Process the Request

Stay organized with collections Save and categorize content based on your preferences.

A real-time bidding interaction begins when Google sends a bid request to your application. This guide explains how to code your application to process the bid request.

Parse request

Google sends a bid request as a serialized protocol buffer attached as the binary payload of an HTTP POST request. The Content-Type is set to application/octet-stream. See Example bid request for an example.

You must parse this request into an instance of the BidRequest message. BidRequest is defined in realtime-bidding.proto, which can be obtained from the reference data page. You can parse the message using the ParseFromString() method in the generated class for the BidRequest. For example, the following C++ code parses a request given a POST payload in a string: 

string post_payload = /* the payload from the POST request */;
BidRequest bid_request;
if (bid_request.ParseFromString(post_payload)) {
  // Process the request.
}

Once you have the BidRequest you can then work with it as an object, extracting and interpreting the fields you need. For example, in C++:

for (int i = 0; i < bid_request.adslot_size(); ++i) {
  const BidRequest_AdSlot& adslot = bid_request.adslot(i);
  // Decide what to bid on adslot.
}

Some information sent in a BidRequest, such as the Google User ID, language, or geographic location, are not always available. If you have pretargeting ad groups that use information that is unknown for a given impression, then those ad groups will not match. In cases where the missing information doesn't matter for the pretargeting conditions, bid requests are sent with the information omitted.

Information about the pretargeting ad group is available in the MatchingAdData group for each AdSlot. It contains the first matching ad group ID of the pretargeting ad group that prompted Google to send out the bid request, that is, the ad group and campaign that are charged if your response wins the auction for the impression. Under certain circumstances, you need to explicitly specify the billing_id for attribution in the BidResponse.AdSlot, for example, when the BidRequest.AdSlot has more than one matching_ad_data. For more information on the constraints on the contents of the bid, refer to therealtime-bidding.proto.

Dictionary files

The bid request uses identifiers defined in dictionary files, which are available on the reference data page.

Bid URL macros

Optionally, some fields of the BidRequest can be inserted into the URL used in the HTTP POST request. This is useful, for example, if you use a lightweight frontend that load balances over multiple backends using a value from the request. Contact your technical account manager to request support for new macros.

MacroDescription
%%GOOGLE_USER_ID%%

Replaced with the google_user_id from the BidRequest. For example, the bidder URL 

http://google.bidder.com/path?gid=%%GOOGLE_USER_ID%%
will be replaced by something like 
http://google.bidder.com/path?gid=dGhpyBhbiBleGFtGxl
at request time.

If the Google User ID is unknown, the empty string is substituted, with a result similar to

http://google.bidder.com/path?gid=
%%HAS_MOBILE%%

Replaced with 1 or 0 when calling BidRequest's has_mobile().
%%HAS_VIDEO%%

Replaced with 1 (true) or 0 (false) when calling BidRequest's has_video().
%%HOSTED_MATCH_DATA%%

Replaced with the value of the hosted_match_data field from the BidRequest.
%%MOBILE_IS_APP%%

Replaced with 1 (true) or 0 (false) from BidRequest's mobile.is_app field.

Find mobile app ID from transaction URL

Mobile application transactions will report URLs that look like this:

mbappgewtimrzgyytanjyg4888888.com

Use a base-32 decoder to decode the portion of the string in bold (gewtimrzgyytanjyg4888888).

You can use an online decoder, but you'll have to capitalize the letters and replace trailing 8s with = values.

So decoding this value: 

GEWTIMRZGYYTANJYG4======
results in: 
1-429610587
The string 429610587 is the app ID for the iOS app iFunny.

Here's another example. The reported URL is: 

mbappgewtgmjug4ytmmrtgm888888.com
Decoding this value: 
GEWTGMJUG4YTMMRTGM======
results in: 
1-314716233
The result 314716233 is the app ID for the iOS app TextNow.

Find mobile app name from transaction URL

Here's an example of getting the app name. The reported URL is as follows: 

mbappMFUXELTDN5WS42DZOBQWQLTJN4XHG3DJORUGK4Q888.com
Decoding this value: 
MFUXELTDN5WS42DZOBQWQLTJN4XHG3DJORUGK4Q===
results in: 
air.com.hypah.io.slither
The result equates to the Android app slither.io.

