#!/usr/bin/env python3
"""达播业务三维度深度分析图表：达人 × 月度 × 平台"""
import openpyxl, glob, os, numpy as np
from datetime import datetime, timedelta
from collections import defaultdict
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import matplotlib.ticker as mticker

plt.rcParams['font.sans-serif'] = ['WenQuanYi Micro Hei', 'Noto Sans CJK SC', 'SimHei', 'DejaVu Sans']
plt.rcParams['axes.unicode_minus'] = False

# === Colors ===
C_BLUE = '#2563EB'; C_GOLD = '#F59E0B'; C_RED = '#EF4444'; C_GREEN = '#10B981'
C_PURPLE = '#8B5CF6'; C_TEAL = '#14B8A6'; C_PINK = '#EC4899'; C_ORANGE = '#F97316'
C_DARK = '#1F2937'; C_GRAY = '#6B7280'; C_LIGHT = '#F3F4F6'
PALETTE = [C_BLUE, C_GOLD, C_RED, C_PURPLE, C_GREEN, C_PINK, C_TEAL, C_ORANGE, '#6366F1', '#84CC16']
PALETTE_SOFT = ['#BFDBFE','#FDE68A','#FECACA','#C4B5FD','#A7F3D0','#FBCFE8','#99F6E4','#FDBA74','#C7D2FE','#BEF264']

# === Data loading ===
files = glob.glob('/root/.openclaw/media/inbound/*3403f15a*')
wb = openpyxl.load_workbook(files[0])
ws = wb['Sheet1']

def excel_date_to_str(val):
    if val is None: return None
    if isinstance(val, datetime): return val.strftime('%Y-%m-%d')
    if isinstance(val, str): return val
    if isinstance(val, (int, float)):
        try: return (datetime(1899,12,30)+timedelta(days=int(val))).strftime('%Y-%m-%d')
        except: return str(val)
    return str(val)

data = []
for row in ws.iter_rows(min_row=2, max_row=ws.max_row, values_only=True):
    name = str(row[0]).strip() if row[0] else ''
    if '合计' in name or '总计' in name or row[3] is None: continue
    data.append({
        'name': name, 'date': excel_date_to_str(row[1]),
        'platform': str(row[2]).strip() if row[2] else '',
        'orders': row[3], 'gmv': row[4] or 0, 'refund_orders': row[5] or 0,
        'refund_amount': row[6] or 0, 'gsv': row[10] or 0,
    })

# Name norm
name_map = {
    '晚柠也是个妈妈了':'晚柠','晚柠':'晚柠','念妈讲学习规划':'念妈','念妈':'念妈',
    '学霸三人行':'学霸三人行','学霸老王':'学霸老王',
    '开心妈妈学习宝藏':'开心妈妈','开心爸育儿':'开心爸',
    '小花生网':'小花生网','小花生':'小花生网','盈姐':'盈姐',
    '百克力':'百克力','亮爸':'亮爸',
    '万物分销':'万物分销','万物内购':'万物分销','万物团购':'万物分销',
    '小小鹰萱妈':'小小鹰萱妈','老狼聊育儿':'老狼聊育儿',
    '海淀妈妈优选':'海淀妈妈优选','神奇瓜妈聊成长':'神奇瓜妈',
    '宣儿妈妈':'宣儿妈妈','宣儿麻麻':'宣儿妈妈',
    '四个娃的组合生活':'四个娃的组合生活','肆个葫芦娃的妈':'四个娃的组合生活',
    '小暖阿姨慢一点':'小暖阿姨',
}
for r in data: r['norm_name'] = name_map.get(r['name'], r['name'])

# Month mapping for raw date strings
MONTH_ORDER = ['2025-09','2025-10','2025-11','2025-12','2026-01','2026-02','2026-03','2026-04','2026-05']
MONTH_LABELS = ['9月','10月','11月','12月','1月','2月','3月','4月','5月']

def parse_month(date_str):
    """Parse date to standardized YYYY-MM"""
    if not date_str: return None
    d = date_str[:7]
    remap = {'2026-09':'2025-09','2026-10':'2025-10','2026-11':'2025-11','2026-12':'2025-12',
             '2025-12':'2025-12'}
    return remap.get(d, d)

# Platform merge
def merge_platform(p):
    if '三开' in p: return '三开'
    if '小红书' in p: return '小红书'
    if '抖音' in p and '视频号' in p: return '抖音+视频号'
    if '视频号' in p: return '视频号'
    if '抖音' in p: return '抖音'
    if '分销' in p: return '分销'
    if '万物' in p: return '万物'
    if '公众号' in p: return '公众号'
    if '社群' in p: return '社群'
    if '微信小店' in p: return '微信小店'
    return p

