Script 417: Campaign Benchmark Performance

Purpose

Tag Dimensions based on CVR, CPL, CPC, CTR Performance compared to previous 6 months.

To Elaborate

The Python script solves the problem of tagging campaign dimensions based on their performance metrics (CVR, CPL, CPC, CTR) compared to the benchmarks set in the previous 6 months. The script calculates the performance ratio of each metric to its benchmark and assigns a performance tag (Over Target, Under Target, On Target) based on predefined criteria. The script also handles cases where the benchmark is not defined or is zero.

Walking Through the Code

  1. The script defines column names for both input and output data.
  2. Empty output columns are set up in the output dataframe.
  3. The script defines a function clean_and_convert_to_numeric to clean and convert values to numeric, handling both numeric and percentage values.
  4. The script defines a function get_col_set_tuple to construct a tuple of column names for marin dimensions based on a naming convention.
  5. The function tag_by_performance is defined to tag dimensions based on performance metrics and benchmarks.
  6. The function takes the input dataframe, output dataframe, column set tuple, and column tag as parameters.
  7. The function cleans and converts the metric, benchmark, and margin columns to numeric values.
  8. If the benchmark is not provided, the median of the metric column is calculated and used as the benchmark.
  9. The function calculates the performance ratio by dividing the metric by the benchmark.
  10. The function separates the dimensions into different performance groups based on predefined criteria.
  11. The function assigns the performance tag to each dimension based on its performance group.
  12. The function returns an array indicating the changed rows in the input dataframe.
  13. The script calls the tag_by_performance function for each of the 4 metrics (CVR, CPL, CPC, CTR) and stores the result in variables changed1, changed2, changed3, and changed4.
  14. The script combines the changes from all metrics using logical OR operation and stores the result in the variable combined_changes.
  15. The script prints the campaigns with changed CVR performance tags for debugging purposes.
  16. The script selects only the campaigns with changed performance tags in the output dataframe.
  17. The script ends.

Vitals

  • Script ID : 417
  • Client ID / Customer ID: 1306925159 / 60269255
  • Action Type: Bulk Upload
  • Item Changed: Campaign
  • Output Columns: Account, Campaign, Conv Rate % - Performance, Cost/Conv $ - Performance, CPC $ - Performance, CTR % - Performance
  • Linked Datasource: M1 Report
  • Reference Datasource: None
  • Owner: ascott@marinsoftware.com (ascott@marinsoftware.com)
  • Created by ascott@marinsoftware.com on 2023-10-25 18:17
  • Last Updated by ascott@marinsoftware.com on 2023-12-06 04:01
> See it in Action

Python Code

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
# columns for output
OUT_COL_MD_DEBUG_PERF_CALC = 'Performance Calc Raw Numbers'
OUT_COL_MD_DEBUG_PERF_CALC_LAST_RUN = 'Performance Calc Last Run Date'

# columns for input
IN_COL_METRIC_CVR = 'Conv. Rate %'
IN_COL_METRIC_CPA = 'Cost/Conv. $'
IN_COL_METRIC_CTR = 'CTR %'
IN_COL_METRIC_CPC = 'Avg. CPC $'


# columns for both input and output
IN_OUT_COL_CAMPAIGN = 'Campaign'
IN_OUT_COL_MD_PERF_CVR = 'Conv Rate % - Performance'
IN_OUT_COL_MD_PERF_CPA = 'Cost/Conv $ - Performance'
IN_OUT_COL_MD_PERF_CTR = 'CTR % - Performance'
IN_OUT_COL_MD_PERF_CPC = 'CPC $ - Performance'

# setup empty output columns
outputDf[OUT_COL_MD_DEBUG_PERF_CALC] = np.nan
#outputDf[OUT_COL_MD_DEBUG_PERF_CALC_LAST_RUN] = datetime.date.today().isoformat()
outputDf[IN_OUT_COL_MD_PERF_CVR] = np.nan
outputDf[IN_OUT_COL_MD_PERF_CPA] = np.nan
outputDf[IN_OUT_COL_MD_PERF_CTR] = np.nan
outputDf[IN_OUT_COL_MD_PERF_CPC] = np.nan

# Ensure that columns with percentage values (e.g., 'Conv. Rate %', 'CTR %') are properly cleaned and converted to numeric
# We need to handle both numeric and percentage values
def clean_and_convert_to_numeric(value):
    if isinstance(value, str):
        cleaned_value = value.replace(',', '')  # Remove commas (e.g., for numbers like '1,000')
        if cleaned_value.endswith('%'):
            numeric_value = float(cleaned_value.rstrip('%')) / 100  # Convert percentage to float
            print(f"Converted {value} to {numeric_value}")
            return numeric_value
        else:
            try:
                numeric_value = float(cleaned_value)
                print(f"Converted {value} to {numeric_value}")
                return numeric_value
            except ValueError:
                print(f"Unable to convert {value} to numeric, setting to np.nan")
                return np.nan
    elif isinstance(value, (float, int)):
        return float(value)
    else:
        return np.nan



