Merge pull request #21 from powersupplyhq/create_google_display_visua…

…lizations

Create google display visualizations

AdWordsUtilities/NAMESPACE

@@ -9,6 +9,7 @@ export(clean_raw_campaign_data)
 export(clean_raw_keyword_data)
 export(click_through_rate_over_time)
 export(contribution_per_click_report)
+export(desktop_gdn_performance_over_time)
 export(desktop_performance_over_time)
 export(impression_share_over_time)
 export(keyword_performance_data)
@@ -17,6 +18,7 @@ export(keywords_campaign_performance_over_time)
 export(keywords_performance_over_time)
 export(keywords_report)
 export(keywords_with_earnings)
+export(mobile_gdn_performance_over_time)
 export(mobile_performance_over_time)
 export(overall_performance_over_time)
 export(raw_campaign_performance_data)

AdWordsUtilities/R/contribution_per_click_report.r

@@ -29,10 +29,10 @@ contribution_per_click_report <- function(keywords_elog,
   }
 
   data <- keywords_elog %>% 
-            group_by(keyword,campaign_name) %>% 
+            group_by(campaign_name,keyword,match_type) %>% 
             AdWordsUtilities::summarize_elog() %>% 
-            filter(cost > 0 & earnings > 0) %>% 
-            group_by(keyword,campaign_name) %>% 
+            filter(cost > 0 & earnings > 0) %>%
+            group_by(campaign_name,keyword,match_type) %>%
             summarize(total_cost = sum(cost),
                       total_contribution = sum(contribution),
                       total_clicks = sum(clicks), 

AdWordsUtilities/R/desktop_gdn_performance_over_time.r

@@ -0,0 +1,42 @@
+#' Create plot of Google Display Network mobile performance over time
+#'
+#' Takes in a AdWords event log and processes performance on desktop devices over time.
+#'
+#' @param gdn_elog A data frame containing keywords and transaction event log data.
+#' @param campaign_filter Optional. (e.g. "Paleo Performers") If provided, generates a data frame and plot for campaigns where the string is included in the campaign name. Default to NULL.
+#' @param plot logical. If TRUE, creates a plot of the data.  If false, creates no plot, Defaults to \code{TRUE}.
+#' @return The data frame used to create a plot of mobile performance over time.
+#' @export
+#' @examples
+#' overall_performance_over_time(elog_data_frame)
+
+desktop_gdn_performance_over_time <- function( gdn_elog, 
+                                              campaign_filter=NULL,
+                                              plot = TRUE){
+  require(plyr)
+  require(dplyr)
+  require(SalvatoUtilities)
+  if( !is.null(campaign_filter) ) {
+    gdn_elog <-  gdn_elog %>% filter(grepl(campaign_filter,campaign_name))
+  }
+  devices_over_time <- gdn_elog %>%
+                        group_by(device,week) %>%
+                        summarize(cost = sum(cost, na.rm = TRUE),
+                                  contribution = sum(money_in_the_bank_paid_to_us,na.rm=TRUE) *.25) %>%
+                        mutate(cum_contribution = cumsum(contribution), 
+                               cum_cost = cumsum(cost),
+                               cum_ROI = cum_contribution - cum_cost) %>%
+                        gather(type,value,cum_cost,cum_contribution,cum_ROI)
+
+	if( plot ) {
+		require(ggplot2)
+    plot(ggplot(devices_over_time %>% 
+            filter(device == "dt"), 
+        aes(week,value,group=type,col=type,fill=type)) + 
+         geom_line() + 
+         ggtitle("GDN Trends on Desktop"))
+	}
+
+	return(devices_over_time)
+
+}

