upload meta data and app files

ecom_cvr_aov_by_month.csv

@@ -0,0 +1,157 @@
+Country,CVR,AOV,PC,DATE
+US,1.19%,671,28%,2022-08-01
+DE,2.23%,449,28%,2022-08-01
+NL,2.60%,400,28%,2022-08-01
+AU,2.08%,632,28%,2022-08-01
+SE,2.69%,419,28%,2022-08-01
+FR,1.53%,434,28%,2022-08-01
+JP,1.22%,410,28%,2022-08-01
+BE,2.29%,475,28%,2022-08-01
+IT,1.02%,469,28%,2022-08-01
+CH,1.96%,684,28%,2022-08-01
+AT,2.30%,545,28%,2022-08-01
+NO,2.22%,684,28%,2022-08-01
+DE,2.60%,517,28%,2022-09-01
+NL,2.18%,472,28%,2022-09-01
+US,2.56%,753,28%,2022-09-01
+SE,2.12%,585,28%,2022-09-01
+BE,2.15%,529,28%,2022-09-01
+FR,1.82%,514,28%,2022-09-01
+AU,1.83%,703,28%,2022-09-01
+JP,0.91%,448,28%,2022-09-01
+IT,1.27%,519,28%,2022-09-01
+CH,2.03%,825,28%,2022-09-01
+AT,1.99%,693,28%,2022-09-01
+NO,2.39%,735,28%,2022-09-01
+DE,2.81%,571,28%,2022-10-01
+NL,1.93%,501,28%,2022-10-01
+US,2.31%,900,28%,2022-10-01
+SE,2.70%,453,28%,2022-10-01
+BE,1.56%,789,28%,2022-10-01
+AU,2.23%,774,28%,2022-10-01
+IT,1.35%,536,28%,2022-10-01
+FR,2.09%,566,28%,2022-10-01
+AT,2.33%,761,28%,2022-10-01
+CH,2.44%,966,28%,2022-10-01
+JP,1.09%,438,28%,2022-10-01
+NO,3.68%,567,28%,2022-10-01
+DE,4.99%,460,28%,2022-11-01
+NL,3.92%,378,28%,2022-11-01
+US,4.09%,688,28%,2022-11-01
+SE,3.93%,431,28%,2022-11-01
+FR,3.49%,462,28%,2022-11-01
+BE,3.90%,516,28%,2022-11-01
+IT,2.81%,462,28%,2022-11-01
+CH,3.79%,727,28%,2022-11-01
+AU,4.23%,525,28%,2022-11-01
+AT,5.05%,573,28%,2022-11-01
+NO,4.24%,549,28%,2022-11-01
+JP,1.56%,435,28%,2022-11-01
+DE,3.70%,458,28%,2022-12-01
+US,4.11%,649,28%,2022-12-01
+NL,2.86%,436,28%,2022-12-01
+FR,2.91%,478,28%,2022-12-01
+BE,3.07%,502,28%,2022-12-01
+SE,2.85%,461,28%,2022-12-01
+IT,1.87%,500,28%,2022-12-01
+CH,3.37%,698,28%,2022-12-01
+AT,3.27%,553,28%,2022-12-01
+AU,1.84%,585,28%,2022-12-01
+JP,1.21%,521,28%,2022-12-01
+NO,2.57%,526,28%,2022-12-01
+DE,3.05%,495,28%,2023-01-01
+NL,2.71%,448,28%,2023-01-01
+US,3.05%,649,28%,2023-01-01
+SE,1.95%,449,28%,2023-01-01
+BE,2.72%,546,28%,2023-01-01
+FR,2.12%,533,28%,2023-01-01
+IT,1.37%,491,28%,2023-01-01
+CH,3.03%,629,28%,2023-01-01
+JP,1.02%,581,28%,2023-01-01
+AT,2.81%,543,28%,2023-01-01
+AU,2.36%,602,28%,2023-01-01
+NO,1.21%,550,28%,2023-01-01
+DE,3.42%,505,28%,2023-02-01
+NL,3.09%,446,28%,2023-02-01
+US,3.07%,664,28%,2023-02-01
+SE,2.27%,487,28%,2023-02-01
+IT,1.75%,504,28%,2023-02-01
+FR,1.98%,502,28%,2023-02-01
+BE,2.97%,504,28%,2023-02-01
+CH,3.54%,708,28%,2023-02-01
+JP,0.84%,605,28%,2023-02-01
+AT,3.65%,582,28%,2023-02-01
+AU,2.61%,586,28%,2023-02-01
+NO,1.92%,532,28%,2023-02-01
+DE,3.59%,513,28%,2023-03-01
+NL,2.71%,445,28%,2023-03-01
+US,3.05%,669,28%,2023-03-01
+SE,2.49%,468,28%,2023-03-01
+IT,1.61%,497,28%,2023-03-01
+FR,1.86%,535,28%,2023-03-01
+BE,3.05%,549,28%,2023-03-01
