first commit

milestone_attribution_dataset_module/milestone_attribution_dataset_module.egg-info/PKG-INFO

@@ -0,0 +1,4 @@
+Metadata-Version: 2.4
+Name: milestone_attribution_dataset_module
+Version: 0.1.0
+Summary: Internal PASBDC graph making scripts used to generate figures for the milestone attribution analysis.

milestone_attribution_dataset_module/milestone_attribution_dataset_module.egg-info/SOURCES.txt

@@ -0,0 +1,9 @@
+pyproject.toml
+milestone_attribution_dataset_module/__init__.py
+milestone_attribution_dataset_module/make_funding_data.py
+milestone_attribution_dataset_module/make_nbs_data.py
+milestone_attribution_dataset_module/shared.py
+milestone_attribution_dataset_module.egg-info/PKG-INFO
+milestone_attribution_dataset_module.egg-info/SOURCES.txt
+milestone_attribution_dataset_module.egg-info/dependency_links.txt
+milestone_attribution_dataset_module.egg-info/top_level.txt

milestone_attribution_dataset_module/milestone_attribution_dataset_module.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

milestone_attribution_dataset_module/milestone_attribution_dataset_module.egg-info/top_level.txt

@@ -0,0 +1 @@
+milestone_attribution_dataset_module

milestone_attribution_dataset_module/milestone_attribution_dataset_module/__init__.py

@@ -0,0 +1,10 @@
+from .make_funding_data import sanitize_funding_data
+from .make_nbs_data import sanitize_nbs_data
+from .shared import tag_documentation_level
+
+__all__ = [
+    'tag_documentation_level',
+    'sanitize_funding_data',
+    'sanitize_nbs_data'
+]
+

milestone_attribution_dataset_module/milestone_attribution_dataset_module/make_funding_data.py

@@ -0,0 +1,81 @@
+# Date Created: 12/29/25
+# Author: Vincent Allen
+# PURPOSE:
+
+# This file tags an exported capital funding milestones dataset with documentation levels based on if there is 
+# a value in the affirmation or Attribution statements column as well as a non-blank value in the attribution source
+
+# This script takes in the following filter's data:
+# https://pasbdc.neoserra.com/activity/list/20?__formid=20&remove=&savename=&sort=DATE&sortdir=DESC&expr=&field_1=DATE&opt_auto_1=pfy&field_2=F_CENTER_ID&opt_2=-1&field_3=IS_REPORTABLE&field_4=&sortdir=DESC
+
+import pandas as pd
+import numpy as np
+import argparse
+import os
+import json
+import sys
+
+from pasbdc_data_cleaning import clean_center_name # pyright:ignore
+from .shared import tag_documentation_level
+from constants_module import NEOSERRA_COLUMNS, OUT_COLUMNS
+
+def sanitize_funding_data(
+        df:pd.DataFrame,
+        col_neo_attribution_source:str,
+        col_neo_affirmation:str,
+        col_out_documentation_level:str,
+        col_neo_center:str,
+):
+    clean_center_name(df)
+
+    # Remove rows that cannot be attributed to a center
+    df = df[df[col_neo_attribution_source] != "Not attributed to center"].copy()
+
+    df = tag_documentation_level(
+        df,
+        col_neo_attribution_source=col_neo_attribution_source,
+        col_neo_affirmation=col_neo_affirmation,
+        col_out_documentation_level=col_out_documentation_level,
+    )
+    df[col_neo_attribution_source] = df[col_neo_attribution_source].fillna("Blank").astype(str)
+    df = df.sort_values(by=col_neo_center)
+    return df
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+   
+    parser.add_argument("-f", "--file", 
+                        type=str, 
+                        required=True, 
+                        help="The data file to de-duplicate and tag with documentation levels")
+
+    parser.add_argument("-o", "--out", 
+                        type=str, 
+                        required=True, 
+                        help="The csv file to write the de-duplicated data to")
+
+    parser.add_argument("-m", "--mapping",
+                        type=str,
+                        required=False,
+                        default="",
+                        help="The path to a JSON file with all of the columns in the DEFAULT_CONFIG of this script which can be used to override column name mappings")
+
+    args = parser.parse_args()
+
+    if args.mapping:
+        NEOSERRA_COLUMNS.apply_json_mapping(args.mapping)
+        OUT_COLUMNS.apply_json_mapping(args.mapping)
+
+    original_df = pd.read_csv(args.file)
+
+    original_df = sanitize_funding_data(
+        original_df,
+        col_neo_attribution_source=NEOSERRA_COLUMNS.milestone_attribution_source,
+        col_neo_affirmation=NEOSERRA_COLUMNS.milestone_affirmation,
+        col_out_documentation_level=OUT_COLUMNS.milestone_documentation_level,
+        col_neo_center=NEOSERRA_COLUMNS.center
+    )
+
+    print(f"DONE!")
+
+    original_df.to_csv(args.out, index=False)