Open Bidding fields

Bid requests sent to exchange and network bidders participating in Open Bidding are similar to those of Authorized Buyers participating in standard real-time bidding. Open Bidding customers will receive a small number of additional fields, and a few existing fields may have alternative uses. These include the following:

OpenRTB Authorized Buyers Details
BidRequest.imp[].ext.dfp_ad_unit_code BidRequest.adslot[].dfp_ad_unit_code

Contains the publisher's Ad Manager network code followed by the ad unit hierarchy, separated by forward slashes.

As an example, this would appear with formatting similar to: /1234/cruises/mars.
BidRequest.user.data[].segment[] BidRequest.adslot[].exchange_bidding.key_value[]

Repeated key-value pairs sent from publisher to exchange bidder.

You can determine that the values are key-value pairs sent by the publisher when BidRequest.user.data[].name is set to “Publisher Passed”.

Declare allowed vendors

Technology vendors which provide services such as research, remarketing, and ad serving may play a role in the interaction between buyers and sellers. Only vendors which Google has vetted for participation in Authorized Buyers interactions are allowed.

To understand the BidRequest and create your BidResponse, you need to be aware of the two different possibilities for declaring technology vendors:

  1. Some vendors do not need to be declared; these vendors are listed in the Authorized Buyers Help.
  2. Other vendors can only participate if they are declared in both the BidRequest and BidResponse:
    • In the BidRequest, the allowed_vendor_type field specifies which vendors the seller allows. Vendors that will be sent in the allowed_vendor_type field of the BidRequest are listed in the Vendors.txt dictionary file.
    • In the BidResponse, the vendor_type field specifies which of those allowed vendors the buyer intends to use.

Example bid request

The following examples represent human-readable samples of the Protobuf and JSON requests.

Google

OpenRTB JSON

OpenRTB Protobuf

To convert the bid request into a binary form, like you would get from the POST payload in a real request, you can do the following (in C++). Note, however, that this is not applicable to OpenRTB JSON.

string text_format_example = /* example from above */;
BidRequest bid_request;
if (TextFormat::ParseFromString(text_format_example, &bid_request)) {
  string post_payload;
  if (bid_request.SerializeToString(&post_payload)) {
    // post_payload is a binary serialization of the protocol buffer
  }
}

Remarketing

Authorized Buyers passes a mobile advertising ID in bid requests from a mobile application. The mobile advertising ID can be an iOS IDFA or Android's advertising ID, which is sent through the %%EXTRA_TAG_DATA%% macro in the JavaScript tag managed by Authorized Buyers.

The %%ADVERTISING_IDENTIFIER%% macro allows buyers to receive iOS IDFA or Android's Advertising ID on impression rendering. It returns an encrypted proto buffer MobileAdvertisingId like %%EXTRA_TAG_DATA%%:

message MobileAdvertisingId {
  optional bytes advertising_id = 1;
  optional int32 user_id_type = 2;
}

The user_id_type is one of the values defined in the enum AdxMobileIdType:

enum AdxMobileIdType {
  MOBILE_ID_UNKNOWN = 0,
  IDFA = 1,
  ANDROID_ID = 2,
};

You can create user lists from mobile advertising IDs using advertising IDs you've collected during impression rendering. These user lists can be maintained on your server or on ours. To create user lists on Google's servers you can use our bulk upload facility.

When the mobile advertising ID matches a user list, you can use it to run remarketing.

Real-time feedback

Real-time feedback is available to Authorized Buyers, as well as exchanges and networks using Open Bidding.

Bid response feedback is supported on the subsequent bid request for both AdX Protocol and OpenRTB. For OpenRTB, it is sent in BidRequestExt.

In addition to the default fields sent in Bid Response Feedback, you can also send custom data in the bid response (in either AdX Proto or OpenRTB) using an event_notification_token that is returned in the BidResponse. The event_notification_token is arbitrary data known only to the bidder that might help with debugging, for example: a new targeting ID or bidding ID representing a new tactic, or metadata associated with the creative known only to the bidder. For details, see OpenRTB Extensions Protocol Buffer for RTB and AdX Proto for AdX.

