প্রজেক্ট: e-commerce ড্যাশবোর্ড

import pandas as pd
import numpy as np
from datetime import datetime, timedelta

np.random.seed(42)

CATEGORIES = ['Mobile','Fashion','Beauty','Home','Books','Sports','Toys']
CITIES = ['Dhaka','Chattogram','Sylhet','Rajshahi','Khulna','Barisal','Rangpur']
CHANNELS = ['Facebook','Google','Direct','Email','Referral','Organic']

def generate_orders(n_days=365, base_orders=50):
    """১২ মাসের synthetic order data — seasonality সহ।"""
    rows = []
    start = datetime(2024, 1, 1)
    for d in range(n_days):
        date = start + timedelta(days=d)
        # seasonality — Eid spike (Apr/Jun), Pohela Boishakh (Apr 14)
        month = date.month
        seasonal = 1.0
        if month in [4, 6]:    seasonal = 1.6  # Eid
        if month in [11, 12]:  seasonal = 1.4  # winter shopping
        if date.weekday() in [4, 5]:  seasonal *= 1.2  # weekend
        # daily orders Poisson
        n_orders = np.random.poisson(base_orders * seasonal)
        for _ in range(n_orders):
            cat = np.random.choice(CATEGORIES,
                p=[0.30,0.22,0.15,0.12,0.08,0.08,0.05])
            city = np.random.choice(CITIES,
                p=[0.45,0.20,0.10,0.08,0.07,0.05,0.05])
            channel = np.random.choice(CHANNELS,
                p=[0.30,0.25,0.20,0.10,0.10,0.05])
            qty = np.random.randint(1, 4)
            base_price = {'Mobile':25000,'Fashion':1500,'Beauty':800,
                          'Home':3000,'Books':400,'Sports':2000,'Toys':1200}[cat]
            price = base_price * np.random.uniform(0.7, 1.3)
            customer_id = f'C{np.random.randint(1, 5000):05d}'
            rows.append({
                'date': date,
                'order_id': f'O{len(rows)+1:07d}',
                'customer_id': customer_id,
                'category': cat,
                'city': city,
                'channel': channel,
                'quantity': qty,
                'unit_price': round(price, 2),
                'gmv': round(qty * price, 2),
            })
    return pd.DataFrame(rows)

def generate_marketing(n_days=365):
    """Daily marketing spend — channel-wise।"""
    rows = []
    start = datetime(2024, 1, 1)
    for d in range(n_days):
        date = start + timedelta(days=d)
        month = date.month
        boost = 2.0 if month in [4, 6, 11, 12] else 1.0
        for ch in ['Facebook','Google','Email']:
            base = {'Facebook':50000,'Google':80000,'Email':5000}[ch]
            spend = base * boost * np.random.uniform(0.8, 1.2)
            rows.append({'date': date, 'channel': ch,
                         'spend': round(spend, 2)})
    return pd.DataFrame(rows)

if __name__ == '__main__':
    orders = generate_orders()
    marketing = generate_marketing()
    orders.to_csv('orders.csv', index=False)
    marketing.to_csv('marketing.csv', index=False)
    print(f'Generated {len(orders)} orders, {len(marketing)} marketing rows')

import pandas as pd
import numpy as np

def total_gmv(df):
    return df['gmv'].sum()

def total_orders(df):
    return len(df)

def aov(df):
    return df['gmv'].sum() / max(len(df), 1)

def unique_customers(df):
    return df['customer_id'].nunique()

def cac(orders_df, marketing_df):
    """Marketing spend / new customers."""
    spend = marketing_df['spend'].sum()
    new_cust = orders_df['customer_id'].nunique()
    return spend / max(new_cust, 1)

def repeat_rate(df):
    """% customers with 2+ orders."""
    counts = df.groupby('customer_id').size()
    return (counts >= 2).sum() / len(counts) * 100

def cohort_retention(df):
    """First-purchase month → retention curve."""
    df = df.copy()
    df['order_month'] = df['date'].dt.to_period('M')
    first = df.groupby('customer_id')['order_month'].min().rename('cohort')
    df = df.merge(first, on='customer_id')
    df['cohort_idx'] = (df['order_month'] - df['cohort']).apply(lambda x: x.n)
    cohort = df.groupby(['cohort','cohort_idx'])['customer_id'].nunique().unstack()
    pct = cohort.div(cohort[0], axis=0) * 100
    return pct.round(1)

def daily_gmv(df):
    return df.groupby('date')['gmv'].sum().reset_index()

def category_gmv(df):
    return df.groupby('category')['gmv'].sum().sort_values(
        ascending=False).reset_index()

def city_gmv(df):
    return df.groupby('city')['gmv'].sum().sort_values(
        ascending=False).reset_index()