# === Prepare cross-dimensional data ===
# 1. By influencer × month
inf_monthly = defaultdict(lambda: defaultdict(lambda: {'gmv':0,'gsv':0,'orders':0,'ref_ords':0}))
for r in data:
    m = parse_month(r['date'])
    if m: inf_monthly[r['norm_name']][m]['gmv'] += r['gmv']
    if m: inf_monthly[r['norm_name']][m]['gsv'] += r['gsv']
    if m: inf_monthly[r['norm_name']][m]['orders'] += r['orders']
    if m: inf_monthly[r['norm_name']][m]['ref_ords'] += r['refund_orders']

# 2. By platform × month
plat_monthly = defaultdict(lambda: defaultdict(lambda: {'gmv':0,'gsv':0,'orders':0,'ref_ords':0}))
for r in data:
    m = parse_month(r['date'])
    p = merge_platform(r['platform'])
    if m: plat_monthly[p][m]['gmv'] += r['gmv']
    if m: plat_monthly[p][m]['gsv'] += r['gsv']
    if m: plat_monthly[p][m]['orders'] += r['orders']
    if m: plat_monthly[p][m]['ref_ords'] += r['refund_orders']

# Get top influencers
inf_total = defaultdict(lambda: {'gmv':0})
for r in data: inf_total[r['norm_name']]['gmv'] += r['gmv']
TOP_N = 8
# Filter out empty names
inf_filtered = [(k,v) for k,v in inf_total.items() if k and k.strip()]
top_inf = [x[0] for x in sorted(inf_filtered, key=lambda x:x[1]['gmv'], reverse=True)[:TOP_N]]

# Get top platforms
plat_total = defaultdict(lambda: {'gmv':0})
for r in data: plat_total[merge_platform(r['platform'])]['gmv'] += r['gmv']
top_plat = [x[0] for x in sorted(plat_total.items(), key=lambda x:x[1]['gmv'], reverse=True) if x[1]['gmv']>0 and x[0].strip()]

print(f"Top influencers: {top_inf}")
print(f"Top platforms: {top_plat}")

# ============================================================
# FIGURE A: 达人维度深度分析 (2 rows × 3 cols)
# ============================================================
figA, axesA = plt.subplots(2, 3, figsize=(22, 14))
figA.patch.set_facecolor('#FAFBFC')

# A1: 达人月度GMV热力图
ax = axesA[0, 0]
heat = np.zeros((len(top_inf), len(MONTH_ORDER)))
for i, name in enumerate(top_inf):
    for j, m in enumerate(MONTH_ORDER):
        heat[i, j] = inf_monthly[name][m]['gmv'] / 10000

masked = np.ma.masked_equal(heat, 0)
im = ax.imshow(masked, cmap='YlOrRd', aspect='auto', vmin=0, vmax=150)
for i in range(len(top_inf)):
    for j in range(len(MONTH_ORDER)):
        v = heat[i, j]
        if v > 0:
            col = 'white' if v > 80 else C_DARK
            ax.text(j, i, f'{v:.0f}', ha='center', va='center', fontsize=8, fontweight='bold', color=col)
ax.set_xticks(range(len(MONTH_ORDER))); ax.set_xticklabels(MONTH_LABELS, fontsize=9)
ax.set_yticks(range(len(top_inf))); ax.set_yticklabels(top_inf, fontsize=9)
ax.set_title('达人月度 GMV 热力图 (万元)', fontsize=13, fontweight='bold', pad=10)
plt.colorbar(im, ax=ax, shrink=0.8, pad=0.02).set_label('万元', fontsize=9)

# A2: 达人月度退款率走势
ax = axesA[0, 1]
for idx, name in enumerate(top_inf):
    rates = []
    for m in MONTH_ORDER:
        s = inf_monthly[name][m]
        rates.append((s['ref_ords']/s['orders']*100) if s['orders']>0 else np.nan)
    ax.plot(range(len(MONTH_ORDER)), rates, 'o-', color=PALETTE[idx], linewidth=2, 
            markersize=7, markerfacecolor='white', markeredgewidth=2, label=name, alpha=0.9)

ax.axhline(y=40, color=C_RED, linestyle='--', alpha=0.3, linewidth=1)
ax.set_xticks(range(len(MONTH_ORDER))); ax.set_xticklabels(MONTH_LABELS, fontsize=9)
ax.set_title('达人月度退款率走势', fontsize=13, fontweight='bold', pad=10)
ax.set_ylabel('退款率 (%)', fontsize=10)
ax.legend(loc='upper left', fontsize=7.5, ncol=2, framealpha=0.9)
ax.grid(alpha=0.3)
ax.set_ylim(0, 85)