When Authorized Buyers sends a bid request to a bidder, the bidder replies with a BidResponse. If the bidder has real-time feedback enabled, then in a subsequent bid request, Authorized Buyers sends feedback on the response in a BidResponseFeedback message, as shown below:

// Feedback on bids submitted in previous responses. This is only set if
// real-time feedback is enabled for your bidder. Contact your account
// manager if you want to enable real-time feedback.
//
message BidResponseFeedback {
  // The unique id from BidRequest.id
  optional bytes request_id = 1;

  // The index of the BidResponse_Ad if there was more than one. The index
  // starts at zero for the first creative.
  optional int32 creative_index = 2;

  // The status code for the ad. See creative-status-codes.txt in the
  // technical documentation for a list of ids.
  optional int32 creative_status_code = 3;

  // If the bid won the auction, this is the price paid in your account
  // currency. If the bid participated in the auction but was out-bid, this
  // is the CPM that should have been exceeded in order to win. This is not
  // set if the bid was filtered prior to the auction, if the publisher or
  // winning bidder has opted out of price feedback or if your account has
  // opted out of sharing winning prices with other bidders. For first-price
  // auctions, minimum_bid_to_win is populated instead of this field.
  optional int64 cpm_micros = 4;

  // The minimum bid value necessary to have won the auction, in micros of
  // your account currency. If your bid won the auction, this is the second
  // highest bid that was not filtered (including the floor price). If your
  // bid did not win the auction, this is the winning candidate's bid. This
  // field will only be populated if your bid participated in a first-price
  // auction, and will not be populated if your bid was filtered prior to the
  // auction.
  optional int64 minimum_bid_to_win = 7;

  // When a publisher uses an RTB auction and waterfall-based SDK mediation on
  // the same query, the winner of the real-time auction must also compete in
  // a mediation waterfall (which is ordered by price) to win the impression.
  // If the bid participated in the auction and there was no waterfall, the
  // value of this field is 0. If the bid participated in the auction and
  // there was a waterfall, the value of this field is a price representing a
  // sample bid from the eligible mediation networks that were higher than the
  // auction winner, weighted by expected fill rate. This field can be used
  // in conjunction with minimum_bid_to_win to train bidding models. The CPM
  // is in micros of your account currency.
  optional int64 sampled_mediation_cpm_ahead_of_auction_winner = 10;

  // Event notification token that was included in the bid response.
  optional bytes event_notification_token = 5;

  // Buyer creative ID that was included in the bid response.
  optional string buyer_creative_id = 6;
}
repeated BidResponseFeedback bid_response_feedback = 44;

From this message, the first field you should check is bid_response_feedback.creative_status_code; you can find the code meaning in creative-status-codes.txt. Note that if you win the bid, you can opt out from the price feedback. For more information, see How to opt-out.

The real-time feedback includes the bid request ID and one of the following:

Auction outcome Real-time feedback
The buyer didn't submit a bid. Nothing.
The buyer submitted a bid that was filtered out before reaching the auction. The creative status code (see creative-status-codes.txt).
The buyer submitted a bid but lost the auction. The creative status code 79 (outbid in auction).
The buyer submitted a bid that won the auction. The clearing price and creative status code 1.

For an app impression and a creative status code of 53, the app publisher could have been using a mediation waterfall and therefore the winning bid would have competed against other demand in the publisher's passback waterfall chain. Learn how to use sampled_mediation_cpm_ahead_of_auction_winner when bidding.

Sample

The following is a sample of real-time feedback as seen in supported protocols:

Google

OpenRTB JSON

OpenRTB Protobuf

Build a bidding model for first-price auctions

After placing a bid in a first-price auction, you will receive real-time feedback including the minimum_bid_to_win and sampled_mediation_cpm_ahead_of_auction_winner fields if the bid was not filtered from the auction. These signals can be used to inform your bidding logic on how much higher or lower your bid could have been in order to win the impression.