AdWordsUtilities/R/mobile_gdn_performance_over_time.r

@@ -0,0 +1,42 @@
+#' Create plot of Google Display Network mobile performance over time
+#'
+#' Takes in a AdWords event log and processes performance on mobile devices over time.
+#'
+#' @param gdn_elog A data frame containing keywords and transaction event log data.
+#' @param campaign_filter Optional. (e.g. "Paleo Performers") If provided, generates a data frame and plot for campaigns where the string is included in the campaign name. Default to NULL.
+#' @param plot logical. If TRUE, creates a plot of the data.  If false, creates no plot, Defaults to \code{TRUE}.
+#' @return The data frame used to create a plot of mobile performance over time.
+#' @export
+#' @examples
+#' overall_performance_over_time(elog_data_frame)
+
+mobile_gdn_performance_over_time <- function( gdn_elog, 
+                                              campaign_filter=NULL,
+                                              plot = TRUE){
+  require(plyr)
+  require(dplyr)
+  require(SalvatoUtilities)
+  if( !is.null(campaign_filter) ) {
+    gdn_elog <-  gdn_elog %>% filter(grepl(campaign_filter,campaign_name))
+  }
+  devices_over_time <- gdn_elog %>%
+                        group_by(device,week) %>%
+                        summarize(cost = sum(cost, na.rm = TRUE),
+                                  contribution = sum(money_in_the_bank_paid_to_us,na.rm=TRUE) *.25) %>%
+                        mutate(cum_contribution = cumsum(contribution), 
+                               cum_cost = cumsum(cost),
+                               cum_ROI = cum_contribution - cum_cost) %>%
+                        gather(type,value,cum_cost,cum_contribution,cum_ROI)
+
+	if( plot ) {
+		require(ggplot2)
+		plot(ggplot(devices_over_time %>% 
+                filter(device == "mb"), 
+            aes(week,value,group=type,col=type,fill=type)) + 
+             geom_line() + 
+             ggtitle("GDN Trends on Mobile"))
+	}
+
+	return(devices_over_time)
+
+}

AdWordsUtilities/R/summarize_elog.r

@@ -17,9 +17,10 @@ summarize_elog <- function(elog_data_frame,
                                    ){
   require(plyr)
   require(dplyr)
-  return (summarize(elog_data_frame, cost = sum(cost, na.rm = TRUE),
+  return (summarize(elog_data_frame, 
+                    cost = sum(cost, na.rm = TRUE),
                     average_position = weighted.mean(average_position,impressions, na.rm=TRUE),
-                    average_quality_score=mean(quality_score, na.rm=TRUE),
+                    average_quality_score=ifelse(is.null(elog_data_frame$quality_score), NA, mean(quality_score, na.rm=TRUE)) ,
                     estimated_available_impressions = sum(impressions/est_search_impression_share, na.rm=TRUE),
                     impressions = sum(impressions, na.rm = TRUE),
                     # imp share is not wholly accurate because of the way the numbers are reported, but close enough. 
@@ -35,8 +36,8 @@ summarize_elog <- function(elog_data_frame,
                     earnings_per_click = earnings/clicks,
                     contribution_per_click= contribution/clicks,
                     cpa = ifelse(num_acquisitions==0, cost, cost/num_acquisitions),
-                    referred_users=sum(new_referred_users, na.rm=TRUE),
-                    referred_earnings=sum(referred_users_transaction_amount,na.rm=TRUE),
+                    referred_users=ifelse(is.null(elog_data_frame$new_referred_users), NA, sum(new_referred_users, na.rm=TRUE)),
+                    referred_earnings=ifelse(is.null(elog_data_frame$referred_users_transaction_amount), NA, sum(referred_users_transaction_amount,na.rm=TRUE)),
                     estimated_ltv = num_acquisitions*average_num_orders_in_lifetime*average_value_per_order*margin,
                     estimated_lifetime_ROAS=(estimated_ltv-cost)/cost,
                     total_earnings = earnings + referred_earnings,

BingUtilities/R/contribution_per_click_report.r

@@ -29,10 +29,10 @@ contribution_per_click_report <- function(keywords_elog,
   }
 
   data <- keywords_elog %>% 
-            group_by(keyword,campaign_name) %>% 
+            group_by(keyword,match_type,campaign_name) %>% 
             BingUtilities::summarize_elog() %>% 
-            filter(cost > 0 & earnings > 0) %>% 
-            group_by(keyword,campaign_name) %>% 
+            filter(cost > 0 & earnings > 0) %>%
+            group_by(keyword,match_type,campaign_name) %>%
             summarize(total_cost = sum(cost),
                       total_contribution = sum(contribution),
                       total_clicks = sum(clicks), 

PowerSupplyUtilities/NAMESPACE

@@ -4,6 +4,7 @@ export(clean_transactions_data_for_adwords)
 export(clean_transactions_data_for_bing)
 export(create_adwords_event_log)
 export(create_bing_event_log)
+export(create_google_display_event_log)
 export(customers_per_month_report)
 export(get_first_transaction_dates_for_users)
 export(get_referrals_data)