+AU,2.26%,581,28%,2023-03-01
+CH,3.68%,725,28%,2023-03-01
+AT,4.26%,602,28%,2023-03-01
+JP,0.97%,567,28%,2023-03-01
+NO,1.37%,634,28%,2023-03-01
+DE,3.08%,504,28%,2023-04-01
+NL,2.80%,484,28%,2023-04-01
+US,2.62%,722,28%,2023-04-01
+SE,1.91%,523,28%,2023-04-01
+FR,1.90%,567,28%,2023-04-01
+IT,1.33%,534,28%,2023-04-01
+BE,2.71%,602,28%,2023-04-01
+AU,2.35%,631,28%,2023-04-01
+AT,2.97%,585,28%,2023-04-01
+CH,2.87%,742,28%,2023-04-01
+JP,0.82%,707,28%,2023-04-01
+NO,1.41%,644,28%,2023-04-01
+DE,3.07%,471,28%,2023-05-01
+NL,2.47%,469,28%,2023-05-01
+US,2.71%,682,28%,2023-05-01
+SE,1.98%,511,28%,2023-05-01
+FR,1.77%,547,28%,2023-05-01
+IT,1.42%,529,28%,2023-05-01
+BE,3.00%,576,28%,2023-05-01
+AU,1.97%,636,28%,2023-05-01
+AT,2.81%,656,28%,2023-05-01
+JP,0.93%,676,28%,2023-05-01
+CH,2.40%,676,28%,2023-05-01
+NO,1.80%,687,28%,2023-05-01
+DE,2.52%,466,28%,2023-06-01
+NL,2.62%,404,28%,2023-06-01
+US,2.25%,622,28%,2023-06-01
+SE,1.98%,453,28%,2023-06-01
+FR,1.83%,524,28%,2023-06-01
+BE,2.82%,521,28%,2023-06-01
+IT,1.48%,519,28%,2023-06-01
+AU,2.51%,576,28%,2023-06-01
+JP,1.34%,575,28%,2023-06-01
+AT,2.06%,569,28%,2023-06-01
+CH,2.75%,761,28%,2023-06-01
+NO,2.46%,623,28%,2023-06-01
+DE,4.48%,432,28%,2023-07-01
+NL,4.55%,397,28%,2023-07-01
+US,4.11%,557,28%,2023-07-01
+SE,2.82%,418,28%,2023-07-01
+AU,4.53%,507,28%,2023-07-01
+BE,4.12%,568,28%,2023-07-01
+FR,3.48%,490,28%,2023-07-01
+IT,2.55%,448,28%,2023-07-01
+JP,2.32%,454,28%,2023-07-01
+NO,3.07%,484,28%,2023-07-01
+AT,3.98%,636,28%,2023-07-01
+CH,3.37%,587,28%,2023-07-01
+DE,3.03%,527,28%,2023-08-01
+US,2.64%,704,28%,2023-08-01
+AU,1.46%,582,28%,2023-08-01
+NL,3.16%,475,28%,2023-08-01
+SE,1.53%,520,28%,2023-08-01
+FR,1.62%,620,28%,2023-08-01
+IT,1.07%,551,28%,2023-08-01
+BE,2.22%,550,28%,2023-08-01
+AT,2.40%,593,28%,2023-08-01
+CH,2.90%,740,28%,2023-08-01
+NO,1.67%,641,28%,2023-08-01
+JP,0.84%,642,28%,2023-08-01

mCPC_dashboard.py

@@ -0,0 +1,128 @@
+import streamlit as st
+from PIL import Image
+from marginal_cpc_calculator import *
+
+st.set_page_config(
+    page_title="mCPC dashboard",
+    page_icon="socks",
+    layout="wide",
+    initial_sidebar_state="auto",
+)
+
+# dataframe operations first time
+def impute_sub_channel(campaign):
+    if any([x in campaign for x in ['| Prospecting','| CPA','| Traffic','| CRM','Lead acquisition']]):
+        return 'TRAFFIC_DRIVING'
+    elif any([x in campaign for x in ['| Conversion','| Advantage+','| ASC+','Sales |','-Conversions']]):
+        return 'CONVERSION'
+    elif any([x in campaign for x in ['| Brand Awareness','| Reach']]):
+        return 'REACH'
+    else:
+        return 'OTHER'
+df = pd.read_csv('meta_daily_ads_level_data/meta_ad_day_level_raw_data.csv.gz',compression='gzip',parse_dates=['Date'])