  • minimum_bid_to_win: The minimum bid that could have been placed to win the real-time bidding auction. If you won the auction, this will be the lowest bid you could have placed while still winning. If you lost the auction, this will be the winning bid.
  • sampled_mediation_cpm_ahead_of_auction_winner: If there are other networks in the mediation chain, the value of this field is a price representing a sample bid from one of the eligible mediation networks that were higher than the auction winner, weighted by expected fill rate. This will be set to 0 if none of the networks in the mediation chain are expected to fill, or if the publisher does not use SDK mediation.

How it works

In order to describe the calculations used to determine the possible values for minimum_bid_to_win and sampled_mediation_cpm_ahead_of_auction_winner, we first need to define the following:

  • The following represents the CPMs in the mediation chain in descending order:
    \[C_1, C_2, …, C_n\]
  • The following represents the corresponding fill rates for the CPMs in the mediation chain:
    \[f_1, f_2, …, f_n\]
  • The following is a function used to determine the expected CPM and its probability from mediation chain element \(i\), based on the given fill rate:
    \(X_i = \{C_i\) with probability \(f_i\); \(0\) with probability \(1 - f_i\}\)
  • The final winning mediation chain will be:
    \[\{C_1, C_2, …, C_K, W\}\]
    where \(W\) is the winning bid, and \(C_K > W >= C_{K+1}\)
  • The reserve price, or floor, is denoted as \(F\).
  • The runner-up bid is denoted as \(R\).
Calculations for auction winner
Field Calculation
minimum_bid_to_win
\(max\{F, R, X_{K+1}, …, X_n\}\)
sampled_mediation_cpm_ahead_
of_auction_winner
\(\{C_i\) with probability \(\prod_{j=1}^{i-1}(1-f_j) \cdot f_i \div \prod_{j=1}^{K}(1-f_j)\}\)
For \(1 <= i <= K\).
Calculations for auction loser
Field Calculation
minimum_bid_to_win
\(max\{F, W\}\)
sampled_mediation_cpm_ahead_
of_auction_winner
\(max\{X_1, …, X_K\}\)

Example with a simple mediation chain

Assume a publisher uses both real-time bidding and an SDK mediation chain as follows:

SDK Mediation Chain Expected CPM Fill Rate
Network 1 \(C_1 = $3.00\) \(f_1 = 5\%\)
Network 2 \(C_2 = $2.00\) \(f_2 = 45\%\)
Network 3 \(C_3 = $0.50\) \(f_3 = 80\%\)
Network 4 \(C_4 = $0.10\) \(f_4 = 85\%\)

Assume the following as the result of the RTB auction:

RTB Auction CPM
Auction Winner (W) $1.00
Auction Runner-UP (R) $0.05
Reserve Price / Floor (F) $0
Bid that won the auction

The following is an example of how values and probabilities for minimum_bid_to_win and sampled_mediation_cpm_ahead_of_auction_winner are calculated for a bid that won.

minimum_bid_to_win Probability
\(max(F, R, C_3) = $0.50\) \(f_3 = 80\%\)
\(max(F, R, C_4) = $0.10\) \((1-f_3) \cdot f_4 = 17\%\)
\(max(F, R, 0) = $0.05\) \((1-f_3) \cdot (1-f_4) = 3\%\)
sampled_mediation_cpm_
ahead_of_auction_winner		 Probability
\(C_1 = $3.00\) \(f_1 \div (1-(1-f_1) \cdot (1-f_2)) =~ 10.5\%\)
\(C_2 = $2.00\) \(((1-f_1) \cdot f_2) \div (1-(1-f_1) \cdot (1-f_2)) =~ 89.5\%\)
Bids that lost the auction

The following is an example of how values and probabilities for minimum_bid_to_win and sampled_mediation_cpm_ahead_of_auction_winner are calculated for a bids that lost.

minimum_bid_to_win Probability
\(max(F, W) = $1.00\) \(100\%\)
sampled_mediation_cpm_
ahead_of_auction_winner		 Probability
\(C_1 = $3.00\) \(f_1 = 5\%\)
\(C_2 = $2.00\) \((1-f_1) \cdot f_2 =~ 42.8\%\)
\(0\) \((1-f_1) \cdot (1-f_2) =~ 52.2\%\)

Bid flattening

Bid flattening describes the processing of a single complex BidRequest into multiple bid requests that are sent to your application. Because they retain identical IDs (BidRequest.google_query_id in the Authorized Buyers RTB Protocol or BidRequestExt.google_query_id in the OpenRTB protocol) you can determine which bid requests are correlated after flattening.