# A3: 达人综合评分雷达 (GMV, GSV, 订单量, 退款率倒数, 单均GMV, 场次)
ax = axesA[0, 2]
# Normalize metrics for radar
metrics = {}
for name in top_inf[:6]:
    s = inf_total[name]
    total_s = inf_monthly[name]
    all_orders = sum(v['orders'] for v in total_s.values())
    all_gmv = sum(v['gmv'] for v in total_s.values())
    all_gsv = sum(v['gsv'] for v in total_s.values())
    all_ref = sum(v['ref_ords'] for v in total_s.values())
    sessions = sum(1 for v in total_s.values() if v['orders']>0)
    metrics[name] = {
        'GMV': all_gmv,
        'GSV': all_gsv,
        '订单量': all_orders,
        '留存率': (1 - all_ref/all_orders)*100 if all_orders>0 else 100,
        '均单额': all_gmv/all_orders if all_orders>0 else 0,
        '场次': sessions,
    }

# Normalize
maxes = {}
for k in ['GMV','GSV','订单量','留存率','均单额','场次']:
    maxes[k] = max(metrics[name][k] for name in metrics)

categories = ['GMV', 'GSV', '订单量', '留存率', '均单额', '场次']
N_cat = len(categories)
angles = np.linspace(0, 2*np.pi, N_cat, endpoint=False).tolist()
angles += angles[:1]

for idx, (name, m) in enumerate(metrics.items()):
    values = [m[k]/maxes[k]*100 for k in categories]
    values += values[:1]
    ax.fill(angles, values, alpha=0.08, color=PALETTE[idx])
    ax.plot(angles, values, 'o-', color=PALETTE[idx], linewidth=2, markersize=5, label=name)

ax.set_xticks(angles[:-1]); ax.set_xticklabels(categories, fontsize=9, fontweight='bold')
ax.set_title('达人能力雷达图 (TOP6)', fontsize=13, fontweight='bold', pad=15)
ax.legend(loc='lower right', fontsize=7.5, bbox_to_anchor=(1.3,0))
ax.set_ylim(0, 110)

# A4: 达人GSV贡献占比 (按时间)
ax = axesA[1, 0]
# Stacked area chart of top5 + others
top5_names = top_inf[:5]
monthly_stacked = defaultdict(lambda: defaultdict(float))
for r in data:
    m = parse_month(r['date'])
    n = r['norm_name'] if r['norm_name'] in top5_names else '其他'
    if m: monthly_stacked[m][n] += r['gsv']/10000

areas = {}
for n in top5_names + ['其他']:
    areas[n] = [monthly_stacked[m].get(n,0) for m in MONTH_ORDER]

bottom = np.zeros(len(MONTH_ORDER))
for idx, name in enumerate(top5_names + ['其他']):
    vals = areas[name]
    ax.fill_between(range(len(MONTH_ORDER)), bottom, bottom+np.array(vals), 
                     color=PALETTE[idx], alpha=0.8, label=name, linewidth=0.5, edgecolor='white')
    bottom += np.array(vals)

ax.set_xticks(range(len(MONTH_ORDER))); ax.set_xticklabels(MONTH_LABELS, fontsize=9)
ax.set_title('月度 GSV 达人贡献占比 (万元)', fontsize=13, fontweight='bold', pad=10)
ax.set_ylabel('GSV (万元)', fontsize=10)
ax.legend(loc='upper left', fontsize=8, ncol=2, framealpha=0.9)
ax.grid(axis='y', alpha=0.3)

# A5: 达人单均GMV & 退款率散点 (by month bubble)
ax = axesA[1, 1]
for idx, name in enumerate(top_inf[:6]):
    x_vals, y_vals, s_vals = [], [], []
    for m in MONTH_ORDER:
        s = inf_monthly[name][m]
        if s['orders'] > 0:
            avg = s['gmv']/s['orders']
            rr = s['ref_ords']/s['orders']*100
            x_vals.append(avg)
            y_vals.append(rr)
            s_vals.append(s['orders']*12)
    if x_vals:
        ax.scatter(x_vals, y_vals, s=s_vals, color=PALETTE[idx], alpha=0.7, edgecolors='white', linewidth=1.2, label=name)
        for x,y,si in zip(x_vals, y_vals, s_vals):
            ax.annotate(name[:3], (x,y), fontsize=6.5, ha='center', va='bottom', color=C_GRAY, alpha=0.7)

ax.axhline(y=40, color=C_RED, linestyle='--', alpha=0.3)
ax.set_xlabel('单均GMV (元)', fontsize=10)
ax.set_ylabel('退款率 (%)', fontsize=10)
ax.set_title('达人效率矩阵：单均GMV × 退款率\n(气泡=月订单量, 每点=一个月)', fontsize=12, fontweight='bold', pad=10)
ax.legend(fontsize=7.5, loc='upper right', framealpha=0.9)
ax.grid(alpha=0.3)