milestone_attribution_dataset_module/milestone_attribution_dataset_module/make_nbs_data.py

@@ -0,0 +1,454 @@
+# Date Created: 12/31/25
+# Author: Vincent Allen
+# PURPOSE:
+
+# This file tags an exported new business starts milestones dataset with documentation levels based on if there is
+# a value in the affirmation or Attribution statements column as well as a non-blank value in the attribution source
+
+# This script is meant to work with the data from this neoserra filter:
+# https://pasbdc.neoserra.com/activity/list/7?__formid=7&remove=&savename=&sort=DATE&sortdir=DESC&expr=&field_1=DATE&opt_auto_1=&opt_a_1=10%2F1%2F2024&opt_b_1=9%2F30%2F2025&field_2=F_CENTER_ID&opt_2=-1&field_3=TYPE&opt_not_3=on&opt_3=&field_4=CLI_TYPE2&opt_4=&opt_4=AC&opt_4=IC&field_5=IS_REPORTABLE&field_6=TYPE&opt_6=&opt_6=%2BBSI&field_7=&sortdir=DESC
+
+import pandas as pd
+import re
+import numpy as np
+import argparse
+#from tabulate import tabulate
+import json
+import sys
+
+from pasbdc_data_cleaning import clean_center_name # pyright:ignore
+from .shared import tag_documentation_level
+from constants_module import NEOSERRA_COLUMNS, OUT_COLUMNS
+
+'''
+def filter_valid_milestone(
+        group: pd.DataFrame,
+        col_neo_center: str = "Center",
+        col_neo_client_id: str = "Client ID",
+        col_neo_milestone_date: str = "Milestone Date",
+        col_neo_attribution_source: str = "Attribution Source",
+        col_neo_attribution_date: str = "Attribution Date",
+        col_neo_affirmation: str = "Affirmation",
+        col_neo_milestone_type: str = "Milestone Type",
+        business_start_impact_val: str = "Business Start Impact",
+        business_established_val: str = "Business Established",
+        log_filtering_file: str = "",
+        col_out_documentation_level: str = "Documentation Level",
+):
+    log_file = None
+    blank_vals = ["", " ", "NaN", "nan", np.nan]
+    if log_filtering_file:
+        log_file_drop = open(log_filtering_file + "drop.txt", 'a')
+        log_file_keep = open(log_filtering_file + "keep.txt", 'a')
+
+    if group.shape[0] == 1:
+        should_return = False
+        milestone_type = group[col_neo_milestone_type].iloc[0]
+
+        # Only one entry for this client, check if it can be counted
+        if milestone_type == business_established_val:
+            log_file_drop.write(
+                f"Dropped milestone: \n{tabulate(group, headers='keys', tablefmt='grid', showindex=False)}\n")
+            #DROPPED_CLIENTS.append(str(group[col_neo_client_id].unique()[0]))
+            # Drop the milestone
+            should_return = False
+            # print(f"Dropped milestone: {group}")
+        elif milestone_type == business_start_impact_val:
+            # Keep the milestone
+            log_file_keep.write(
+                f"Kept milestone: \n{tabulate(group, headers='keys', tablefmt='grid', showindex=False)}\n")
+            #KEPT_CLIENTS.append(str(group[col_neo_client_id].unique()[0]))
+            should_return = True
+
+        if log_file_drop is not None:
+            log_file_drop.close()
+            log_file_keep.close()
+
+        if should_return:
+            return group
+        else:
+            return None
+    else:
+        should_return = False
+
+        has_impact_milestone = (group[col_neo_milestone_type] == business_start_impact_val).any()
+
+        combined_milestone = None
+        if has_impact_milestone:
+            # We have a valid business start impact
+            should_return = True
+
+            log_file_keep.write(
+                f"Kept milestone: \n{tabulate(group, headers='keys', tablefmt='grid', showindex=False)}\n")
+            #KEPT_CLIENTS.append(str(group[col_neo_client_id].unique()[0]))
+
+            affirmation_concat_logic = lambda x: ', '.join(x.astype(str)[~x.isin(blank_vals)].unique())
+            attribution_source_concat_logic = lambda x: ', '.join(x.astype(str)[~x.isin(blank_vals)].unique())