Ad formats

Some ad opportunities can accept multiple formats. With bid flattening, each format is sent in a distinct bid request where attributes such as eligible billing IDs are relevant to the format specified in the request.

Bid requests containing the following formats will be flattened into distinct bid requests:

  • Banner
  • Video
  • Audio
  • Native

Ad format flattening example

Below is an example showing a simplified OpenRTB JSON bid request without ad format flattening in comparison to an equivalent set of flatenned requests:

Pre-flatten

Post-flatten

Deals

An ad opportunity for a given bidder can be applicable to various deal types, in addition to the open auction. With bid flattening for deals, one bid request will be sent for the open auction, and one for each type of fixed-price deal. In practice, ad constraints can differ between auctions and fixed-price deal types, for example, for a given video ad opportunity that is available to both the open auction and a fixed-price deal, a bidder will receive distinct bid requests for each where constraints such as maximum ad duration and whether skippable ads are allowed can differ. As a result, flattening applied to the ad opportunity lets you more easily discern the ad constraints for the open auction and the fixed-price deal.

Max skippable video duration

Google's protocol and OpenRTB implementation support the following fields for video duration and skipability:

Duration Skippable duration Skipability
Google protocol max_ad_duration skippable_max_ad_duration video_ad_skippable
OpenRTB maxduration n/a skip

This means that while the Google protocol can have a granular skippable and non-skippable duration, the OpenRTB implementation only has a single maximum duration value.

Prior to bid flattening, OpenRTB's maxduration would be set to the lower of the Google protocol's max_ad_duration and skippable_max_ad_duration fields. This behavior has now changed to sending two separate bid requests when these values differ: one representing the maxduration for skippable and the other for non-skippable opportunities.

The following examples show how a Google protocol request translates to OpenRTB before and after bid flattening. The equivalent Google protocol request has a max_ad_duration of 15 and a skippable_max_ad_duration of 60.

Example max_ad_duration skip (true OR false)
Original request without flattening 15 true
Flattened request #1: Non-skippable 15 false
Flattened request #2: Skippable 60 true

Skippable video duration bid request flattening will only take place when these conditions are met:

  • The request allows video.
  • Both skip and no-skip videos are allowed, and the two respective max durations differ in value.
  • This request is Private Auction or Open Auction-eligible.
  • The bidder account has active OpenRTB endpoints.

You can opt out from this type of flattening by contacting your technical account manager.

Video pods

Bid requests for a video pod with multiple ad opportunities are flattened, such that each bid request is for an individual ad opportunity from that pod. This enables you to bid on multiple ad opportunities for a given pod.

Open Measurement

Open Measurement lets you specify third-party vendors that provide independent measurement and verification services for ads served to mobile app environments.

You can determine whether a publisher supports Open Measurement in the bid request by checking whether the ad opportunity excludes the OmsdkType: OMSDK 1.0 attribute found in Publisher-excludable creative attributes. For the Authorized Buyers protocol, this would be found under BidRequest.adslot[].excluded_attribute. For the OpenRTB protocol, this would be found under the battr attribute for Banner or Video, depending on the format.

For more information on how to interpret bid requests containing Open Measurement signals, refer to the Open Measurement SDK Help Center article.

Sample bid requests

The following sections show sample bid requests for different ad types.

App banner

Google

OpenRTB JSON

OpenRTB Protobuf

App interstitial

Google

OpenRTB JSON

OpenRTB Protobuf

App interstitial video

Google

OpenRTB JSON

OpenRTB Protobuf

App native

Google

OpenRTB JSON

OpenRTB Protobuf

Web video

Google

OpenRTB JSON

OpenRTB Protobuf

Mobile web banner for exchange bidder

OpenRTB Protobuf