# A6: 达人月度活跃度 (何时开播)
ax = axesA[1, 2]
# Binary heatmap - which months each influencer streamed
active = np.zeros((len(top_inf), len(MONTH_ORDER)))
for i, name in enumerate(top_inf):
    for j, m in enumerate(MONTH_ORDER):
        if inf_monthly[name][m]['orders'] > 0:
            active[i, j] = 1

ax.imshow(active, cmap='RdYlGn', aspect='auto', alpha=0.8)
for i in range(len(top_inf)):
    for j in range(len(MONTH_ORDER)):
        if active[i,j] == 1:
            ax.text(j, i, '●', ha='center', va='center', fontsize=14, color='#065F46', fontweight='bold')
        else:
            ax.text(j, i, '—', ha='center', va='center', fontsize=10, color='#D1D5DB')

# Color background by month phase
phase_colors = ['#FEF3C7','#FEF3C7','#FEF3C7','#FEF3C7','#DBEAFE','#DBEAFE','#FEE2E2','#FEE2E2','#D1FAE5']
for j in range(len(MONTH_ORDER)):
    ax.axvspan(j-0.5, j+0.5, alpha=0.12, color=phase_colors[j], zorder=0)

ax.set_xticks(range(len(MONTH_ORDER))); ax.set_xticklabels(MONTH_LABELS, fontsize=9)
ax.set_yticks(range(len(top_inf))); ax.set_yticklabels(top_inf, fontsize=9)
ax.set_title('达人月度开播活跃度', fontsize=13, fontweight='bold', pad=10)

figA.suptitle('维度一：达人维度深度分析', fontsize=17, fontweight='bold', y=1.01, color=C_DARK)
plt.tight_layout(pad=3.5)
outA = '/root/.openclaw/workspace/output/daren_dim_people.png'
figA.savefig(outA, dpi=180, bbox_inches='tight', facecolor=figA.get_facecolor())
plt.close()
print(f"✅ Figure A: {outA} ({os.path.getsize(outA)/1024:.0f} KB)")

# ============================================================
# FIGURE B: 月度维度深度分析 (2 rows × 3 cols)
# ============================================================
figB, axesB = plt.subplots(2, 3, figsize=(22, 14))
figB.patch.set_facecolor('#FAFBFC')

# Monthly totals computed from data
m_totals = {}
for r in data:
    m = parse_month(r['date'])
    if not m: continue
    if m not in m_totals:
        m_totals[m] = {'gmv':0,'gsv':0,'orders':0,'ref_ords':0,'ref_amt':0,'sessions':set()}
    m_totals[m]['gmv'] += r['gmv']
    m_totals[m]['gsv'] += r['gsv']
    m_totals[m]['orders'] += r['orders']
    m_totals[m]['ref_ords'] += r['refund_orders']
    m_totals[m]['ref_amt'] += r['refund_amount']
    m_totals[m]['sessions'].add(r['date'])

# B1: 月度核心指标面板
ax = axesB[0, 0]
x = np.arange(len(MONTH_ORDER))

# GMV bar + GSV bar overlay
gmv_arr = [m_totals[m]['gmv']/10000 if m in m_totals else 0 for m in MONTH_ORDER]
gsv_arr = [m_totals[m]['gsv']/10000 if m in m_totals else 0 for m in MONTH_ORDER]
refund_arr = np.array(gmv_arr) - np.array(gsv_arr)
order_arr = [m_totals[m]['orders'] if m in m_totals else 0 for m in MONTH_ORDER]

bars_b1 = ax.bar(x, gsv_arr, color=C_GREEN, alpha=0.85, label='GSV', zorder=3)
bars_b2 = ax.bar(x, refund_arr, bottom=gsv_arr, color=C_RED, alpha=0.5, label='退款', zorder=3)

for i in range(len(MONTH_ORDER)):
    if gmv_arr[i] > 0:
        ax.text(i, gmv_arr[i]+1, f'{gmv_arr[i]:.0f}万\n{order_arr[i]}单', ha='center', fontsize=8, fontweight='bold', color=C_DARK)

ax.set_xticks(x); ax.set_xticklabels(MONTH_LABELS, fontsize=9)
ax.set_title('月度 GMV 构成 & 订单量', fontsize=13, fontweight='bold', pad=10)
ax.set_ylabel('万元', fontsize=10)
ax.legend(fontsize=8)
ax.grid(axis='y', alpha=0.3, zorder=0)

