[Computational Ads]1. Bidding Method


This is the first article about computational advertisement. Here I am sharing my learning notes and thoughts when reading Computational Advertising by Prof. Liu.


There are various advertising bidding methods in the market for the customers to choose. The automatic bidding methods can be described in the following diagram:

Bidding Methods You can see that from left to right, the ads platform's risk is increasing and the advertiser's optimizing space is decreasing. I am going to explain these models in details in the next section.


CPT(Cost per time)

  • This is usually aiming for big advertiser's specific advertising event. It gives a advertising place exculsively to an advertiser.
  • This is a sale method and not really a bidding method, since the price is agreed ahead.
  • It's suitable for some advertisements that need strong exposure, and usually for branding advertisement.
  • It's getting less and less popular nowadays.

CPM(Cost per Mille/Thousand Impressions)

  • In this method, your bid price is for one thousand impressions. The system will deliver your ads to get the most exposure and reach to as many as possible within your budget.
  • The advertiser is the person who need to estimate the click value and the click rate of the advertiser and control the risk whether they can make profit through the bidding.
  • This is the mainstream bidding method in video advertisemnet.
  • Variants:
  • oCPM(Optimized CPM):
    • The bidding method is similar to standard CPM, which means it charges by impressions number.
    • Advertisers can choose the aspects that they want to optimize, like increasing conversion rate, lower conversion cost and increasing impression gaining speed.
  • vCPM(Viewable CPM):
    • The bidding type is setting your maximum costs on a viewable 1000 impressions.
    • It counts as a viewable impression after 2 seconds of a video played on YouTube, or 1 second of a display ad is shown on the Display network.

CPC(Cost Per Click)

  • This means your bid is the cost you are willing to pay per click. The system will deliver your ads to users who are most likely to click them at a cost that is as close to your bid as possible.
  • It's used by searching advertising first, and adpoted by many other advertising product soon too. It's the provider's job to estimate the clicking rate and the advertiser's duty to estimate the click value, which decouples the two factors of computing eCPM. So the two sides can use their data optimize their responsible part, which can be more accurate on both sides. So this is the most popular bidding method in the market.
  • Variants:
  • ECPC(Enhanced CPC)/oCPC(Optimized CPC):
    • It's a mix of manual and smart bidding. You set the basic CPC for your ad groups and keywords, but the algorithm gets to optimize them.
    • You can choose whether you want the algorithm to enhance your set bids based on a flat number of conversions, or to optimize the value of conversions, or to target impression share.
  • Maunal CPC:
    • This is the method where you set bids for different ad groups or keywords on your own. This requires a lot of work to optimize the result.

CPS(Cost Per Sale)/CPA(Cost Per Action)/ROI

  • Target CPA is a fully automated bid strategy where advertisers set a target cost per conversion. This is an extreme method where the risks are taken just by the ads provider. The advertisers only need to compute the profit of each conversion to ensure they make money, while the provider need to control the click rate and click value at the same time.
  • It sounds ideal for the advertisers, but when applied to real production there are a lot of issues:
  • The provider can't monitor the actions of users in the advertisers' sites.
  • The advertiser may intentionally cancel orders to decrease conversion rate.
  • The provider may use consignment/hijacking to cheat the advertisers.
  • The data for optimizing one single advertisement is not enough.
  • Given above concerns, we need a platform that can monitor the advertising and conversion process to ensure the benefit for the both sides:
  • App downloading, all the conversion data is accumulated in AppStore, which is a trusted third-party.
  • alimama: The conversion is totally provided by Taobao, which is a trusted platform for both sides.


  1. Bussiness Help Center - Tiktok
  2. The Pros & Cons of Every Automated Bidding Strategy in Google
  3. Google Ads Bidding Strategies: The Ultimate Guide
  4. Computational Advertising, 1st Edition - Peng Liu, Chao Wang