+            milestone_type_concat_logic = lambda x: ', '.join(x.astype(str)[~x.isin(blank_vals)].unique())
+            milestone_date_concat_logic = lambda x: ', '.join(x.astype(str)[~x.isin(blank_vals)].unique())
+            attribution_date_concat_logic = lambda x: ', '.join(x.astype(str)[~x.isin(blank_vals)].unique())
+
+            combined_milestone = pd.DataFrame({
+                col_neo_client_id: [group[col_neo_client_id].iloc[0]],
+                col_neo_center: [group[col_neo_center].iloc[0]],
+                col_neo_milestone_type: [milestone_type_concat_logic(group[col_neo_milestone_type])],
+                col_neo_affirmation: [affirmation_concat_logic(group[col_neo_affirmation])],
+                col_neo_attribution_source: [attribution_source_concat_logic(group[col_neo_attribution_source])],
+                col_neo_milestone_date: [milestone_date_concat_logic(group[col_neo_milestone_date])],
+                col_neo_attribution_date: [attribution_date_concat_logic(group[col_neo_attribution_date])]
+            })
+        else:
+            # print(f"Dropped milestone: {group}")
+            log_file_drop.write(
+                f"Dropped milestone: \n{tabulate(group, headers='keys', tablefmt='grid', showindex=False)}\n")
+            #DROPPED_CLIENTS.append(str(group[col_neo_client_id].unique()[0]))
+            should_return = False
+
+        if log_file is not None:
+            log_file_drop.close()
+            log_file_keep.close()
+        if should_return:
+            return combined_milestone
+        else:
+            return None
+'''
+
+'''
+def tag_documentation_level(
+        nbs_data: pd.DataFrame,
+        col_neo_attribution_source: str = "Attribution Source",
+        col_neo_affirmation: str = "Affirmation",
+        col_neo_attribution_statement: str = "Attribution Statement",
+        col_out_documentation_level: str = "Documentation Level"
+):
+    # 1. Define your search lists
+    no_doc_terms = ['Requested on eCenter']
+    documented_terms = [
+        'eCenter',
+        'Email from Client',
+        'Director Confirmed through Session Note Review',
+        'Impact & Outcomes Form'
+    ]
+
+    # 2. Create Regex patterns
+    # We join terms with '|' (OR). We use re.escape to handle symbols like '&' or '.' safely.
+    no_doc_pattern = '|'.join([re.escape(x) for x in no_doc_terms])
+    documented_pattern = '|'.join([re.escape(x) for x in documented_terms])
+
+    # 3. Define Conditions
+
+    # Check if string contains "Requested on eCenter" OR is null/empty
+    no_doc_condition = (
+            nbs_data[col_neo_attribution_source].str.contains(no_doc_pattern, case=False, na=False) |
+            nbs_data[col_neo_attribution_source].isin(['', 'NaN', np.nan])
+    )
+
+    # Note: For wrong_doc, we usually want to ensure it DOES NOT contain the "empty" markers.
+    # We keep .isin() here because checking for specific "empty" strings is usually exact.
+    wrong_doc_condition = (
+                              ~nbs_data[col_neo_attribution_source].isin(['', 'NaN', np.nan])
+                          ) & (
+                              nbs_data[col_neo_affirmation].isin(['', 'NaN', np.nan])
+                          ) & (
+                              nbs_data[col_neo_attribution_statement].isin(['', 'NaN', np.nan])
+                          )
+
+    # Check if source contains any of the documented terms
+    documented_condition = (
+            nbs_data[col_neo_attribution_source].str.contains(documented_pattern, case=False, na=False)
+            & (
+                    ~nbs_data[col_neo_attribution_statement].isin(['', 'NaN', np.nan])
+                    |
+                    ~nbs_data[col_neo_affirmation].isin(['', 'NaN', np.nan])
+            )
+    )
+
+    choices = [
+        'Not Documented',
+        'Affirmation Missing',
+        'Documented'
+    ]
+
