Savvy marketers and programmatic traders like to take control of their campaigns, run multiple tests, evaluate different attributes and decide on the next stage of campaign optimization based on these analysis. Since a lot of this data is not trivial to analyze, by default we rely on ML models to take correct optimization decisions. However, often times, user (campaign manager) context cannot be ignored and there are instances when giving more control to users to impact the final bid price adds a lot of value

Our internal bidding system and performance models were already leveraging multipliers to drive scale and performance for the campaigns. However, today, we have released an updated version of multipliers by allowing the users to add Custom Bid Multipliers that will work in tandem with internal algorithm driven multipliers

Feature Description

When you run a CPC/CPI campaign on Kayzen platform, our ML models calculate the probability of an action (click/install) for each bid request. Apart from using this probability to calculate the bid price for an individual impression, the system also uses an internal multiplier algorithm which boosts bid prices up or down based on some heuristics. For instance, if a specific publisher app is driving good performance for one campaign of the advertiser app but not for the other, the internal bid multiplier algorithm will increase/decrease the bid price for bids placed via that specific publisher depending on the campaign's performance. This works in tandem with the ML model to ensure we can deliver good performance at meaningful scale.

With this release, we are now giving control to the user as well, to impact the final bid price by adding multipliers for specific campaign attributes. This user enabled bid multiplier will work along with ML models and internal multiplier algorithm , thereby giving the user an opportunity to take decisions which leverage both the manual campaign learnings as well as the power of prediction models. In addition, users can also use Bid multiplier for CPM campaigns where the above model heuristics do not apply

Feature Screenshots

  1. A new section called Bid Multiplier is added in campaign create/edit flow

2. You can add bid multipliers for specific values of the following attributes

a.Apps (publisher app bundle, exchange-app site id)

b. Location (Country/State/City) and

c.Exchanges

3. You will be able to view and edit bid multipliers via campaign detail page as well


Sample Use Cases

  1. Increasing spend on specific countries

    You are running a campaign targeting multiple EU countries and are seeing good performance from two of them, say Austria and Spain. You will like to spend more budget on these two countries over the next couple of weeks. While the ML model will figure out the performance angle and change the bidding accordingly, you want to take control of your campaign and boost up the bid price for Austria and Spain before the weekend kicks in. You can now use Custom Bid Multiplier to increase bid price for Austria and Spain thereby increasing the spends for these two countries without the need to create a separate campaign
  2. New Exchange Boost

    You saw an exchange performing well for one of your apps and want to start targeting this exchange in campaigns belonging to another app of yours. Rather than creating a separate campaign to boost up the price for this exchange, you can now just add this exchange in the existing campaign and assign a bid multiplier to increase spend on this exchange so that you can observe and evaluate results faster
  3. Placement Optimization

    Few placements in your campaign are driving good number of installs, within the target CPI, have high win rate but the registration rate is on the lower side. They don't seem to be fraudulent apps and the performance is not that bad that they'll get auto blacklisted by the algorithm. You calculate that if you can reduce the CPI for such placements further by 25%, you can get closer to the cost per registration target at the expense of some installs. You can now leverage bid multiplier to reduce bidding for some of these apps and optimize your campaign accordingly

Key Restrictions

  1. Conflicting bid multiplier values- The system does not allow conflicting bid multiplier values for the same request. For instance, you cannot put multiplier as 1.5 for Germany and 2.1 for Berlin since a bid request from Berlin is also a bid request from Germany
  2. Max applicable bid multiplier value- This is the maximum bid multiplier value that can be set at a campaign level. The sum value of all the attributes multiplied together cannot be greater than max applicable bid multiplier value. The max multiplier that can be applied at campaign level is defined by advertiser admin. For instance, consider the following case-
    a. Max multiplier for an advertiser app is 15
    b. User adds various multipliers for Berlin (5), OSVersion 13.0 (2), Exchange Mopub(4)

If a bid request comes from Mopub for a user in Berlin having a phone with OSVersion 13.0 , the max multiplier that will be applied to this bid will be 15 and not 40 (5*2*4). This safety check ensures that the users can add multiplier values for different dimensions without worrying about the potential cascading effect that can happen while calculating the final bid price.

Custom Bid Multiplier can be a very handy tool for campaign managers to take control of bidding decisions when needed. Should you have more questions or feedback about the release, please do reach out to your Programmatic success specialist


Did this answer your question?