delivery algorithm has a built in pause and brake mechanism which prevents the goals from over-delivering and finishing too early, leaving you without ads to display until the end of the goal's lifetime. The ideal delivery curve has a built in safety margin which prevents under-delivery at the end of a goal's lifetime and goals strive to be as close to this ideal delivery curve as possible.

ideal_delivery.png

 

Dealing with Under- and Over-Delivery

 

Possible reasons for under-delivery:

 

  • Less traffic than expected.
  • Uneven site traffic.

    Example: You have a goal that is a week long and needs to deliver 1000 impressions. From historical data we know you get most of your traffic on Thursdays because of a popular show. After 3 days, you are 3/7 into your goal time-wise, and 10% into the expected traffic. This is because Pulse expects the Thursday spike so it delivers fewer ads the first 3 days and saves the impressions for Thursday. For the first 3 days the goal might seem like it is under-delivering, but after the traffic spike on Thursday the delivery should be on track. If the delivery is not on track after Thursday, then Pulse tries to catch up.

  • Too specific targeting that limits the ad to a very small number of potential viewers.
  • Global targeting rules are different from the campaign/goal targeting rules and you did not override them.
  • You are overbooked by selling 1 million impressions for a day, but only have 500 000 views in a day, which means the higher prioritised campaigns/goals will deliver before the lower prioritised camp/goals.
  • You changed campaign priority but forgot to change the goal priority so they match, or you do not have the “inherit campaign priority settings” option on.

Possible reasons for over-delivery:

 

  • Configuring goals with a short lifetime and a low delivery goal compared to available inventory.
  • Issues with 3rd party ads causing passbacks to be used.

Note: A common cause for concern is related to subsequent changes made to goal configuration during its lifetime, for example shortening or extending the goal's lifetime and/or increasing or decreasing the delivery goal, which can make it seem like the goal is under- or over-delivering. However, this is not an issue with the delivery, just a reaction to the change you made until the delivery pace adjusts to the new configuration.

 

How Pulse prevents under- and over-delivery

 

  1. The image below shows a goal under-delivering and trying to catch up.
    under_delivery.png
    Catching up: when the goal is under the linear delivery, in this case the black "midroll linear aggregation" line, Pulse tries to catch up by applying a correction factor to the intended delivery so it actually delivers more.

  2. The image below shows a goal constantly over-delivering and therefore hovering around the upper boundary where the delivery is paused.
    pulse_delivery_pause_brake_curves.png

    Braking: when the goal is between the grey "Brake" curve and pink "Brake completely" curve, Pulse brakes the goal in a way that it still delivers, but at a slower pace.

    Pausing: when the goal is above the pink "Brake completely" curve, Pulse completely pauses the delivery and waits for the goal to fall under the pink curve so it can start delivering again.

 

Daily Goal Caps vs Pulse Dynamic Design

 

  • With the Pulse simulation based forecasting approach, which forecasts future available inventory based on historical user traffic and current campaign and account settings, you see whether your campaign is going to deliver as expected or not, this way benefiting from Pulse's dynamic design. For more details, refer to the Planner.
  • The daily goal cap approach is static and following it leads to wasted inventory in case of traffic pattern changes. It implies there is a 100% certainty you are going to have enough inventory to deliver on the daily cap. However:
    • if you get less inventory than you needed, you are not going to reach the daily cap.
    • if you get more inventory than you expected, the platform cannot make use of that extra inventory in a good way. This is why most display ad servers run empty during sudden traffic spikes.