+df.rename(columns={'Ad Set Name':'ad_set_name', 'Campaign Name':'campaign', 'Country':'country',
+                   'Date':'date', 'Amount Spent (SEK)':'cost', 'Clicks (all)':'clicks','Impressions':'impressions',
+                   'Purchases Conversion Value (SEK)':'platform_revenue','Ad Set Name':'ad_name',},inplace=True)
+df.set_index('date',inplace=True) 
+df['cpc']=df['cost']/df['clicks']
+df['platform']='meta'
+df['sub_channel']=df.campaign.apply(impute_sub_channel)
+
+# dashboard operations
+def add_logo(logo_path, width, height):
+    """Read and return a resized logo"""
+    logo = Image.open(logo_path)
+    modified_logo = logo.resize((width, height))
+    return modified_logo
+
+def get_date(input_min_date_month,input_min_date_year,input_max_date_month,input_max_date_year):
+    month_converter = {'Jun':'06','Jul':'07','Aug':'08','Sep':'09',
+                       'Oct':'10','Nov':'11','Dec':'12','Jan':'01','Feb':'02',
+                       'Mar':'03','Apr':'04','May':'05'}
+    min_date = input_min_date_year+'-'+month_converter[input_min_date_month]+'-01'
+    max_date = input_max_date_year+'-'+month_converter[input_max_date_month]+'-01'
+    return min_date,max_date
+
+my_logo = add_logo(logo_path="hs_logo.png", width=180, height=60)
+st.sidebar.image(my_logo)
+
+st.title("mCPC & ad spend insights ✨")
+
+# INLCUDE SECTION FOR INPUTS AND BUTTON TO GENERATE INSIGHTS
+with st.sidebar:
+    st.write("Enter  selections for mCPC spend profile")
+    input_country = st.selectbox('Country:',
+                                ('','DE', 'US', 'UK','NL','SE','AU','CH','FR','BE'))
+    input_platform = st.selectbox('Platform:',
+                                ('','meta'))
+    input_sub_channel = st.selectbox('Tactic / Sub channel:',
+                                ('','CONVERSION','TRAFFIC_DRIVING','REACH'))
+    col1, col2 = st.columns(2)
+    with col1:
+        input_min_date_month = st.selectbox('From Month:',
+                                ('','Jun','Jul','Aug','Sep','Oct','Nov','Dec','Jan','Feb','Mar','Apr','May'))
+    with col2:
+        input_min_date_year = st.selectbox('From Year:',
+                                ('','2022','2023'))
+    col3, col4 = st.columns(2)
+    with col3:
+        input_max_date_month = st.selectbox('To Month:',
+                                ('','Jun','Jul','Aug','Sep','Oct','Nov','Dec','Jan','Feb','Mar','Apr','May'))
+    with col4:
+        input_max_date_year = st.selectbox('To Year:',
+                                ('','2022','2023'))   
+    start = st.button("Generate mCPC curves", type='primary') 
+    text_error_value = ''
+    if start:
+        for var,varname in [(input_country,'Country'),
+                    (input_platform,'Platform'),
