modules/campaign_analysis.py

from flask import Blueprint, render_template, request, flash, redirect, url_for
import pandas as pd
import numpy as np
import json

campaign_analysis_bp = Blueprint('campaign_analysis', __name__, url_prefix='/campaign')

_current_sales_df = None
_current_campaign_df = None

def normalize(s):
    return str(s).strip().lower()

# OPTIONAL: channel mapping between campaign and sales if desired
CHANNEL_MAP = {
    'whatsapp': 'online',
    'email': 'online',
    'mobile app': 'online',
    'tv': 'offline',
    # add more as needed
}

@campaign_analysis_bp.route('/', methods=['GET'])
def show_campaign_analysis():
    return render_template('campaign_analysis.html', title="Campaign Analysis Engine")

@campaign_analysis_bp.route('/upload', methods=['POST'])
def upload_files():
    global _current_sales_df, _current_campaign_df

    sales_file = request.files.get('sales_file')
    campaign_file = request.files.get('campaign_file')
    if not sales_file or not campaign_file:
        flash('Please upload both sales and campaign CSV files.')
        return redirect(url_for('campaign_analysis.show_campaign_analysis'))

    try:
        _current_sales_df = pd.read_csv(sales_file, parse_dates=['date'], dayfirst=True)
        _current_campaign_df = pd.read_csv(campaign_file, parse_dates=['start_date', 'end_date'], dayfirst=True)
    except Exception as e:
        flash(f'Error reading CSVs: {e}')
        return redirect(url_for('campaign_analysis.show_campaign_analysis'))

    try:
        analysis = analyze_campaigns(_current_sales_df, _current_campaign_df)
        plot_data = make_plot_data(analysis['campaigns'])
    except Exception as e:
        flash(f'Error during analysis: {e}')
        return redirect(url_for('campaign_analysis.show_campaign_analysis'))

    return render_template('campaign_analysis.html',
        title="Campaign Analysis Engine",
        campaigns=analysis['campaigns'],
        summary=analysis['summary'],
        columns=_current_campaign_df.columns.tolist(),
        plot_data=plot_data  # <--- DO NOT json.dumps here!
    )

def make_plot_data(campaigns):
    # For bar and line charts
    names = [c['name'] if c['name'] else c['id'] for c in campaigns]
    roi = [c['roi'] if c['roi']==c['roi'] else 0 for c in campaigns]
    uplift = [c['uplift_pct'] if c['uplift_pct']==c['uplift_pct'] else 0 for c in campaigns]
    revenue = [c['total_revenue'] for c in campaigns]
    start_dates = [str(c['start_date']) for c in campaigns]
    types = [c['name'] for c in campaigns]
    regions = [c['region'] for c in campaigns]

    # Pie chart of types
    from collections import Counter
    type_counts = Counter(types)
    region_counts = Counter(regions)

    return {
        "bar_uplift": {"x": names, "y": uplift},
        "bar_roi": {"x": names, "y": roi},
        "line_revenue": {"x": start_dates, "y": revenue, "names": names},
        "pie_type": {"labels": list(type_counts.keys()), "values": list(type_counts.values())},
        "pie_region": {"labels": list(region_counts.keys()), "values": list(region_counts.values())}
    }

def map_campaign_channel_to_sales(campaign_channel):
    # Normalize and map campaign channel to sales channel if needed
    ch = normalize(campaign_channel)
    return CHANNEL_MAP.get(ch, ch)

def analyze_campaigns(sales_df, campaign_df):
    """
    For each campaign, compute total units/revenue during the campaign period,
    estimate uplift vs. pre-campaign, and simple ROI.
    """
    sales_df['channel'] = sales_df['channel'].astype(str)
    campaign_results = []
    for _, camp in campaign_df.iterrows():
        # Split target_skus and iterate
        target_skus = [sku.strip() for sku in str(camp['target_skus']).split(',')]
        region = camp['region']
        # Map campaign channel to sales channel if needed
        campaign_channel = camp['channel']
        mapped_channel = map_campaign_channel_to_sales(campaign_channel)
        start_date = camp['start_date']
        end_date = camp['end_date']

        # Filter sales for this campaign
        sales_mask = (
            sales_df['sku'].isin(target_skus)
            & (sales_df['region'] == region)
            & (sales_df['channel'].str.lower() == mapped_channel)
            & (sales_df['date'] >= start_date)
            & (sales_df['date'] <= end_date)
        )
        sales_during = sales_df[sales_mask]
        total_units = int(sales_during['sales_units'].sum())
        total_revenue = float(sales_during['sales_revenue'].sum())

        # Uplift: compare average daily units in campaign vs. 14 days prior
        pre_mask = (
            sales_df['sku'].isin(target_skus)
            & (sales_df['region'] == region)
            & (sales_df['channel'].str.lower() == mapped_channel)
            & (sales_df['date'] >= (start_date - pd.Timedelta(days=14)))
            & (sales_df['date'] < start_date)
        )
        sales_pre = sales_df[pre_mask]
        days_campaign = (end_date - start_date).days + 1
        days_pre = (sales_pre['date'].max() - sales_pre['date'].min()).days + 1 if not sales_pre.empty else 1
        avg_daily_campaign = total_units / days_campaign if days_campaign > 0 else 0
        avg_daily_pre = sales_pre['sales_units'].sum() / days_pre if days_pre > 0 and not sales_pre.empty else 0
        uplift_pct = ((avg_daily_campaign - avg_daily_pre) / avg_daily_pre * 100) if avg_daily_pre > 0 else np.nan

        # ROI: (Incremental revenue - budget) / budget
        budget = camp.get('budget', np.nan)
        try:
            budget = float(budget)
        except Exception:
            budget = np.nan
        incremental_revenue = total_revenue - sales_pre['sales_revenue'].sum() if not sales_pre.empty else total_revenue
        roi = ((incremental_revenue - budget) / budget * 100) if pd.notna(budget) and budget > 0 else np.nan

        campaign_results.append({
            'id': camp.get('campaign_id', ''),
            'name': camp.get('type', ''),
            'sku': ','.join(target_skus),
            'region': region,
            'channel': campaign_channel,
            'start_date': start_date.date() if not pd.isnull(start_date) else '',
            'end_date': end_date.date() if not pd.isnull(end_date) else '',
            'budget': budget if not pd.isnull(budget) else '-',
            'total_units': total_units,
            'total_revenue': total_revenue,
            'uplift_pct': uplift_pct,
            'roi': roi
        })

    # Simple summary for display
    summary = {
        'total_campaigns': len(campaign_results),
        'total_revenue': float(np.nansum([c['total_revenue'] for c in campaign_results])),
        'avg_uplift_pct': float(np.nanmean([c['uplift_pct'] for c in campaign_results if not np.isnan(c['uplift_pct'])])) if campaign_results else 0,
        'avg_roi': float(np.nanmean([c['roi'] for c in campaign_results if not np.isnan(c['roi'])])) if campaign_results else 0,
    }
    return {'campaigns': campaign_results, 'summary': summary}