def channel_performance(orders_df, marketing_df):
    """Channel-wise revenue, orders, spend, CAC."""
    rev = orders_df.groupby('channel').agg(
        orders=('order_id','count'),
        revenue=('gmv','sum'),
        customers=('customer_id','nunique')
    ).reset_index()
    spend = marketing_df.groupby('channel')['spend'].sum().reset_index()
    out = rev.merge(spend, on='channel', how='left').fillna(0)
    out['cac'] = out['spend'] / out['customers'].replace(0, 1)
    out['roas'] = out['revenue'] / out['spend'].replace(0, 1)
    return out

import plotly.express as px
import plotly.graph_objects as go

def chart_daily_gmv(df):
    fig = px.line(df, x='date', y='gmv',
                  title='Daily GMV',
                  labels={'gmv':'GMV (BDT)','date':'Date'})
    fig.update_traces(line=dict(color='#2563eb', width=2))
    fig.update_layout(template='plotly_white', height=400)
    return fig

def chart_category(df):
    fig = px.bar(df, x='gmv', y='category', orientation='h',
                 title='Category-wise GMV',
                 labels={'gmv':'GMV (BDT)','category':''},
                 color='gmv', color_continuous_scale='Blues')
    fig.update_layout(template='plotly_white', height=400,
                      yaxis={'categoryorder':'total ascending'})
    return fig

def chart_city(df):
    fig = px.bar(df, x='city', y='gmv',
                 title='City-wise GMV',
                 color='gmv', color_continuous_scale='Greens')
    fig.update_layout(template='plotly_white', height=400)
    return fig

def chart_channel(df):
    fig = go.Figure()
    fig.add_trace(go.Bar(x=df['channel'], y=df['revenue'],
                         name='Revenue', marker_color='#2563eb',
                         yaxis='y'))
    fig.add_trace(go.Scatter(x=df['channel'], y=df['roas'],
                             mode='lines+markers', name='ROAS',
                             marker_color='#dc2626', yaxis='y2'))
    fig.update_layout(
        title='Channel — Revenue vs ROAS',
        yaxis=dict(title='Revenue (BDT)', side='left'),
        yaxis2=dict(title='ROAS', side='right', overlaying='y'),
        template='plotly_white', height=400,
        legend=dict(x=0.01, y=0.99))
    return fig

def chart_cohort(retention_df):
    fig = px.imshow(retention_df.values,
                    labels=dict(x='Months Since First Purchase',
                                y='Cohort', color='Retention %'),
                    x=retention_df.columns,
                    y=[str(c) for c in retention_df.index],
                    color_continuous_scale='RdYlGn',
                    aspect='auto',
                    text_auto='.0f')
    fig.update_layout(title='Cohort Retention Heatmap',
                      template='plotly_white', height=500)
    return fig

import streamlit as st
import pandas as pd
import plotly.express as px

# local modules
from data_loader import generate_orders, generate_marketing
from kpi import (total_gmv, total_orders, aov, unique_customers, cac,
                 repeat_rate, cohort_retention, daily_gmv,
                 category_gmv, city_gmv, channel_performance)
from charts import (chart_daily_gmv, chart_category, chart_city,
                    chart_channel, chart_cohort)

# ─────── Page config ───────
st.set_page_config(
    page_title='Daraz-Style Dashboard',
    page_icon='🛍️',
    layout='wide',
    initial_sidebar_state='expanded'
)

# ─────── Cache ───────
@st.cache_data
def load_data():
    orders = generate_orders()
    marketing = generate_marketing()
    orders['date'] = pd.to_datetime(orders['date'])
    marketing['date'] = pd.to_datetime(marketing['date'])
    return orders, marketing

orders, marketing = load_data()

# ─────── Sidebar ───────
with st.sidebar:
    st.image('https://via.placeholder.com/200x60?text=YourBrand', width=200)
    st.markdown('### 🎯 Filters')

    min_d = orders['date'].min().date()
    max_d = orders['date'].max().date()
    date_range = st.date_input('Date Range', value=(min_d, max_d),
                                min_value=min_d, max_value=max_d)

    cities = ['All'] + sorted(orders['city'].unique().tolist())
    sel_city = st.selectbox('City', cities)

    cats = sorted(orders['category'].unique().tolist())
    sel_cats = st.multiselect('Categories', cats, default=cats)

    st.markdown('---')
    st.caption('Built with Streamlit + Plotly')
    st.caption('© ABCL TECH 2025')

# ─────── Apply filters ───────
df = orders.copy()
if len(date_range) == 2:
    df = df[(df['date'].dt.date >= date_range[0]) &
            (df['date'].dt.date <= date_range[1])]
if sel_city != 'All':
    df = df[df['city'] == sel_city]
df = df[df['category'].isin(sel_cats)]

mkt = marketing.copy()
if len(date_range) == 2:
    mkt = mkt[(mkt['date'].dt.date >= date_range[0]) &
              (mkt['date'].dt.date <= date_range[1])]