+                    (input_min_date_month,'From Date'),
+                    (input_min_date_year,'From Date'),
+                    (input_max_date_month,'To Date'),
+                    (input_max_date_year,'To Date')]:
+            if var=='':
+                text_error_value = f'{varname} cant be empty'
+                st.error(text_error_value)
+                break    
+
+# CHECKS TO ENSURE INPUTS ARE CLEAN BEFORE CALLING FUNCTIONS
+if start:
+    if text_error_value=='':
+        min_date, max_date = get_date(input_min_date_month,
+                                      input_min_date_year,
+                                      input_max_date_month,
+                                      input_max_date_year)
+        st.markdown("""---""")
+        mid, mad = get_comparison_dates(min_date,max_date,period='previous_period')
+        st.subheader(f"Calculating for dates between {mid} & {mad}")
+        fig,output_msg = calculate_max_spend(df,mid,mad,
+                                            revenue_or_ebitda='revenue',
+                                            country=input_country,
+                                            platform=input_platform,
+                                            sub_channel=input_sub_channel,
+                                            pprint=True)
+        st.write(fig)
+        for l in output_msg:
+            st.write(l)
+        st.markdown("""---""")
+        mid, mad = get_comparison_dates(min_date,max_date,period='LY')
+        st.subheader(f"Calculating for dates between {mid} & {mad}")
+        fig,output_msg = calculate_max_spend(df,mid,mad,
+                                            revenue_or_ebitda='revenue',
+                                            country=input_country,
+                                            platform=input_platform,
+                                            sub_channel=input_sub_channel,
+                                            pprint=True)
+        st.write(fig)
+        for l in output_msg:
+            st.write(l)
+
+    else:
+        st.write(f'Text error value is {text_error_value}')
+# DISPLAY INSIGHTS FOR PREVIOUS PERIOD AND LAST YEAR
+# TEST FUNCTIONALITY OF APP
+# DEPLOY TO HUGGINGFACE
+# INCORPORATE GOOGLE
+# INCORPORATE TIKTOK
+
+st.markdown("<br><hr><center>Made with ❤️ for Happy Socks by <a href='mailto:vivekbharadwaj.ca@gmail.com?subject=mCPC dashboard queries&body=Please specify the issue you are facing with the dashboard.'><strong>Vivek</strong></a> ✨</center><hr>", unsafe_allow_html=True)

marginal_cpc_calculator.py

@@ -0,0 +1,210 @@
+import numpy as np
+from scipy.optimize import curve_fit
+import matplotlib.pyplot as plt
+import pandas as pd
+
+def hello():
+    print ("world!")
