Prebid Server - Module Analytics Tags Conventions

• Overview
• Analytics Tag Conventions
• Designing Analytics Tags
• Document the Analytics Tags Produced
• Parsing the Invocation Context
• Further Reading

Overview

Analytics Tags (aka ‘ATags’) are a log mechanism provided by PBS-core to allow modules to inform analytics adapters about what happened in the request. Use of the Analytics Tag structure is completely optional. It’s meant to be used when there are application or reporting reasons for sharing module results. See the Prebid Server module overview for background information.

This document defines a convention aimed at allowing module developers to create consistent ATags that can be easily read by analytics adapters.

1. Modules may return a data structure containing ATags to PBS-Core.
2. PBS-Core adds any ATags to the ‘invocation context’ data structure.
3. Analytics modules have access to the invocation context.

Analytics Tag Conventions

The general idea is that ATags are a list of module-specific “activities” that have these attributes:

• activity name: should be unique within the context of this module. e.g. ‘enrich-request’
• an overall status
• an array of specific results within the activity
• within the results:
  • status of each result
  • scope of the result
  • module-specific values for the result

Here’s an example from the ORTB2 Blocking module:

[{
   // scenario: response from bidderA blocked on badv for imp=1
   activities: [{
    name: "enforce-blocking",
    status: "success", // no errors from the module
    results: [{
        status: "success-block",
        values: { // these are module-specific details about the result
                  "attributes": ["badv"],
                  "adomain": ["bad.com"]
        },
        appliedto: {
          "bidder": "bidderA",
           impids: ["1"]
        }
    },{
        status: "success-allow",
        // no details needed when the response isn't blocked
        appliedto: {
          "bidder": "bidderA",
          "impids": ["2","3","4"]
        }
    }]
 }]

The following table contains the field conventions.

Designing Analytics Tags

ATags are for reporting. Start by considering what the module’s doing that consumers might want to display. Each processing stage the module operates in may be reported as a separate activity, or perhaps everything the module does is lumped as one activity.

Once the activities are defined, determine what reportable metric would be useful. Examples:

• percentage of impressions enriched
• percentage of imps where a specific bidder was removed due to optimization
• A/B testing data
• percentage and value of bid responses that were rejected due to creative validations

Case study: for the ORTB2 Blocking module, the requirement was to be able to report on what percentage of responses from each bidder were being thrown away due to blocking rules. This could have been done by defining a separate ‘activity’ for each of the 4 types of enforcement, but it was decided to just have one kind of activity (‘enforce-blocking’) and get the specific details as part of the ‘value’. There was no requirement to report on the outbound part of what the module does, so no ATags are created for that part of the functionality.

Once you know what reports are desired, figure out which activity ‘results’ are needed to generate those numbers.

For the ORTB2 Blocking module, the numbers needed are how often a given bidder has a response compared to how often their responses are rejected. So overall the block rate is: successBlock divided by (successBlock + successAllow).

Document the Analytics Tags Produced

Be sure to detail the results in your module documentation so that analytics adapters are aware of what they can look for.

Let them know:

• which activities your module supports
• what kind of results to expect
• whether the results objects have module-specific values

Parsing the Invocation Context

If you’re an analytics adapter, you will be given the entire PBS ‘invocation context’, which contains a wealth of information about the auction.

In short, to get analytics tags, you’ll need to parse this data structure:

• Loop through stages[]
  • If the stage has relevant info, loop through outcomes[]
    • Outcomes.entity is a label describing the contents
    • Loop through groups[]
      • Loop through invocationresults[]
        • if hookid.modulecode is relevant for your analytics, grab analyticstags

Here’s an example of the data structured as JSON, though the details of the actual object will differ in PBS-Java and PBS-Go.

          "stages": [
            {
              "stage": "raw-auction-request",
              "outcomes": [
                {
                  "entity": "bidrequest",
                  "executiontimemillis": 246,
                  "groups": [
                    {
                      "executiontimemillis": 190,
                      "invocationresults": [
                        {
                          "hookid": {
                            "modulecode": "MODULE1",
                            "hookimplcode": "HOOK_A"
                          },
                          "executiontimemillis": 123,
                          "status": "success",
                          "action": "update",
                          "debugmessages": [
                            "debug message"
                          ],
                         "analyticstags": {
                            "activities": [
                              {
                                "name": "device-id",
                                "status": "success",
                                "results": [
                                  {
                                    "status": "success",
                                    "values": {
                                      "some-field": "some-value"
                                    },
                                    "appliedto": {
                                      "impids": [ "impId1" ]
                                    }
                                  }
                                ]
                              }
                            ]
                          }
                        },
                        {
                          "hookid": {
                            "modulecode": "MODULE1",
                            "hookimplcode": "HOOK_B"
                          },
                          "status": "success",
                          "message": "inventory has no value",
                          "action": "reject"
                        }
                      ]
                    }
                  ]
                }
              ]
            },
            {
              "stage": "bidder-request",
              "outcomes": [
                {
                  "entity": "pubmatic", // entity is biddercode for some stages
                  "executiontimemillis": 246,
                  "groups": [
                    "invocationresults": {
                      "hookid": { ... }
                      "analyticstags": [{
            ...
                      ]
                    }
                  ]
                },
                {
                  "entity": "adform",
                  "executiontimemillis": 246,
                  "groups": [...]
                }
              ]
            }
          ]

See the implementation guide for your platform for specific syntax.

Further Reading

• PBS Module Overview
• PBS Module Implementation