Experimentos en la misma campaña

Los experimentos dentro de la campaña se utilizan para probar una función específica dentro de una sola campaña. A diferencia de los experimentos administrados por el sistema, en los que el tráfico se divide entre las campañas de control y de tratamiento, los experimentos dentro de la campaña dividen el tráfico dentro de la campaña, según si la función está habilitada o no.

Este flujo de trabajo es compatible con los siguientes valores de ExperimentType:

ADOPT_AI_MAX
Prueba el impacto de las funciones de IA Max en las campañas de Búsqueda.
ADOPT_BROAD_MATCH_KEYWORDS
Prueba el impacto de las palabras clave de concordancia amplia en las campañas de Búsqueda.
PMAX_TEXT_CUSTOMIZATION_FINAL_URL_EXPANSION
Se utiliza específicamente para probar las funciones de las campañas de máximo rendimiento: personalización de texto y expansión de la URL final. Permite que la IA de Google envíe tráfico a páginas de destino pertinentes y genere recursos de texto para correlacionar mejor las búsquedas.

Configuración

  1. Define el Experiment (con status establecido en ENABLED) y proporciona un tipo de experimento, un ExperimentArm de control y un ExperimentArm de tratamiento:
    • Para ADOPT_AI_MAX y ADOPT_BROAD_MATCH_KEYWORDS, tanto el grupo de control como el grupo de tratamiento deben hacer referencia a la misma campaña en campaigns.
    • Para PMAX_TEXT_CUSTOMIZATION_FINAL_URL_EXPANSION, solo el grupo de tratamiento especifica la campaña de máximo rendimiento en campaigns (deja campaigns sin establecer en el grupo de control).

  2. Para ADOPT_AI_MAX, crea una actualización de CampaignOperation con una máscara de campo para habilitar ai_max_setting.enable_ai_max y configurar asset_automation_settings.

  3. Envía una solicitud GoogleAdsService.Mutate que incluya operaciones de mutación para crear el experimento y los grupos experimentales, y (para ADOPT_AI_MAX) habilitar la función de prueba en la campaña.

Para ADOPT_AI_MAX y ADOPT_BROAD_MATCH_KEYWORDS, el tráfico se debe dividir de manera uniforme (50/50) dentro de la campaña, de modo que el 50% del tráfico esté expuesto a la función habilitada (el grupo de tratamiento) y el 50% no (el grupo de control). PMAX_TEXT_CUSTOMIZATION_FINAL_URL_EXPANSION admite divisiones del tráfico personalizadas que suman 100.

Java

// Create the experiment resource name using a temporary ID.
String experimentResourceName = ResourceNames.experiment(customerId, -1L);

// Create the experiment.
Experiment experiment =
    Experiment.newBuilder()
        .setResourceName(experimentResourceName)
        .setName("ADOPT_AI_MAX Experiment #" + UUID.randomUUID())
        .setType(ExperimentType.ADOPT_AI_MAX)
        .build();
MutateOperation experimentOperation =
    MutateOperation.newBuilder()
        .setExperimentOperation(ExperimentOperation.newBuilder().setCreate(experiment).build())
        .build();

// Create the control arm. Both arms in an intra-campaign experiment reference the same base
// campaign.
ExperimentArm controlArm =
    ExperimentArm.newBuilder()
        .setExperiment(experimentResourceName)
        .setName("Control Arm")
        .setControl(true)
        .setTrafficSplit(50)
        .addCampaigns(ResourceNames.campaign(customerId, campaignId))
        .build();
MutateOperation controlArmOperation =
    MutateOperation.newBuilder()
        .setExperimentArmOperation(
            ExperimentArmOperation.newBuilder().setCreate(controlArm).build())
        .build();

// Create the treatment arm.
ExperimentArm treatmentArm =
    ExperimentArm.newBuilder()
        .setExperiment(experimentResourceName)
        .setName("Treatment Arm")
        .setControl(false)
        .setTrafficSplit(50)
        .addCampaigns(ResourceNames.campaign(customerId, campaignId))
        .build();
MutateOperation treatmentArmOperation =
    MutateOperation.newBuilder()
        .setExperimentArmOperation(
            ExperimentArmOperation.newBuilder().setCreate(treatmentArm).build())
        .build();