+    nbs_data[col_out_documentation_level] = np.select(
+        condlist=[no_doc_condition, wrong_doc_condition, documented_condition],
+        choicelist=choices,
+        default='Not Determined'
+    )
+
+    return nbs_data
+'''
+
+
+
+
+'''
+def tag_documentation_level(
+    nbs_data: pd.DataFrame,
+    col_neo_attribution_source:str="Attribution Source",
+    col_neo_affirmation:str="Affirmation",
+    col_neo_attribution_statement:str="Attribution Statement",
+    col_out_documentation_level:str="Documentation Level"
+):
+    no_doc_condition = (
+    nbs_data[col_neo_attribution_source].isin(['Requested on eCenter', '', 'NaN', np.nan])
+    )
+
+    wrong_doc_condition = (
+        ~nbs_data[col_neo_attribution_source].isin(['', 'NaN', np.nan])
+    ) & (
+        nbs_data[col_neo_affirmation].isin(['', 'NaN', np.nan])
+    ) & (
+        nbs_data[col_neo_attribution_statement].isin(['', 'NaN', np.nan])
+        )
+    
+    documented_condition = (
+        nbs_data[col_neo_attribution_source].isin(['eCenter', 'Email from Client', 'Director Confirmed through Session Note Review', 'Impact & Outcomes Form']) 
+        & (
+            ~nbs_data[col_neo_attribution_statement].isin(['', 'NaN', np.nan])
+            |
+            ~nbs_data[col_neo_affirmation].isin(['', 'NaN', np.nan])
+        )
+    )
+    
+    choices = [
+        'Not Documented',
+        'Affirmation Missing',
+        'Documented'
+    ]
+
+    nbs_data[col_out_documentation_level] = np.select(
+        condlist=[no_doc_condition, wrong_doc_condition, documented_condition], 
+        choicelist=choices, 
+        default='Not Determined'
+    )
+'''
+
+
+def filter_valid_milestone(
+        group: pd.DataFrame,
+        col_neo_center: str = "Center",
+        col_neo_client_id: str = "Client ID",
+        col_neo_milestone_date: str = "Milestone Date",
+        col_neo_attribution_source: str = "Attribution Source",
+        col_neo_attribution_date: str = "Attribution Date",
+        col_neo_affirmation: str = "Affirmation",
+        col_neo_milestone_type: str = "Milestone Type",
+        col_neo_reportable: str = "Reportable?",
+        business_start_impact_val: str = "Business Start Impact",
+        business_established_val: str = "Business Established",
+        log_filtering_file: bool = "",
+):
+    log_file = None
+    blank_vals = ["", " ", "NaN", "nan", np.nan]
+    """
+    if log_filtering_file:
+        log_file_drop = open(log_filtering_file + "drop.txt", 'a')
+        log_file_keep = open(log_filtering_file + "keep.txt", 'a')
+    """
+    should_return = False
+    if group.shape[0] == 1:
+        milestone_type = group[col_neo_milestone_type].iloc[0]
+
+        # Only one entry for this client, check if it can be counted
+        if milestone_type == business_established_val:
+            """
+            log_file_drop.write(
+                f"Dropped milestone: \n{tabulate(group, headers='keys', tablefmt='grid', showindex=False)}\n")
+            """
+            # Drop the milestone
+            should_return = False
+            # print(f"Dropped milestone: {group}")
+        elif milestone_type == business_start_impact_val:
+            # Keep the milestone
+            """
+            log_file_keep.write(
+                f"Kept milestone: \n{tabulate(group, headers='keys', tablefmt='grid', showindex=False)}\n")
+            """
+            should_return = True
+
+        """
+        if log_file_drop is not None:
+            log_file_drop.close()
+            log_file_keep.close()
+        """
+
+        if should_return:
+            return group
+        else:
+            return None
+    else:
+        has_impact_milestone = (group[col_neo_milestone_type] == business_start_impact_val).any()
+
+        combined_milestone = None
+        if has_impact_milestone:
+            # We have a valid business start impact
+            should_return = True
+