+
+def polynomial_func(x, a,b,c,d):
+    y = a * x - b * x**2 + c * x**3 + d
+    return y
+
+def get_country_metrics(country,min_date='2023-07-01',max_date='2023-09-01'):
+    df_country = pd.read_csv('ecom_cvr_aov_by_month.csv',parse_dates=['DATE'])
+    df_country['CVR']=df_country['CVR'].apply(lambda x: x.replace("%","")).astype(float)/100
+    df_country['PC']=df_country['PC'].apply(lambda x: x.replace("%","")).astype(float)/100
+    if country in df_country.Country.unique():
+        return df_country[(df_country.Country==country) & 
+                           (df_country.DATE>=min_date) & 
+                           (df_country.DATE<max_date)].groupby('Country')[['CVR','AOV','PC']].mean().values[0]
+    else:
+        return df_country.loc[(df_country.DATE>=min_date) & 
+                               (df_country.DATE<max_date),['CVR','AOV','PC']].mean().values
+
+def get_comparison_dates(min_date,max_date,period='previous_period'):
+    '''
+    period is either previous period or last year
+    '''
+    if period=='previous_period':
+        delta = (pd.to_datetime(max_date).to_period('M') - pd.to_datetime(min_date).to_period('M')).n
+    else:
+        delta = 12
+    mid,mad=(pd.DataFrame([pd.to_datetime(min_date), pd.to_datetime(max_date)]) - pd.DateOffset(months=delta)).values
+    return pd.to_datetime(mid[0]).strftime('%Y-%m-%d'),pd.to_datetime(mad[0]).strftime('%Y-%m-%d')
+
+
+def calculate_max_spend(df,min_date,max_date,
+                        revenue_or_ebitda='revenue',
+                        country='DE',
+                        platform='meta',
+                        sub_channel='',
+                        pprint=False):
+    # consider only ads with positive click outcomes and sort by cpc
+    if sub_channel=='':
+        df_temp = df[(df.clicks>0) & (df.country==country) & 
+                    (df.platform==platform) & 
+                    (df.index>min_date) & (df.index<max_date)
+                        ].sort_values('cpc').reset_index()
+    else:
+        df_temp = df[(df.clicks>0) & (df.country==country) & 
+                    (df.platform==platform) & 
+                    (df.sub_channel==sub_channel) &
+                    (df.index>min_date) & (df.index<max_date)
+                        ].sort_values('cpc').reset_index()
+
+    # calculate cumulative spend and cumulative clicks
+    df_temp['cum_cost']=df_temp['cost'].cumsum()
+    df_temp['cum_clicks']=df_temp['clicks'].cumsum()
+    df_temp['cum_cpc']=df_temp['cum_cost']/df_temp['cum_clicks']
+
+    # calculate revenue and income
+    cvr,aov,pc = get_country_metrics(country,min_date,max_date)
+    df_temp['cum_revenue']=df_temp['cum_clicks']*cvr*aov
+    # at some stage, the ebitda becomes negative
+    if revenue_or_ebitda=='revenue':
+        df_temp['cum_income'] = df_temp['cum_revenue']-df_temp['cum_cost']
+    else:
+        df_temp['cum_income'] = df_temp['cum_revenue'] * pc - df_temp['cum_cost']
+    
+    # calculate marginal cpc and marginal income
+    df_temp['marginal_cpc'] = (df_temp.cum_cost - df_temp.cum_cost.shift(1))/(df_temp.cum_clicks - df_temp.cum_clicks.shift(1))
+    df_temp['marginal_income'] = df_temp.cum_income - df_temp.cum_income.shift(1) - df_temp['marginal_cpc']
+    
+    try:
+        params, cv = curve_fit(polynomial_func, df_temp['cum_cost'],df_temp['marginal_income'].fillna(0), p0=(1, 1,1,1))
+        a,b,c,d = params
+
+        # determine quality of the fit
+        squaredDiffs = np.square(df_temp['marginal_income'].fillna(0) - polynomial_func(df_temp['cum_cost'], a,b,c,d))
+        squaredDiffsFromMean = np.square(df_temp['marginal_income'].fillna(0) - np.mean(df_temp['marginal_income'].fillna(0)))
+        rSquared = 1 - np.sum(squaredDiffs) / np.sum(squaredDiffsFromMean)
+        output_msg=[]
+        if pprint==True:
+            output_msg.append(f"R² of fit for marginal income = {rSquared}")
+            # print(f"R² of fit for marginal income = {rSquared}")
+
+        # inspect the parameters
+        if pprint==True:
+            output_msg.append(f"Marginal income equation Y = {a} * x - {b} * x^2 + {c} * x^3 + {d}")
+