// Create a campaign operation with an update mask to enable AI Max and configure asset
// automation settings.
// Note: For intra-campaign experiments, these settings are applied to the base campaign but are
// only active for the treatment traffic split.
Campaign campaign =
    Campaign.newBuilder()
        .setResourceName(ResourceNames.campaign(customerId, campaignId))
        .setAiMaxSetting(AiMaxSetting.newBuilder().setEnableAiMax(true).build())
        .addAssetAutomationSettings(
            AssetAutomationSetting.newBuilder()
                .setAssetAutomationType(AssetAutomationType.TEXT_ASSET_AUTOMATION)
                .setAssetAutomationStatus(AssetAutomationStatus.OPTED_IN)
                .build())
        .addAssetAutomationSettings(
            AssetAutomationSetting.newBuilder()
                .setAssetAutomationType(
                    AssetAutomationType.FINAL_URL_EXPANSION_TEXT_ASSET_AUTOMATION)
                .setAssetAutomationStatus(AssetAutomationStatus.OPTED_IN)
                .build())
        .build();

CampaignOperation campaignOp =
    CampaignOperation.newBuilder()
        .setUpdate(campaign)
        .setUpdateMask(FieldMasks.allSetFieldsOf(campaign))
        .build();
MutateOperation campaignMutateOperation =
    MutateOperation.newBuilder().setCampaignOperation(campaignOp).build();

// Send all mutate operations in a single Mutate request.
List<MutateOperation> mutateOperations =
    ImmutableList.of(
        experimentOperation,
        controlArmOperation,
        treatmentArmOperation,
        campaignMutateOperation);

try (GoogleAdsServiceClient googleAdsServiceClient =
    googleAdsClient.getLatestVersion().createGoogleAdsServiceClient()) {

  MutateGoogleAdsRequest request =
      MutateGoogleAdsRequest.newBuilder()
          .setCustomerId(Long.toString(customerId))
          .addAllMutateOperations(mutateOperations)
          .build();

  MutateGoogleAdsResponse response = googleAdsServiceClient.mutate(request);

C#

// Create the experiment resource name using a temporary ID.
string experimentResourceName = ResourceNames.Experiment(customerId, -1);

// Create the experiment.
MutateOperation experimentOperation = new MutateOperation()
{
    ExperimentOperation = new ExperimentOperation()
    {
        Create = new Experiment()
        {
            ResourceName = experimentResourceName,
            Name = $"ADOPT_AI_MAX Experiment #{ExampleUtilities.GetRandomString()}",
            Type = ExperimentType.AdoptAiMax
        }
    }
};

// Create the control arm. Both arms in an intra-campaign experiment
// reference the same base campaign.
MutateOperation controlArmOperation = new MutateOperation()
{
    ExperimentArmOperation = new ExperimentArmOperation()
    {
        Create = new ExperimentArm()
        {
            Experiment = experimentResourceName,
            Name = "Control Arm",
            Control = true,
            TrafficSplit = 50,
            Campaigns = { ResourceNames.Campaign(customerId, campaignId) }
        }
    }
};

// Create the treatment arm.
MutateOperation treatmentArmOperation = new MutateOperation()
{
    ExperimentArmOperation = new ExperimentArmOperation()
    {
        Create = new ExperimentArm()
        {
            Experiment = experimentResourceName,
            Name = "Treatment Arm",
            Control = false,
            TrafficSplit = 50,
            Campaigns = { ResourceNames.Campaign(customerId, campaignId) }
        }
    }
};

// Create a campaign operation with an update mask to enable AI Max and
// configure asset automation settings.
// Note: For intra-campaign experiments, these settings are applied to the
// base campaign but are only active for the treatment traffic split.
Campaign campaign = new Campaign()
{
    ResourceName = ResourceNames.Campaign(customerId, campaignId),
    AiMaxSetting = new Campaign.Types.AiMaxSetting { EnableAiMax = true }
};

campaign.AssetAutomationSettings.Add(new Campaign.Types.AssetAutomationSetting
{
    AssetAutomationType = AssetAutomationType.TextAssetAutomation,
    AssetAutomationStatus = AssetAutomationStatus.OptedIn
});

campaign.AssetAutomationSettings.Add(new Campaign.Types.AssetAutomationSetting
{
    AssetAutomationType = AssetAutomationType.FinalUrlExpansionTextAssetAutomation,
    AssetAutomationStatus = AssetAutomationStatus.OptedIn
});