# constructs tuple of 4 column names with names of marin dimensions columns based on convention used
def get_col_set_tuple(col_name):
	col_name_clean = col_name.replace('.','')
	return (
		col_name,
		col_name_clean + ' - Avg - Benchmark',
		col_name_clean + ' - High - Criteria',
		col_name_clean + ' - Low - Criteria'
	)


# set col_tag column on outDf with performance tag by comparing col_metric with col_benchmark
# returns array of boolean indicating changed rows in inDf
def tag_by_performance(inDf, outDf, col_set_tuple, col_tag):

	(col_metric, col_benchmark, col_over_margin, col_under_margin) = col_set_tuple

	# setup tmp columns for intrim values
	col_tag_name_new_value = col_tag + "_new"
	col_perf_ratio = col_tag + "_perf_ratio"

	inDf[col_tag_name_new_value] = np.nan

	# Ensure that columns with percentage values (e.g., 'Conv. Rate %', 'CTR %') are properly cleaned and converted to numeric
	inDf[col_metric] = inDf[col_metric].apply(clean_and_convert_to_numeric)
	inDf[col_benchmark] = inDf[col_benchmark].apply(clean_and_convert_to_numeric)
	inDf[col_over_margin] = inDf[col_over_margin].apply(clean_and_convert_to_numeric)
	inDf[col_under_margin] = inDf[col_under_margin].apply(clean_and_convert_to_numeric)

	# calculate benchmark if none provided
	median_metric = inputDf[col_metric].median()
	print(col_metric, "median", median_metric)
	inDf[col_benchmark].fillna({col_benchmark: median_metric}, inplace=True)
		
	# calc relative performance to benchmark
	inDf[col_perf_ratio] = inputDf[col_metric] / inputDf[col_benchmark]

	# separate into different performance groups
	array_benchmark_not_defined = inDf[col_benchmark] == np.nan
	array_benchmark_zero = inDf[col_benchmark] <= 0.0
	array_no_benchmark =  np.logical_or(array_benchmark_not_defined, array_benchmark_zero)

	array_over_perf = (inDf[col_perf_ratio] - 1.0) > inDf[col_over_margin]
	array_under_perf = (1.0 - inDf[col_perf_ratio]) > inDf[col_under_margin]
	array_on_target = np.logical_not(np.logical_or(array_no_benchmark, np.logical_or(array_over_perf, array_under_perf)))

	inDf.loc[ array_over_perf, col_tag_name_new_value ] = 'Over Target'
	inDf.loc[ array_under_perf, col_tag_name_new_value ] = 'Under Target'
	inDf.loc[ array_on_target, col_tag_name_new_value ] = 'On Target'
	#inDf.loc[ array_no_benchmark, col_tag_name_new_value ] = 'No Benchmark'

	# find changed perf tags
	changed = inDf[col_tag_name_new_value].notnull() & (inDf[col_tag] != inDf[col_tag_name_new_value])

	outDf[col_tag] = inDf[col_tag_name_new_value]

	return changed


# tag by performance for each of the 4 metrics

changed1 = tag_by_performance(
				inputDf, 
				outputDf,
				get_col_set_tuple(IN_COL_METRIC_CVR), 
				IN_OUT_COL_MD_PERF_CVR)

changed2 = tag_by_performance(
				inputDf, 
				outputDf,
				get_col_set_tuple(IN_COL_METRIC_CPA), 
				IN_OUT_COL_MD_PERF_CPA)

changed3 = tag_by_performance(
				inputDf, 
				outputDf,
				get_col_set_tuple(IN_COL_METRIC_CTR), 
				IN_OUT_COL_MD_PERF_CTR)

changed4 = tag_by_performance(
				inputDf, 
				outputDf,
				(IN_COL_METRIC_CPC,
				'CPC $ - Avg - Benchmark',
				'CPC $ - High - Criteria',
				'CPC $ - Low - Criteria'
				),
				IN_OUT_COL_MD_PERF_CPC)

combined_changes = np.logical_or(changed4, np.logical_or(changed3, np.logical_or(changed1, changed2)))

# debug print what's changed
print("== Campaigns with CVR Perf Tag changes ==")
in_cols = [IN_OUT_COL_CAMPAIGN, IN_OUT_COL_MD_PERF_CVR] + list(get_col_set_tuple(IN_COL_METRIC_CVR))
out_cols = [IN_OUT_COL_CAMPAIGN, IN_OUT_COL_MD_PERF_CVR]
inDf_changed = inputDf.loc[changed1, in_cols]
outDf_changed = outputDf.loc[changed1, out_cols]
debug_df = inDf_changed.join(outDf_changed, rsuffix="_out")
print(debug_df.to_string())


# only include campaigns with changed perf tag in bulk file
outputDf = outputDf[ combined_changes ]

Post generated on 2024-05-15 07:44:05 GMT

comments powered by Disqus