+        # calculate max costs when it becomes negative ROAS
+        df_marginal = pd.DataFrame(zip(np.arange(500000), 
+                                       polynomial_func(np.arange(500000), a,b,c,d)
+                                      )).rename(columns={0:'cumulative_cost',1:'marginal_income'})
+        # join actuals to get the full picture
+        _ = df_temp[['cost','cpc','cum_cost', 'cum_clicks', 'cum_cpc', 'cum_revenue', 'cum_income']]
+        # converting cost to integer in order to join with the marginal dataset
+        _['cum_cost']=_['cum_cost'].apply(lambda x:np.round(x,0).astype(int))
+        df_marginal = pd.merge(df_marginal, 
+                               _.groupby('cum_cost')[
+                                   ['cost','cpc','cum_clicks', 'cum_cpc', 'cum_revenue', 'cum_income']
+                                    ].max().reset_index().rename(columns={#'cum_cost':'cumulative_cost',
+                                                                          'cpc':'marginal_cpc'}),
+                              how='left',left_on='cumulative_cost',right_on='cum_cost').fillna(method='ffill')
+        max_spend_threshold = df_marginal.loc[(df_marginal.marginal_income.shift(1)>=0) & (df_marginal.marginal_income.shift(-1)<0),'cumulative_cost'].min()
+        highest_amount_spent = df_marginal.cum_cost.max()
+    except:
+        if pprint==True:
+            output_msg.append("Something happened during marginal CPC and equation calculations. We couldn't calculate max_spend_threshold and highest_amount_spent...")
+        max_spend_threshold=np.nan
+        highest_amount_spent=np.nan
+        rSquared=np.nan
+        total_negative_roas_spend=np.nan
+
+    if pd.notnull(max_spend_threshold):
+        if (max_spend_threshold>highest_amount_spent):
+            marginal_cpc,cum_clicks,cum_cpc,cum_revenue = np.nan, np.nan, np.nan, np.nan
+            if pprint==True:
+                output_msg.append(f"Highest amount spent during this period is: {highest_amount_spent}")
+        else:
+            marginal_cpc,cum_clicks,cum_cpc,cum_revenue = df_marginal.loc[df_marginal.cumulative_cost==max_spend_threshold,
+                                                                      ['marginal_cpc','cum_clicks','cum_cpc','cum_revenue']].values[0]
+            # any spend beyond the max_spend_threshold needs to be saved
+            total_negative_roas_spend = df_temp.loc[df_temp.cum_cost>max_spend_threshold,['cost']].sum().values[0]
+    #     # so to get the max cost for this tactic or strategy, 
+        if pprint==True:
+            output_msg.append(f"Max spend threshold is: {max_spend_threshold}") 
+            output_msg.append(f"For this spend, marginal_cpc={marginal_cpc}, cumulative_clicks={cum_clicks}, cumulative_cpc={cum_cpc}, cumulative_revenue={cum_revenue}")
+            output_msg.append(f"Total amount spent in negative ROAS={total_negative_roas_spend}")
+            print (output_msg)
+
+            plt.style.use('ggplot')
+            fig = plt.figure(figsize=(18,6))
+            ax1 = fig.add_subplot(121)
+            plt.plot(df_temp['cum_cost'],df_temp['marginal_income'], '.', label="data")
+            plt.plot(df_temp['cum_cost'], polynomial_func(df_temp['cum_cost'], a,b,c,d), '--', color='blue', label="fitted")
+            plt.xlabel('cumulative cost'); plt.ylabel('marginal income')
+            ax1.title.set_text('Net income at different spend levels')
+            ax2 = fig.add_subplot(122);
+            plt.scatter(df_temp['cum_cost'],df_temp['cpc']);
+            plt.axhline(1*cvr*aov,color='black',linestyle='--');plt.axhline(1*cvr*aov*pc,color='blue',linestyle='--');
+            plt.annotate(f'Any clicks above {np.round(cvr*aov,2)} SEK cpc has negative ROAS \n Above {np.round(cvr*aov*pc,2)} SEK is negative PC',size=12,xy=[0,60], color="black")
+            plt.xlabel('cumulative cost'); plt.ylabel('cpc')
+            # ax2.axes.get_xaxis().set_visible(False)
+            ax2.title.set_text('CPC for at different spend levels');
+            return fig,output_msg
+    else:
+            marginal_cpc,cum_clicks,cum_cpc,cum_revenue = np.nan, np.nan, np.nan, np.nan