MutateOperation campaignOperation = new MutateOperation()
{
    CampaignOperation = new CampaignOperation()
    {
        Update = campaign,
        UpdateMask = FieldMasks.AllSetFieldsOf(campaign)
    }
};

// Send all mutate operations in a single Mutate request.
List<MutateOperation> mutateOperations = new List<MutateOperation>
{
    experimentOperation,
    controlArmOperation,
    treatmentArmOperation,
    campaignOperation
};

MutateGoogleAdsResponse response = googleAdsService.Mutate(
    customerId.ToString(), mutateOperations);

PHP

This example is not yet available in PHP; you can take a look at the other languages.

Python

# Create the experiment resource name using a temporary ID.
experiment_resource_name = googleads_service.experiment_path(
    customer_id, "-1"
)

# Create the experiment.
experiment_operation = client.get_type("MutateOperation")
experiment = experiment_operation.experiment_operation.create
experiment.resource_name = experiment_resource_name
experiment.name = f"ADOPT_AI_MAX Experiment #{uuid4()}"
experiment.type_ = client.enums.ExperimentTypeEnum.ADOPT_AI_MAX

# Create the control arm. Both arms in an intra-campaign experiment
# reference the same base campaign.
control_arm_operation = client.get_type("MutateOperation")
control_arm = control_arm_operation.experiment_arm_operation.create
control_arm.experiment = experiment_resource_name
control_arm.name = "Control Arm"
control_arm.control = True
control_arm.traffic_split = 50
control_arm.campaigns.append(
    googleads_service.campaign_path(customer_id, campaign_id)
)

# Create the treatment arm.
treatment_arm_operation = client.get_type("MutateOperation")
treatment_arm = treatment_arm_operation.experiment_arm_operation.create
treatment_arm.experiment = experiment_resource_name
treatment_arm.name = "Treatment Arm"
treatment_arm.control = False
treatment_arm.traffic_split = 50
treatment_arm.campaigns.append(
    googleads_service.campaign_path(customer_id, campaign_id)
)

# Create a campaign operation with an update mask to enable AI Max and
# configure asset automation settings.
# Note: For intra-campaign experiments, these settings are applied to the
# base campaign but are only active for the treatment traffic split.
campaign_operation = client.get_type("MutateOperation")
campaign = campaign_operation.campaign_operation.update
campaign.resource_name = googleads_service.campaign_path(
    customer_id, campaign_id
)
campaign.ai_max_setting.enable_ai_max = True

for asset_automation_type_enum in [
    client.enums.AssetAutomationTypeEnum.TEXT_ASSET_AUTOMATION,
    client.enums.AssetAutomationTypeEnum.FINAL_URL_EXPANSION_TEXT_ASSET_AUTOMATION,
]:
    asset_automation_setting = client.get_type(
        "Campaign"
    ).AssetAutomationSetting()
    asset_automation_setting.asset_automation_type = (
        asset_automation_type_enum
    )
    asset_automation_setting.asset_automation_status = (
        client.enums.AssetAutomationStatusEnum.OPTED_IN
    )
    campaign.asset_automation_settings.append(asset_automation_setting)

client.copy_from(
    campaign_operation.campaign_operation.update_mask,
    protobuf_helpers.field_mask(None, campaign._pb),
)

# Send all mutate operations in a single Mutate request.
mutate_operations = [
    experiment_operation,
    control_arm_operation,
    treatment_arm_operation,
    campaign_operation,
]

response = googleads_service.mutate(
    customer_id=customer_id,
    mutate_operations=mutate_operations,
)

Ruby

# Create the experiment resource name using a temporary ID.
experiment_resource_name = client.path.experiment(customer_id, "-1")

# Create the experiment.
experiment_operation = client.operation.mutate do |m|
  m.experiment_operation = client.operation.create_resource.experiment do |e|
    e.resource_name = experiment_resource_name
    e.name = "ADOPT_AI_MAX Experiment #{(Time.new.to_f * 1000).to_i}"
    e.type = :ADOPT_AI_MAX
  end
end