# B2: 月度退款率 + GSV率 双轴
ax = axesB[0, 1]
ax2 = ax.twinx()
ref_rates = [(m_totals[m]['ref_ords']/m_totals[m]['orders']*100) if m in m_totals and m_totals[m]['orders']>0 else 0 for m in MONTH_ORDER]
gsv_rates = [(m_totals[m]['gsv']/m_totals[m]['gmv']*100) if m in m_totals and m_totals[m]['gmv']>0 else 0 for m in MONTH_ORDER]

ax.bar(x, ref_rates, 0.5, color=C_RED, alpha=0.25, zorder=2)
ax.plot(x, ref_rates, 'o-', color=C_RED, linewidth=2.5, markersize=9, zorder=4)
ax2.plot(x, gsv_rates, 's--', color=C_GREEN, linewidth=2.5, markersize=9, zorder=4)

for i, (r, g) in enumerate(zip(ref_rates, gsv_rates)):
    if r > 0:
        ax.annotate(f'{r:.0f}%', (i, r), textcoords="offset points", xytext=(0,12), ha='center', fontsize=9, fontweight='bold', color=C_RED)
    if g > 0:
        ax2.annotate(f'{g:.0f}%', (i, g), textcoords="offset points", xytext=(0,-16), ha='center', fontsize=9, fontweight='bold', color=C_GREEN)

ax.set_xticks(x); ax.set_xticklabels(MONTH_LABELS, fontsize=9)
ax.set_title('月度退款率 & GSV率 (净收入率)', fontsize=13, fontweight='bold', pad=10)
ax.set_ylabel('退款率 (%)', color=C_RED, fontsize=10)
ax2.set_ylabel('GSV率 (%)', color=C_GREEN, fontsize=10)
ax.tick_params(axis='y', colors=C_RED)
ax2.tick_params(axis='y', colors=C_GREEN)
ax.grid(axis='y', alpha=0.3, zorder=0)
ax.set_ylim(0, 80); ax2.set_ylim(0, 100)

# B3: 月度场次 vs 场均GMV vs 场均单量
ax = axesB[0, 2]
sessions_arr = [len(m_totals[m]['sessions']) if m in m_totals else 0 for m in MONTH_ORDER]
avg_gmv_session = [m_totals[m]['gmv']/len(m_totals[m]['sessions'])/10000 if m in m_totals and len(m_totals[m]['sessions'])>0 else 0 for m in MONTH_ORDER]
avg_orders_session = [m_totals[m]['orders']/len(m_totals[m]['sessions']) if m in m_totals and len(m_totals[m]['sessions'])>0 else 0 for m in MONTH_ORDER]

ax3_twin = ax.twinx()
bars_b3 = ax.bar(x, sessions_arr, 0.5, color=C_PURPLE, alpha=0.3, label='直播场次', zorder=2)
ax.plot(x, avg_gmv_session, 'D-', color=C_BLUE, linewidth=2.5, markersize=10, zorder=4, label='场均GMV(万)')
ax3_twin.plot(x, avg_orders_session, '^--', color=C_ORANGE, linewidth=2, markersize=9, zorder=4, label='场均单量')

for i, (s, a, o) in enumerate(zip(sessions_arr, avg_gmv_session, avg_orders_session)):
    if s > 0:
        ax.text(i-0.25, s+0.3, str(s), fontsize=9, color=C_PURPLE, fontweight='bold')
        ax.annotate(f'¥{a:.0f}万', (i, a), textcoords="offset points", xytext=(0,12), ha='center', fontsize=8.5, color=C_BLUE, fontweight='bold')
        ax3_twin.annotate(f'{o:.0f}单', (i, o), textcoords="offset points", xytext=(0,-16), ha='center', fontsize=8.5, color=C_ORANGE, fontweight='bold')

ax.set_xticks(x); ax.set_xticklabels(MONTH_LABELS, fontsize=9)
ax.set_title('月度场次 & 场均效率', fontsize=13, fontweight='bold', pad=10)
ax.set_ylabel('场次', color=C_PURPLE, fontsize=10)
ax3_twin.set_ylabel('场均单量', color=C_ORANGE, fontsize=10)
ax.tick_params(axis='y', colors=C_PURPLE)
ax3_twin.tick_params(axis='y', colors=C_ORANGE)
ax.grid(axis='y', alpha=0.3, zorder=0)

# Handle combined legend
l1, l1a = ax.get_legend_handles_labels()
l2, l2a = ax3_twin.get_legend_handles_labels()
ax.legend(l1+l2, l1a+l2a, loc='upper left', fontsize=7.5)

# B4: 月度平台GMV分布 (堆叠柱状)
ax = axesB[1, 0]
plat_order = ['小红书','三开','抖音+视频号','视频号','抖音','分销','微信小店','公众号','万物','社群']
stack_bottom = np.zeros(len(MONTH_ORDER))
for pidx, p in enumerate(plat_order):
    vals = [plat_monthly[p][m]['gmv']/10000 for m in MONTH_ORDER]
    if sum(vals) > 0:
        ax.bar(x, vals, bottom=stack_bottom, color=PALETTE[pidx % len(PALETTE)], alpha=0.85, label=p, zorder=3)
        stack_bottom += np.array(vals)

ax.set_xticks(x); ax.set_xticklabels(MONTH_LABELS, fontsize=9)
ax.set_title('月度平台 GMV 分布', fontsize=13, fontweight='bold', pad=10)
ax.set_ylabel('GMV (万元)', fontsize=10)
ax.legend(loc='upper left', fontsize=8, ncol=2, framealpha=0.9)
ax.grid(axis='y', alpha=0.3, zorder=0)