+            # Filter the group such that we only use the impact milestones for the attribution and affirmation data
+            # This ensures that we do not take documented status from other milestones like business established
+            impact_group = group[group[col_neo_milestone_type] == business_start_impact_val]
+
+            """
+            log_file_keep.write(
+                f"Kept milestone: \n{tabulate(group, headers='keys', tablefmt='grid', showindex=False)}\n")
+            """
+
+            affirmation_concat_logic = lambda x: ', '.join(x.astype(str)[~x.isin(blank_vals)].unique())
+            attribution_source_concat_logic = lambda x: ', '.join(x.astype(str)[~x.isin(blank_vals)].unique())
+
+            # Get all of the attribution sources
+            group[col_neo_attribution_source]
+
+            milestone_type_concat_logic = lambda x: ', '.join(x.astype(str)[~x.isin(blank_vals)].unique())
+            milestone_date_concat_logic = lambda x: ', '.join(x.astype(str)[~x.isin(blank_vals)].unique())
+            attribution_date_concat_logic = lambda x: ', '.join(x.astype(str)[~x.isin(blank_vals)].unique())
+            reportable_concat_logic = lambda x: x.max(skipna=True)
+
+            combined_milestone = pd.DataFrame({
+                col_neo_client_id: [group[col_neo_client_id].iloc[0]],
+                col_neo_center: [group[col_neo_center].iloc[0]],
+                col_neo_milestone_type: [milestone_type_concat_logic(group[col_neo_milestone_type])],
+                col_neo_affirmation: [affirmation_concat_logic(impact_group[col_neo_affirmation])],
+                col_neo_attribution_source: [attribution_source_concat_logic(impact_group[col_neo_attribution_source])],
+                col_neo_milestone_date: [milestone_date_concat_logic(group[col_neo_milestone_date])],
+                col_neo_attribution_date: [attribution_date_concat_logic(group[col_neo_attribution_date])],
+                col_neo_reportable: [reportable_concat_logic(group[col_neo_reportable])]
+            })
+        else:
+            # print(f"Dropped milestone: {group}")
+            """
+            log_file_drop.write(
+                f"Dropped milestone: \n{tabulate(group, headers='keys', tablefmt='grid', showindex=False)}\n")
+            """
+            should_return = False
+
+        """
+        if log_file is not None:
+            log_file_drop.close()
+            log_file_keep.close()
+        """
+
+        if should_return:
+            return combined_milestone
+        else:
+            return None
+
+def sanitize_nbs_data(
+        df: pd.DataFrame,
+        col_neo_center:str,
+        col_neo_client_id: str,
+        col_neo_milestone_date: str,
+        col_neo_attribution_date: str,
+        col_neo_attribution_source: str,
+        col_neo_affirmation: str,
+        col_neo_milestone_type: str,
+        col_out_documentation_level: str,
+        col_neo_reportable: str,
+        business_start_impact_val: str,
+        business_established_val: str,
+) -> pd.DataFrame:
+    # Turn the neoserra center names into the correct ones for visualization
+    clean_center_name(df)
+
+    # Collapse any duplicate milestones within the data, passing in the variable columns
+    # Note: include_groups=False prevents the grouping key from being included in the applied result
+
+    # Remove rows with this
+    df[col_neo_attribution_source] = df[col_neo_attribution_source].str.strip() #pyright:ignore
+    df = df[df[col_neo_attribution_source] != "Not attributed to center"]
+
+    # If the input is empty, return an empty dataframe with the expected columns
+    if df.empty:
+        df_empty = pd.DataFrame(columns=df.columns.tolist() + [col_out_documentation_level])
+        return df_empty
+
+    df_clean = df.groupby(col_neo_client_id).apply(
+        filter_valid_milestone,
+        include_groups=True,
+        col_neo_center=col_neo_center,
