limata3

/* We will only be offering code samples for the original attribution models (we won;t have the to_lead and non_direct variations in here */
/* Note: We are using WITH statements in here, but in a real life example you would want to create actual tables for each step instead having each step as a WITH statement */
/* Note: This is based on BigQuery SQL dialect */

/* This starts with our customer touchpoints table - filtered by Customer XYZ only though */
WITH master_customer_touchpoints_and_ranking AS (
    SELECT 
        /* Master Customer Touchpoints Table Original Fields */
        customer_id
        , device_id

limata3

/* First Touch Attribution */
SELECT
    customer_id
    , order_id
    , touchpoint_id
    , CAST(MAX(total_net_orders) OVER (PARTITION BY buying_period_id) AS FLOAT64) AS first_touchpoint_credit
FROM master_ranked_customer_tocuhpoints
WHERE
    touchpoints_customer_order_rank_asc = 1 /* This ensures all credit goes to the first touchpoint */
;    

limata3

SELECT
    customer_id
    , touchpoint_id
    , lead_id
    , order_id
    , touchpoint_channel
    , touchpoint_campaign
    , touchpoint_datetime
    , RANK() OVER (PARTITION BY customer_id ORDER BY touchpoint_datetime ASC, rand() ) AS touchpoints_customer_rank_asc /* touchpoint -> customer */
    , RANK() OVER (PARTITION BY customer_id ORDER BY touchpoint_datetime DESC, rand() ) AS touchpoints_customer_rank_des /* touchpoint -> customer */    

limata3

limata3

limata3

import pandas as pd
import numpy as np


raw_test_data = pd.read_csv('medium_ppc_inc_evaluate_dataset.csv')
# You can access the csv above in this link: https://drive.google.com/open?id=1IOuneJr-QFDYGsJPRjQ8ra-5C4VczeOW

raw_test_data['diff_conversions'] = raw_test_data['test_conversions'] - raw_test_data['ctl_conversions']

N = len(raw_test_data)

limata3

def create_experiment(experiment_name, base_campaign_id, draft_id, split_percent=50, split_type='COOKIE'):
    adwords_client = connect_to_adwords_api()
    trial_service = adwords_client.GetService('TrialService', version='v201809')
    trial = {
        'draftId': draft_id,
        'baseCampaignId': base_campaign_id,
        'name': experiment_name,
        'trafficSplitPercent': split_percent,
        'trafficSplitType': split_type
    }

limata3

import pandas as pd


def get_all_ad_groups_from_campaign(campaign_id):
    adwords_client = connect_to_adwords_api()
    ad_group_service = adwords_client.GetService('AdGroupService', version='v201809')
    selector = {
        'fields': ['Id', 'Name', 'Status','TargetCpa'],
        'paging': {
            'startIndex': str(0),

limata3

with control_campaigns as (
select
    'Control' as measurement_group
  , campaign_name as base_campaign
  , date
  , cost
  , conversions
  , conversion_value
from
  fivetran.adwords.google_campaign_performance

limata3

def change_draft_target_cpa(original_campaign_tcpa, original_campaign_name, draft_campaign_id):
    # Create Portfolio for new campaign
    draft_target_cpa = round(original_campaign_tcpa * (1-0.30), 2)
    draft_portfolio_name = '{0}-Test_XYZ_Draft_Portfolio'.format(original_campaign_name)
    adwords_client = connect_to_adwords_api()
    bidding_strategy_service = adwords_client.GetService('BiddingStrategyService', version='v201809')
    portfolio_config = {
        'name': draft_portfolio_name,
        'biddingScheme': {
            'xsi_type': 'TargetCpaBiddingScheme',