# ─────── Title ───────
st.title('🛍️ Daraz-Style E-commerce Dashboard')
st.caption('A capstone project — ABCL TECH Data Science Course')

# ─────── KPI cards ───────
col1, col2, col3, col4 = st.columns(4)
col1.metric('GMV', f'৳ {total_gmv(df)/1e7:.2f} Cr', '+১২%')
col2.metric('Orders', f'{total_orders(df):,}', '+৮%')
col3.metric('AOV', f'৳ {aov(df):,.0f}', '+২%')
col4.metric('Unique Customers', f'{unique_customers(df):,}', '+১৫%')

col5, col6, col7, col8 = st.columns(4)
col5.metric('CAC', f'৳ {cac(df, mkt):,.0f}', '-৫%', delta_color='inverse')
col6.metric('Repeat Rate', f'{repeat_rate(df):.1f}%', '+৩%')
col7.metric('Total Marketing Spend', f'৳ {mkt["spend"].sum()/1e7:.2f} Cr')
col8.metric('Avg ROAS', f'{df["gmv"].sum()/max(mkt["spend"].sum(),1):.2f}x')

st.markdown('---')

# ─────── Tabs ───────
tab1, tab2, tab3, tab4, tab5 = st.tabs([
    '📈 Trends', '🛒 Categories', '🗺️ Geography',
    '📡 Channels', '👥 Cohorts'])

with tab1:
    st.subheader('Daily GMV Trend')
    daily = daily_gmv(df)
    st.plotly_chart(chart_daily_gmv(daily), use_container_width=True)

    # 7-day moving average overlay
    daily['ma7'] = daily['gmv'].rolling(7, min_periods=1).mean()
    fig2 = px.line(daily, x='date', y=['gmv','ma7'],
                   title='GMV with 7-day Moving Average',
                   labels={'value':'GMV (BDT)','variable':''})
    fig2.update_layout(template='plotly_white', height=400)
    st.plotly_chart(fig2, use_container_width=True)

with tab2:
    st.subheader('Category Performance')
    cat_df = category_gmv(df)
    st.plotly_chart(chart_category(cat_df), use_container_width=True)

    st.subheader('Top 10 Best-Selling SKUs')
    top10 = df.groupby('category').agg(
        orders=('order_id','count'),
        revenue=('gmv','sum'),
        avg_price=('unit_price','mean')
    ).sort_values('revenue', ascending=False).head(10)
    st.dataframe(top10.style.format({
        'revenue':'৳{:,.0f}','avg_price':'৳{:,.0f}'}),
        use_container_width=True)

with tab3:
    st.subheader('Geographic Distribution')
    city_df = city_gmv(df)
    st.plotly_chart(chart_city(city_df), use_container_width=True)

    st.subheader('City × Category Heatmap')
    pivot = df.pivot_table(index='city', columns='category',
                            values='gmv', aggfunc='sum').fillna(0)
    fig3 = px.imshow(pivot.values, x=pivot.columns, y=pivot.index,
                     color_continuous_scale='Blues',
                     aspect='auto', text_auto='.2s',
                     labels=dict(color='GMV'))
    fig3.update_layout(template='plotly_white', height=500)
    st.plotly_chart(fig3, use_container_width=True)

with tab4:
    st.subheader('Marketing Channel Performance')
    ch_df = channel_performance(df, mkt)
    st.plotly_chart(chart_channel(ch_df), use_container_width=True)

    st.subheader('Channel Details')
    st.dataframe(ch_df.style.format({
        'revenue':'৳{:,.0f}','spend':'৳{:,.0f}',
        'cac':'৳{:,.0f}','roas':'{:.2f}x'}),
        use_container_width=True)

    st.info('🔍 ROAS < 1 মানে loss-making channel।')

with tab5:
    st.subheader('Cohort Retention Heatmap')
    st.write('প্রতিটি row একটি প্রথম-purchase month-এর cohort। '
             'Column — সেই cohort-এর কত % পরবর্তী মাসগুলোয় ফিরে এসেছেন।')
    if len(df) > 100:
        retention = cohort_retention(df)
        st.plotly_chart(chart_cohort(retention), use_container_width=True)

        # M+1 retention summary
        if 1 in retention.columns:
            avg_m1 = retention[1].mean()
            st.metric('Average M+1 Retention', f'{avg_m1:.1f}%')
    else:
        st.warning('Not enough data — filters relax করুন।')

# ─────── Footer ───────
st.markdown('---')
st.caption('Data is synthetic — for demo purposes only.')