# Create the control arm. Both arms in an intra-campaign experiment
# reference the same base campaign.
campaign_resource_name = client.path.campaign(customer_id, campaign_id)
control_arm_operation = client.operation.mutate do |m|
  m.experiment_arm_operation = client.operation.create_resource.experiment_arm do |arm|
    arm.experiment = experiment_resource_name
    arm.name = "Control Arm"
    arm.control = true
    arm.traffic_split = 50
    arm.campaigns << campaign_resource_name
  end
end

# Create the treatment arm.
treatment_arm_operation = client.operation.mutate do |m|
  m.experiment_arm_operation = client.operation.create_resource.experiment_arm do |arm|
    arm.experiment = experiment_resource_name
    arm.name = "Treatment Arm"
    arm.control = false
    arm.traffic_split = 50
    arm.campaigns << campaign_resource_name
  end
end

# Create a campaign operation with an update mask to enable AI Max and
# configure asset automation settings.
# Note: For intra-campaign experiments, these settings are applied to the
# base campaign but are only active for the treatment traffic split.
campaign_operation = client.operation.mutate do |m|
  m.campaign_operation = client.operation.update_resource.campaign(campaign_resource_name) do |c|
    c.ai_max_setting = client.resource.ai_max_setting do |aims|
      aims.enable_ai_max = true
    end

    [:TEXT_ASSET_AUTOMATION, :FINAL_URL_EXPANSION_TEXT_ASSET_AUTOMATION].each do |type|
      c.asset_automation_settings << client.resource.asset_automation_setting do |setting|
        setting.asset_automation_type = type
        setting.asset_automation_status = :OPTED_IN
      end
    end
  end
end

# Send all mutate operations in a single Mutate request.
mutate_operations = [
  experiment_operation,
  control_arm_operation,
  treatment_arm_operation,
  campaign_operation,
]

response = client.service.google_ads.mutate(
  customer_id: customer_id,
  mutate_operations: mutate_operations,
)

Perl

This example is not yet available in Perl; you can take a look at the other languages.

curl

Informe sobre el experimento

Debido a que el tráfico de control y de tratamiento se mezclan en una sola campaña, debes utilizar los informes de experimentos directos para comparar las métricas entre los grupos de control y de tratamiento. Los informes estándar a nivel de la campaña solo muestran métricas agregadas para toda la campaña y no pueden distinguir entre los dos grupos.

La siguiente consulta de GAQL recupera las estadísticas de clics de un experimento de ADOPT_AI_MAX dentro de la campaña durante los últimos 30 días:

SELECT
  experiment.resource_name,
  experiment.name,
  metrics.clicks,
  metrics.control_clicks,
  metrics.clicks_point_estimate,
  metrics.clicks_p_value
FROM experiment
WHERE segments.date DURING LAST_30_DAYS
  AND experiment.type = 'ADOPT_AI_MAX'

Promociona, gradúa o finaliza el experimento

Después de evaluar los resultados, puedes completar el experimento con ExperimentService:

  • Finalizar: Si no estás conforme con los resultados, usa EndExperiment. La función se inhabilita y la campaña vuelve a publicar todo el tráfico sin la función experimental. Esta es una operación síncrona.
  • Promocionar (ADOPT_AI_MAX y ADOPT_BROAD_MATCH_KEYWORDS): Si estás conforme con los resultados de un experimento de campaña de Búsqueda, usa PromoteExperiment. Esto aplica el cambio experimental como el nuevo estado permanente de la campaña. Esta es una operación asíncrona. Consulta Errores asíncronos para obtener más detalles.
  • Graduar (PMAX_TEXT_CUSTOMIZATION_FINAL_URL_EXPANSION): PromoteExperiment no se admite para PMAX_TEXT_CUSTOMIZATION_FINAL_URL_EXPANSION. En su lugar, usa GraduateExperiment para aplicar la personalización de texto y la configuración de expansión de la URL final del grupo de tratamiento en la campaña de máximo rendimiento existente. Incluye una entrada CampaignBudgetMapping en campaign_budget_mappings que especifique la campaña del grupo de tratamiento (experiment_campaign) y su presupuesto (campaign_budget). Esta es una operación síncrona. (GraduateExperiment no se admite para ADOPT_AI_MAX ni ADOPT_BROAD_MATCH_KEYWORDS porque no hay una campaña de tratamiento independiente para graduar).
Anterior
Experimentos administrados por el sistema
Siguiente
Experimentos de optimización de recursos