# B5: 月度新老达人对比 (首月开播 vs 复播)
ax = axesB[1, 1]
first_appear = {}
for r in data:
    n = r['norm_name']; m = parse_month(r['date'])
    if m and (n not in first_appear or m < first_appear[n]):
        first_appear[n] = m

new_count = defaultdict(int); old_count = defaultdict(int)
new_gmv = defaultdict(float); old_gmv = defaultdict(float)
for r in data:
    n = r['norm_name']; m = parse_month(r['date'])
    if not m: continue
    if first_appear.get(n) == m:
        new_count[m] += 1; new_gmv[m] += r['gmv']/10000
    else:
        old_count[m] += 1; old_gmv[m] += r['gmv']/10000

new_arr = [new_gmv[m] for m in MONTH_ORDER]
old_arr = [old_gmv[m] for m in MONTH_ORDER]

ax.bar(x, old_arr, color=C_BLUE, alpha=0.85, label='复播达人', zorder=3)
ax.bar(x, new_arr, bottom=old_arr, color=C_GOLD, alpha=0.85, label='新达人', zorder=3)

for i in range(len(MONTH_ORDER)):
    if new_arr[i]+old_arr[i] > 0:
        ax.text(i, new_arr[i]+old_arr[i]+3, f'新{new_count[MONTH_ORDER[i]]}\n复{old_count[MONTH_ORDER[i]]}', 
                ha='center', fontsize=7.5, color=C_DARK)

ax.set_xticks(x); ax.set_xticklabels(MONTH_LABELS, fontsize=9)
ax.set_title('月度新老达人 GMV 贡献\n(按首月开播分类)', fontsize=13, fontweight='bold', pad=10)
ax.set_ylabel('GMV (万元)', fontsize=10)
ax.legend(fontsize=8)
ax.grid(axis='y', alpha=0.3, zorder=0)

# B6: 月度GMV增长率 & 环比变化
ax = axesB[1, 2]  # B6 (last subplot in 2x3 grid)
m_gmv_arr = [m_totals[m]['gmv']/10000 if m in m_totals else 0 for m in MONTH_ORDER]
mom_growth = []
for i, g in enumerate(m_gmv_arr):
    if i == 0: mom_growth.append(0)
    else: mom_growth.append((g-m_gmv_arr[i-1])/m_gmv_arr[i-1]*100 if m_gmv_arr[i-1]>0 else 0)

growth_colors = [C_GREEN if v>=0 else C_RED for v in mom_growth]
ax.bar(x, mom_growth, color=growth_colors, alpha=0.7, zorder=3)
ax.plot(x, m_gmv_arr, 'D-', color=C_DARK, linewidth=2, markersize=10, markerfacecolor=C_GOLD, zorder=4)

for i, (g, gm) in enumerate(zip(mom_growth, m_gmv_arr)):
    if i > 0:
        label = f'{g:+.0f}%'
    else: label = '—'
    ax.text(i, mom_growth[i]+(8 if mom_growth[i]>=0 else -12), label, ha='center', fontsize=8, fontweight='bold', color=C_DARK)
    ax.text(i, gm+3, f'{gm:.0f}万', ha='center', fontsize=8, color=C_BLUE, fontweight='bold')

ax.axhline(y=0, color=C_DARK, linewidth=0.5, alpha=0.5)
ax.set_xticks(x); ax.set_xticklabels(MONTH_LABELS, fontsize=9)
ax.set_title('月度 GMV 环比增长率', fontsize=13, fontweight='bold', pad=10)
ax.set_ylabel('环比增长 (%)', fontsize=10)
ax.grid(axis='y', alpha=0.3, zorder=0)

figB.suptitle('维度二：月度维度深度分析', fontsize=17, fontweight='bold', y=1.01, color=C_DARK)
plt.tight_layout(pad=3.5)
outB = '/root/.openclaw/workspace/output/daren_dim_month.png'
figB.savefig(outB, dpi=180, bbox_inches='tight', facecolor=figB.get_facecolor())
plt.close()
print(f"✅ Figure B: {outB} ({os.path.getsize(outB)/1024:.0f} KB)")

# ============================================================
# FIGURE C: 平台维度深度分析 (2 rows × 3 cols)
# ============================================================
figC, axesC = plt.subplots(2, 3, figsize=(22, 14))
figC.patch.set_facecolor('#FAFBFC')