+        col_neo_client_id=col_neo_client_id,
+        col_neo_milestone_date=col_neo_milestone_date,
+        col_neo_attribution_source=col_neo_attribution_source,
+        col_neo_attribution_date=col_neo_attribution_date,
+        col_neo_affirmation=col_neo_affirmation,
+        col_neo_milestone_type=col_neo_milestone_type,
+        col_neo_reportable=col_neo_reportable,
+        business_start_impact_val=business_start_impact_val,
+        business_established_val=business_established_val
+    )
+
+    # If all groups were filtered out, df_clean will be empty and reset_index will fail
+    if df_clean.empty:
+        df_empty = pd.DataFrame(columns=df.columns.tolist() + [col_out_documentation_level])
+        return df_empty
+
+    # Move the Client ID back to the columns (it becomes the index after groupby)
+    # Drop just tells it to lose the index copy of Client ID and keep the row based copy included with include_groups=True
+    df_clean = df_clean.reset_index(level=col_neo_client_id, drop=True)
+
+    # Tag the documentation level
+    df_clean = tag_documentation_level(
+        df_clean,
+        col_neo_attribution_source=col_neo_attribution_source,
+        col_neo_affirmation=col_neo_affirmation,
+        col_out_documentation_level=col_out_documentation_level
+    )
+
+    # Can be uncommented to drop any not determined documentation levels. It's good to keep this in as it will show you
+    # if something may have slipped through the cracks.
+    # df_clean = df_clean[df_clean[col_out_documentation_level] != "Not Determined"]
+
+    # Fill in any blanks to allow proper visualization of nan values
+    df_clean[col_neo_attribution_source] = df_clean[col_neo_attribution_source].fillna("Blank").astype(str).replace(['', ' ', 'nan', 'NaN'], "Blank") #pyright:ignore
+
+    df_clean = df_clean.sort_values(by=col_neo_center)
+
+    return df_clean #pyright:ignore
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument("-f", "--file",
+                        type=str,
+                        required=True,
+                        help="The data file to de-duplicate and tag with documentation levels")
+
+    parser.add_argument("-o", "--out",
+                        type=str,
+                        required=True,
+                        help="The csv file to write the de-duplicated data to")
+
+    parser.add_argument("-m", "--mapping",
+                        type=str,
+                        required=False,
+                        default="",
+                        help="The path to a JSON file with all of the columns in the DEFAULT_CONFIG of this script which can be used to override column name mappings")
+
+    args = parser.parse_args()
+
+    if args.mapping:
+        NEOSERRA_COLUMNS.apply_json_mapping(args.mapping)
+        OUT_COLUMNS.apply_json_mapping(args.mapping)
+
+    # Load Data
+    nbs_df = pd.read_csv(args.file, parse_dates=[NEOSERRA_COLUMNS.milestone_date, NEOSERRA_COLUMNS.attribution_date])
+
+    # Process Data using the new library function
+    nbs_df = sanitize_nbs_data(
+        nbs_df,
+        col_neo_center=NEOSERRA_COLUMNS.center,
+        col_neo_client_id=NEOSERRA_COLUMNS.client_id,
+        col_neo_milestone_date=NEOSERRA_COLUMNS.milestone_date,
+        col_neo_attribution_date=NEOSERRA_COLUMNS.attribution_date,
+        col_neo_attribution_source=NEOSERRA_COLUMNS.milestone_attribution_source,
+        col_neo_affirmation=NEOSERRA_COLUMNS.milestone_affirmation,
+        col_neo_milestone_type=NEOSERRA_COLUMNS.milestone_type_name,
+        col_out_documentation_level=OUT_COLUMNS.milestone_documentation_level,
+        col_neo_reportable=NEOSERRA_COLUMNS.reportable,
+        business_start_impact_val=NEOSERRA_COLUMNS.business_start_impact_val,
+        business_established_val=NEOSERRA_COLUMNS.business_established_val
+    )
+
+    # Save Data
+    nbs_df.to_csv(args.out, index=False)
+    print(f"DONE!")

