This module allows publishers the chance to ramp-up on Prebid Server, testing the impact of server-side header bidding before fully switching. Without this module, the s2sConfig settings direct all requests to Prebid Server. By including this module in the PrebidJS build, there are additional options for controlling how requests are sent:
The package is built by specifying the s2sTesting module on the build command. For example:
gulp build --modules= ... , prebidServerBidAdapter,s2sTesting
The expectation is the Server-to-Server Testing module will used for a limited period. During the test period, the publisher analyzes metrics from various places to gauge the impact of moving header bidding to the server. After the test period is complete, regardless of outcome, A/B testing would be turned off and this module no longer included in the PrebidJS build.
With the Server-to-Server Testing module, the following enhancements are provided:
New ‘bidderControl’ options in s2sConfig. E.g.
pbjs.setConfig({
s2sConfig: {
bidders: [ "bidderA", "bidderB" ],
enabled: true,
defaultVendor: 'vendor1',
testing: true,
testServerOnly: false,
bidderControl: {
"bidderA": {
bidSource: {server:10, client:90},
includeSourceKvp: true
},
"bidderB": {
bidSource: {server: 25, client:75},
includeSourceKvp: true
}
}
},
// needed since PBJS 5.0 to generate the hb_source key
targetingControls: {
addTargetingKeys: ['SOURCE']
}
});
The values in the bidSource object are weights. These examples show them as percentages that add to 100, but they could also be “{server:1, client:9}” and “{server:1, client:3}” respectively.
The testing: true attribute is required to enable the bidderControl and
bidSource features. This shouldn’t be confused with the enabled: true flag
which enables the entire server-to-server feature.
When ‘includeSourceKvp’ is specified in s2sConfig and the appropriate addTargetingKeys targetingControl is added, the system will log an additional Key Value Pair (KVP) to the ad server. This will allow reporting to confirm the ratio of client-vs-server administered requests, as well as more advanced reporting.
hb_source_BIDDER=client
OR
hb_source_BIDDER=s2s
There’s also a new bidSource option in AdUnits that overrides the global bidSource. E.g.
AdUnit={
[...]
bids=[{
bidder: "bidderA",
bidSource: {client:50, server:50} // precedence over s2sConfig.bidderControl
[...]
}]
}
Determining the success of the test is up to each publisher, but keep in mind the various data sources you may have access to:
hb_source key value pair can be used to confirm some elements on the performance.As a Publisher, I want to ramp up server traffic across my site to monitor revenue impact before going entirely to server-side header bidding. I don’t want to modify AdUnits because that’s time consuming in the CMS.
Example S2S Config defining that 10% of Rubicon requests and 100% of AppNexus requests go through the server:
pbjs.setConfig({
s2sConfig: {
account: "PREBID-SERVER-ACCOUNT",
bidders: [ "bidderA", "bidderB" ],
defaultVendor: 'vendor1',
enabled: true,
testing: true,
bidderControl: {
"bidderA": {
bidSource: {client:90, server:10},
includeSourceKvp: true
}
}
},
targetingControls: {
addTargetingKeys: ['SOURCE']
}
});
The additional hb_source_bidderA KVP will be sent to the ad server for additional reporting.
As a Publisher, I want to get metrics on the difference between the server approach and the client approach so we can gauge the impact of Prebid Server, but we want to do this on a small number of AdUnits.
Example S2S Config defining that the client route is the default path for bidderA:
pbjs.setConfig({
s2sConfig: {
account: "PREBID-SERVER-ACCOUNT",
bidders: [ "bidderA", "bidderB" ],
defaultVendor: 'vendor1',
enabled: true,
testing: true,
bidderControl: {
bidderA: {
bidSource: {client:100},
includeSourceKvp: true
}
}
},
targetingControls: {
addTargetingKeys: ['SOURCE']
}
});
And then changes to override one particular AdUnit for server testing:
AdUnit={
[...]
bids=[{
bidder: "bidderA",
bidSource: {server:25, client:75}
[...]
}]
}
Requests will go to bidderA on the server path 25% of the time and the client path the rest of the time. The additional hb_source_bidderA KVP will be sent to the ad server for additional reporting.
As a Publisher, I want to get metrics on the difference between a ‘pure server’
approach and a mixed client-server approach so we can gauge the impact of Prebid Server running alone, without any client requests. I’ll use the testServerOnly: true flag to suppress all client requests whenever the ‘A/B test group’ results in a server request.
Using the testServerOnly flag means that all client requests will be suppressed (those requests will not be made) whenever any bid requests from the ‘A/B test group’ result in a ‘server’ bid request. The ‘A/B test group’ includes any requests whose source is controled by ‘s2sConfig.bidderControl’ or ‘bidSource’ at the adUnit level. This may give a clearer picture of how s2s performs without interference from client bid requests.
For best results, all bidders/bids in the ‘A/B test group’ should be configured with the same client/server allocation. Because use of this flag will result in turning off client bids a certain percentage of the time, it could negatively affect revenue, and should be used with caution. Thus it should only be used when ‘server’ is allocated a small percentage (i.e. <= 5%) of bid requests.
Example S2S Config defining that 5% of the time all bid requests will go ‘server’ and 95% of the time a mix of ‘server’ and ‘client’:
pbjs.setConfig({
s2sConfig: {
account: "PREBID-SERVER-ACCOUNT",
bidders: [ "bidderA", "bidderB" ],
defaultVendor: 'vendor1',
enabled: true,
testing: true,
testServerOnly: true,
bidderControl: {
bidderA: {
bidSource: {client:95, server 5},
includeSourceKvp: true
},
bidderB: {
bidSource: {client:95, server 5},
includeSourceKvp: true
}
}
},
targetingControls: {
addTargetingKeys: ['SOURCE']
}
});
AdUnit={
[...]
bids=[{
bidder: "bidderA",
[...]
},{
bidder: "bidderB",
[...]
},{
bidder: "bidderC",
[...]
}]
}
5% of the time bidderA and bidderB will use s2s bid requests while bidderC does not bid, and the other 95% of the time all bidders use client bid requests.
Addtional details:
s2sConfig.bidderControl or AdUnit.bids[].bidSource, then it will not affect testServerOnly. i.e. It’s going on the server path will not exclude client side adapters.As a Publisher, I’d like to run tests on one part or my site per one of the
other use cases above. I’ll use the test KVP to confirm relative responses,
so would like to have the hb_source test KVP coming in even on pages where
the server test isn’t running.
Example S2S Config defining that client is the default path for bidderA:
pbjs.setConfig({
s2sConfig: {
...
enabled: false,
testing: true,
bidderControl: {
bidderA: {
bidSource: {client:100},
includeSourceKvp: true
}
}
},
targetingControls: {
addTargetingKeys: ['SOURCE']
}
});
The test KVP hb_source_bidderA on this page will always sent with the value “client”.