# Filter platforms with meaningful data
meaningful_plats = [p for p in plat_order if sum(plat_monthly[p][m]['gmv'] for m in MONTH_ORDER) > 50000]

# C1: 平台 GMV 占比饼图
ax = axesC[0, 0]
plat_pie_gmv = [sum(plat_monthly[p][m]['gmv'] for m in MONTH_ORDER)/10000 for p in meaningful_plats]
wedges, texts, autotexts = ax.pie(plat_pie_gmv, labels=meaningful_plats, autopct='%1.1f%%',
    colors=PALETTE[:len(meaningful_plats)], startangle=90, pctdistance=0.82,
    textprops={'fontsize':8.5})
for at in autotexts: at.set_fontweight('bold'); at.set_fontsize(8)
ax.set_title('平台 GMV 占比', fontsize=13, fontweight='bold', pad=10)

# C2: 平台退款率对比 (柱状+标注)
ax = axesC[1, 0]
plat_ref = []
for p in meaningful_plats:
    total_o = sum(plat_monthly[p][m]['orders'] for m in MONTH_ORDER)
    total_r = sum(plat_monthly[p][m]['ref_ords'] for m in MONTH_ORDER)
    plat_ref.append({'name':p, 'orders':total_o, 'ref_rate':total_r/total_o*100 if total_o>0 else 0,
                     'gmv':sum(plat_monthly[p][m]['gmv'] for m in MONTH_ORDER)/10000})

plat_ref.sort(key=lambda x:x['ref_rate'])
names_r = [x['name'] for x in plat_ref]
rates_r = [x['ref_rate'] for x in plat_ref]
gmv_r = [x['gmv'] for x in plat_ref]
rate_colors = [C_GREEN if r<25 else C_GOLD if r<40 else C_RED for r in rates_r]
bars_c2 = ax.barh(range(len(names_r)), rates_r, color=rate_colors, height=0.6, zorder=3)
for bar, r, g in zip(bars_c2, rates_r, gmv_r):
    ax.text(bar.get_width()+1, bar.get_y()+bar.get_height()/2, 
            f'{r:.1f}% | GMV ¥{g:.0f}万', va='center', fontsize=8.5, color=C_DARK)
ax.axvline(x=40, color=C_RED, linestyle='--', alpha=0.3, linewidth=1)
ax.set_yticks(range(len(names_r))); ax.set_yticklabels(names_r, fontsize=9)
ax.set_title('平台退款率对比\n(按退款率升序)', fontsize=13, fontweight='bold', pad=10)
ax.set_xlabel('退款率 (%)', fontsize=10)
ax.invert_yaxis()
ax.grid(axis='x', alpha=0.3, zorder=0)

# C3: 平台达人来源多样性
ax = axesC[0, 1]
# How many unique influencers per platform
plat_inf = defaultdict(set)
for r in data:
    p = merge_platform(r['platform'])
    plat_inf[p].add(r['norm_name'])

plat_div = [(p, len(inf_set), sum(plat_monthly[p][m]['gmv'] for m in MONTH_ORDER)/10000) 
            for p, inf_set in plat_inf.items() if p in meaningful_plats]
plat_div.sort(key=lambda x:x[1], reverse=True)

names_d = [x[0] for x in plat_div]
divers = [x[1] for x in plat_div]
gmv_d = [x[2] for x in plat_div]

ax_twin = ax.twinx()
bars_d = ax.bar(range(len(names_d)), divers, color=C_PURPLE, alpha=0.7, label='合作达人数', zorder=3)
ax_twin.plot(range(len(names_d)), gmv_d, 'D-', color=C_BLUE, linewidth=2.5, markersize=10, zorder=4, label='GMV(万)')

for i, (d, g) in enumerate(zip(divers, gmv_d)):
    ax.text(i, d+0.3, str(d), ha='center', fontsize=10, fontweight='bold', color=C_PURPLE)
    ax_twin.annotate(f'¥{g:.0f}万', (i, g), textcoords="offset points", xytext=(0,12), ha='center', fontsize=8.5, color=C_BLUE, fontweight='bold')

ax.set_xticks(range(len(names_d))); ax.set_xticklabels(names_d, fontsize=8.5, rotation=20)
ax.set_title('平台达人多样性\n(达人数 vs GMV)', fontsize=13, fontweight='bold', pad=10)
ax.set_ylabel('合作达人数', color=C_PURPLE, fontsize=10)
ax_twin.set_ylabel('GMV (万元)', color=C_BLUE, fontsize=10)
ax.tick_params(axis='y', colors=C_PURPLE)
ax_twin.tick_params(axis='y', colors=C_BLUE)
ax.grid(axis='y', alpha=0.3)
l1,l1a = ax.get_legend_handles_labels(); l2,l2a = ax_twin.get_legend_handles_labels()
ax.legend(l1+l2, l1a+l2a, loc='upper right', fontsize=7.5)