milestone_attribution_dataset_module/milestone_attribution_dataset_module/shared.py

@@ -0,0 +1,76 @@
+import pandas as pd
+import re
+import numpy as np
+
+def tag_documentation_level(
+        nbs_data: pd.DataFrame,
+        col_neo_attribution_source: str = "Attribution Source",
+        col_neo_affirmation: str = "Affirmation",
+        col_out_documentation_level: str = "Documentation Level"
+):
+    # Terms for the attribution source that if found by individually constitute no documentation
+    no_doc_terms = [
+        'Director Confirmed through Session Note Review',
+        'Requested on eCenter'
+    ]
+
+    # Terms that if seen, even if beside something in no_doc_terms, will constitute a documentable attribution source
+    documented_terms = [
+        'eCenter',
+        'Email from Client',
+        'Impact & Outcomes Form',
+        'Quarterly Assessment Survey'
+    ]
+
+    no_doc_pattern = '|'.join([re.escape(x) for x in no_doc_terms])
+    documented_pattern = '|'.join([re.escape(x) for x in documented_terms])
+
+    # Remove Requested on eCenter so it doesn't trigger false positives.
+    clean_source = nbs_data[col_neo_attribution_source].astype(str).str.replace(no_doc_pattern, '', regex=True)
+
+    def is_populated(series):
+        return series.astype(str).str.strip().replace({'nan': '', 'NaN': ''}) != ''
+
+    has_affirmation = is_populated(nbs_data[col_neo_affirmation])
+    has_raw_source = is_populated(nbs_data[col_neo_attribution_source])
+
+    # Check if the CLEAN source contains a valid term "Email from Client"
+    has_valid_term = clean_source.str.contains(documented_pattern, case=False, na=False)
+
+    # PRIORITY 1: Documented
+    # Has a valid term has affirmation
+    documented_condition = (
+            has_valid_term &
+            has_affirmation
+    )
+
+    # PRIORITY 2: Affirmation Missing
+    # Has a valid term  BUT is missing affirmation.
+    wrong_doc_condition = (
+            has_valid_term &
+            (~has_affirmation)
+    )
+
+    # PRIORITY 3: Not Documented
+    # Matches "Requested..." OR Source is empty
+    # We don't need to check clean_source here. If it had a valid term,
+    # it would have been caught by Priority 1 or 2.
+    no_doc_condition = (
+            nbs_data[col_neo_attribution_source].str.contains(no_doc_pattern, case=False, na=False) |
+            (~has_raw_source)
+    )
+
+    choices = [
+        'Documented',  # Priority 1
+        'Affirmation Missing',  # Priority 2
+        'Not Documented'  # Priority 3
+    ]
+
+    # Apply the condition across the dataset creating a new column
+    nbs_data[col_out_documentation_level] = np.select(
+        condlist=[documented_condition, wrong_doc_condition, no_doc_condition],
+        choicelist=choices,
+        default='Not Determined'
+    )
+
+    return nbs_data

milestone_attribution_dataset_module/pyproject.toml

@@ -0,0 +1,12 @@
+# dataset/pyproject.toml
+[build-system]
+requires = ["setuptools", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "milestone_attribution_dataset_module"
+version = "0.1.0"
+description = "Internal PASBDC graph making scripts used to generate figures for the milestone attribution analysis."
+
+[tool.setuptools]
+packages = ["milestone_attribution_dataset_module"]