PowerSupplyUtilities/R/create_google_display_event_log.r

@@ -0,0 +1,95 @@
+#' Create Google Display Event Log (elog) for AdWords
+#'
+#'
+#' @param from Start date in either format <"yyyy-mm-dd"> or <yyyymmdd>. Inclusive.
+#' @param to End date in either format <"yyyy-mm-dd"> or <yyyymmdd>. Inclusive.
+#' @return A data frame with all events (AdWords clicks, Transactions and Referrals).
+#' @export
+#' @examples
+#' create_google_display_event_log(from=20150101, to=20151231)
+
+create_google_display_event_log <- function(from=Sys.Date(), 
+                                  to=Sys.Date()){
+  require(AdWordsUtilities)
+  require(PowerSupplyUtilities)
+  require(SalvatoUtilities)
+
+  ppc_events <- all_ppc_raw_completed_order_events( from = from, to = to )
+  mixpanel_adwords_conversions <- ppc_events[["adwords"]]
+  mixpanel_adwords_conversions <- clean_completed_order_events(mixpanel_adwords_conversions)
+
+  ##### Retrieve AdWords Spend/Click Data
+  adwords_campaigns_data <- AdWordsUtilities::campaign_performance_data(from=as.Date(from), to=as.Date(to))
+
+  # Retrieve revenue data
+  db_transactions <- get_transactions_data(from=from, to=to)
+  #db_influencer_metrics <- get_referrals_data(from=from, to=to)
+  db_first_transactions  <- get_first_transaction_dates_for_users(from=from, to=to, users=db_transactions$app_user_id)
+
+  # Join Mixpanel Conversion Data with transaction data
+  unique_users <- distinct(mixpanel_adwords_conversions, app_user_id)
+  db_transactions  <- db_transactions %>% inner_join(unique_users, by="app_user_id")
+
+  #Filter out people where their first order was not in the specified start date and end date
+  db_transactions <- db_transactions %>% filter(is.element(app_user_id, db_first_transactions$id))
+
+  db_transactions <- clean_transactions_data_for_adwords(db_transactions)
+
+  # Add campaign names to db_transactions log
+  db_transactions <- campaign_lookup_table(from=from, to=to) %>%
+                       right_join(db_transactions, by=c(campaign_id = "campaign_id"))
+
+
+  ###################################### CREATE ELOGS ################################################
+  # Create keywords elog
+  keywords_elog <- rbind.fill(db_transactions, adwords_campaigns_data)
+  keywords_elog$week <- as.week(keywords_elog$date)
+  keywords_elog <- keywords_elog %>% arrange(week)
+
+  # Create join table for user_id and the keyword and campaign_name of first purchase
+  user_first_acquisition_metrics <- keywords_elog %>%
+                                    filter(!is.na(user_id)) %>% #Remove NA user_ids (which means they are not monetary transactions)
+                                    group_by(user_id) %>%
+                                    summarize(keyword = first(keyword),
+                                              campaign_name=first(campaign_name),
+                                              campaign_id=first(campaign_id),
+                                              device=first(device),
+                                              match_type=first(match_type))
+
+  #Add influencer metrics to the event log
+  db_influencer_metrics <- get_referrals_data(from=from, to=to)
+  influencer_metrics_with_user_data <- db_influencer_metrics %>%
+                                      rename(week=week_start,
+                                            user_id=influencer_id) %>%
+                                      mutate(week = as.Date(week, format = '%Y-%m-%d')) %>%
+                                      inner_join(user_first_acquisition_metrics, by=c(user_id="user_id")) %>%
+                                      filter(week >= from, week <= to)
+  keywords_elog <- rbind.fill(keywords_elog, influencer_metrics_with_user_data)
+
+  keywords_elog <- keywords_elog %>% filter(grepl("Rmktg",campaign_name))
+
+  keywords_elog <- keywords_elog %>% 
+                    select( -transaction_date,
+                            -event,
+                            -time,
+                            -distinct_id,
+                            -X.created,
+                            -X.email,
+                            -X.first_name,
+                            -X.last_name,
+                            -X.lib_version,
+                            -X.name,
+                            -currency,
+                            -discount,
+                            -mp_country_code,
+                            -mp_lib,
+                            -products,
+                            -tax,
+                            -total,
+                            -lastOrderedFrom,
+                            -utm_content,
+                            -mp_keyword,
+                            -traits)
+
+  return(keywords_elog)
+}