# C4: 头部平台月度趋势
ax = axesC[0, 2]
top_4_plats = [x[0] for x in plat_div[:4]]
for idx, p in enumerate(top_4_plats):
    vals = [plat_monthly[p][m]['gmv']/10000 for m in MONTH_ORDER]
    ax.plot(range(len(MONTH_ORDER)), vals, 'o-', color=PALETTE[idx], linewidth=2.5, markersize=9, 
            markerfacecolor='white', markeredgewidth=2, label=p)

ax.set_xticks(range(len(MONTH_ORDER))); ax.set_xticklabels(MONTH_LABELS, fontsize=9)
ax.set_title('TOP4 平台月度 GMV 趋势', fontsize=13, fontweight='bold', pad=10)
ax.set_ylabel('GMV (万元)', fontsize=10)
ax.legend(fontsize=9, framealpha=0.9)
ax.grid(alpha=0.3)

# C5: 平台退款率月度热力图
ax = axesC[1, 1]
heat_plat = np.zeros((len(meaningful_plats), len(MONTH_ORDER)))
for i, p in enumerate(meaningful_plats):
    for j, m in enumerate(MONTH_ORDER):
        s = plat_monthly[p][m]
        heat_plat[i, j] = (s['ref_ords']/s['orders']*100) if s['orders']>0 else np.nan

masked_p = np.ma.masked_invalid(heat_plat)
im = ax.imshow(masked_p, cmap='RdYlGn_r', aspect='auto', vmin=0, vmax=80)
for i in range(len(meaningful_plats)):
    for j in range(len(MONTH_ORDER)):
        v = heat_plat[i, j]
        if not np.isnan(v):
            col = 'white' if v > 45 else C_DARK
            ax.text(j, i, f'{v:.0f}%', ha='center', va='center', fontsize=8.5, fontweight='bold', color=col)

ax.set_xticks(range(len(MONTH_ORDER))); ax.set_xticklabels(MONTH_LABELS, fontsize=9)
ax.set_yticks(range(len(meaningful_plats))); ax.set_yticklabels(meaningful_plats, fontsize=9)
ax.set_title('平台退款率月份热力图', fontsize=13, fontweight='bold', pad=10)
plt.colorbar(im, ax=ax, shrink=0.8, pad=0.02).set_label('退款率 (%)', fontsize=9)

# C6: 平台场均效率 & 稳定性
ax = axesC[1, 2]
# Per-platform per-session avg GMV and stddev
plat_session_stats = defaultdict(list)
for r in data:
    p = merge_platform(r['platform'])
    plat_session_stats[p].append(r['gmv']/10000)

plat_eff = []
for p in meaningful_plats:
    vals = plat_session_stats[p]
    if vals:
        plat_eff.append({'name': p, 'avg': np.mean(vals), 'std': np.std(vals), 'count': len(vals)})

plat_eff.sort(key=lambda x:x['avg'], reverse=True)
eff_names = [x['name'] for x in plat_eff]
eff_avg = [x['avg'] for x in plat_eff]
eff_std = [x['std'] for x in plat_eff]

y_pos = np.arange(len(eff_names))
bars_c6 = ax.barh(y_pos, eff_avg, xerr=eff_std, color=C_BLUE, alpha=0.75, height=0.6, 
                   ecolor=C_DARK, capsize=4, zorder=3)

for bar, avg, std, cnt in zip(bars_c6, eff_avg, eff_std, [x['count'] for x in plat_eff]):
    cv = std/avg*100 if avg>0 else 0
    ax.text(bar.get_width()+0.3, bar.get_y()+bar.get_height()/2,
            f'场均 ¥{avg:.1f}万 ±{std:.1f}万 (CV:{cv:.0f}%)', 
            va='center', fontsize=8, color=C_DARK)

ax.set_yticks(y_pos); ax.set_yticklabels(eff_names, fontsize=9)
ax.set_title('平台场均 GMV & 稳定性\n(误差线=标准差)', fontsize=13, fontweight='bold', pad=10)
ax.set_xlabel('场均 GMV (万元)', fontsize=10)
ax.invert_yaxis()
ax.grid(axis='x', alpha=0.3, zorder=0)

figC.suptitle('维度三：平台/渠道维度深度分析', fontsize=17, fontweight='bold', y=1.01, color=C_DARK)
plt.tight_layout(pad=3.5)
outC = '/root/.openclaw/workspace/output/daren_dim_platform.png'
figC.savefig(outC, dpi=180, bbox_inches='tight', facecolor=figC.get_facecolor())
plt.close()
print(f"✅ Figure C: {outC} ({os.path.getsize(outC)/1024:.0f} KB)")

print("\n🎉 All 3 dimensional analysis charts generated!")