+            total_negative_roas_spend=np.nan
+            if pprint==True:
+                print (f"Not enough data for mCPC analysis. Max threshold resulted in {max_spend_threshold}. Highest amount spent during this period is: {highest_amount_spent}") 
+    #     return df_marginal
+    return max_spend_threshold,highest_amount_spent,total_negative_roas_spend,marginal_cpc,cum_clicks,cum_cpc,cum_revenue,rSquared
+
+def loop_mCPC_countries(df,min_date,max_date,
+                        country_list=['UK', 'DE', 'US', 'NL', 'SE', 'CH', 'BE', 'EU', 'FR', 'AU'],
+                        platform='meta'):
+    country_flag = ''
+    df_result = pd.DataFrame()
+
+    for i,d in df[(df.index>min_date) & 
+                  (df.index<max_date) & (df.platform==platform) & 
+                  (df.country.isin(country_list))].groupby(['country','sub_channel'])['cost'].sum().reset_index().iterrows():
+        if country_flag!=d['country']:
+            country_flag=d['country']
+            print (f"Processing {country_flag}...")
+
+        # run numbers for prior delta
+        mid, mad = get_comparison_dates(min_date,max_date,period='previous_period')
+        mcost = calculate_max_spend(
+                        df,min_date=mid,max_date=mad,
+                        revenue_or_ebitda='revenue',
+                        country=d['country'],
+                        platform=platform,
+                        sub_channel=d['sub_channel'],
+                        pprint=False)
+        
+        # rerun values for LY this time
+        mid, mad = get_comparison_dates(min_date,max_date,period='LY')
+        mcost_ly = calculate_max_spend(
+                        df,min_date=mid,max_date=mad,
+                        revenue_or_ebitda='revenue',
+                        country=d['country'],
+                        platform=platform,
+                        sub_channel=d['sub_channel'],
+                        pprint=False)
+        # max_spend_threshold,highest_amount_spent,marginal_cpc,cum_clicks,cum_cpc,cum_revenue,rSquared = \
+        # combine the 2 results
+        df_result = pd.concat([df_result,
+                            pd.Series({'platform':platform,
+                                        'country':d['country'], 
+                                        'sub_channel':d['sub_channel'],
+                                        'budget':'',
+                                        'max_spend_PP':mcost[0],
+                                        'amt_spent_PP':mcost[1],
+                                        'rSquared_PP':mcost[6], # can you trust this mcpc analysis for this market?                                     
+                                        'max_spend_LY':mcost_ly[0],
+                                        'amt_spent_LY':mcost_ly[1],
+                                        'rSquared_LY':mcost_ly[6], # can you trust this mcpc analysis for this market?                                     
+                                        'cpc_PP':mcost[2],
+                                        'cpc_LY':mcost_ly[2],
+                                        'cum_clicks_PP':mcost[3],
+                                        'cum_clicks_LY':mcost_ly[3],
+                                        'cum_cpc_PP':mcost[4],
+                                        'cum_cpc_LY':mcost_ly[4],                                     
+                                        })],axis=1)
+    df_result = df_result.T
+    df_result['diff']=df_result['max_spend_PP']-df_result['amt_spent_PP']
+    print (f"Total impact due to unrealized revenue and negative ROAS spend between planned and threshold costs is {(df_result['diff'].apply(abs)*-1).sum()} SEK")
+    return df_result.sort_values('diff',ascending=True).reset_index(drop=True)

meta_daily_ads_level_data/meta_ad_day_level_raw_data.csv.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb08473861a94a374e979a765c03e66a572e948379422d4bedaeb69223763c80
+size 1982487

requirements.txt

@@ -0,0 +1,7 @@
+matplotlib==3.7.3
+numpy==1.24.4
+pandas==2.0.3
+Pillow==9.5.0
+requests==2.31.0
+scipy==1.10.1
+huggingface_hub==0.17.2