app.R

# ============================================================================
# Quality Growth Strategy Dashboard
# グローバルマーケティング戦略及びソートリーダーシップの高度化
# データドリブンな経営ナラティブによる「Quality Growth」の実現
# ============================================================================

library(shiny)
library(shinydashboard)
library(ggplot2)
library(dplyr)
library(tidyr)
library(plotly)
library(DT)
library(scales)

# ============================================================================
# データ生成（架空データ）
# ============================================================================

set.seed(2026)

# 1. 海外ブランディング時系列データ（24ヶ月）
create_branding_data <- function() {
  n_months <- 24
  dates <- seq(as.Date("2024-01-01"), by = "month", length.out = n_months)
  
  # トレンド + 季節性 + ノイズ
  trend <- seq(200, 1000, length.out = n_months)
  seasonal <- 100 * sin(seq(0, 4*pi, length.out = n_months))
  noise <- rnorm(n_months, 0, 50)
  
  spend <- trend + seasonal + noise
  search <- 0.8 * lag(spend, 1, default = 200) + 1.2 * lag(spend, 2, default = 200) + rnorm(n_months, 0, 30)
  revenue <- 2000 + 3.5 * lag(spend, 2, default = 300) + 2.8 * lag(search, 1, default = 400) + rnorm(n_months, 0, 200)
  
  data.frame(
    Date = dates,
    Spend = round(spend, 0),
    Search = round(pmax(search, 100), 0),
    Revenue = round(pmax(revenue, 2000), 0)
  )
}

# 2. 地域別ROIデータ
create_regional_data <- function() {
  regions <- c("US", "Europe", "Asia", "LATAM", "MEA")
  n_per_region <- 12
  
  regional_data <- do.call(rbind, lapply(regions, function(region) {
    base_spend <- switch(region,
      "US" = 800, "Europe" = 600, "Asia" = 400, "LATAM" = 200, "MEA" = 150
    )
    roi_mult <- switch(region,
      "US" = 3.99, "Europe" = 3.53, "Asia" = 6.48, "LATAM" = 4.2, "MEA" = 5.1
    )
    
    spend <- base_spend + rnorm(n_per_region, 0, base_spend * 0.2)
    search <- spend * runif(n_per_region, 1.2, 1.8) + rnorm(n_per_region, 0, 50)
    revenue <- spend * roi_mult + rnorm(n_per_region, 0, spend * 0.3)
    
    data.frame(
      Region = region,
      Month = seq(as.Date("2025-01-01"), by = "month", length.out = n_per_region),
      Spend = round(spend, 0),
      Search = round(search, 0),
      Revenue = round(revenue, 0),
      Tech_Score = round(runif(n_per_region, 55, 90), 1),
      Design_Score = round(runif(n_per_region, 55, 90), 1),
      Trust_Score = round(runif(n_per_region, 55, 90), 1)
    )
  }))
  
  regional_data
}

# 3. ソートリーダーシップデータ
create_tl_data <- function() {
  n_sessions <- 600
  channels <- c("Organic", "Paid", "Social", "Direct", "Referral")
  reports <- c(
    "AI活用事例2025", "サステナビリティ報告", 
    "DX推進白書", "次世代インフラ展望", "グローバル市場分析"
  )
  
  channel_probs <- c(0.32, 0.19, 0.20, 0.19, 0.10)
  
  tl_data <- data.frame(
    session_id = paste0("S", sprintf("%04d", 1:n_sessions)),
    channel = sample(channels, n_sessions, replace = TRUE, prob = channel_probs),
    report = sample(reports, n_sessions, replace = TRUE),
    duration_sec = NA,
    scrolled = NA,
    pages_viewed = NA
  )
  
  # チャネル別の行動パターン
  for (i in 1:nrow(tl_data)) {
    ch <- tl_data$channel[i]
    if (ch == "Referral") {
      tl_data$duration_sec[i] <- round(rnorm(1, 130, 40))
      tl_data$scrolled[i] <- runif(1) < 0.85
      tl_data$pages_viewed[i] <- rpois(1, 3) + 1
    } else if (ch == "Organic") {
      tl_data$duration_sec[i] <- round(rnorm(1, 105, 35))
      tl_data$scrolled[i] <- runif(1) < 0.75
      tl_data$pages_viewed[i] <- rpois(1, 2.5) + 1
    } else if (ch == "Direct") {
      tl_data$duration_sec[i] <- round(rnorm(1, 70, 30))
      tl_data$scrolled[i] <- runif(1) < 0.60
      tl_data$pages_viewed[i] <- rpois(1, 2) + 1
    } else if (ch == "Social") {
      tl_data$duration_sec[i] <- round(rnorm(1, 65, 35))
      tl_data$scrolled[i] <- runif(1) < 0.50
      tl_data$pages_viewed[i] <- rpois(1, 1.5) + 1
    } else {  # Paid
      tl_data$duration_sec[i] <- round(rnorm(1, 35, 20))
      tl_data$scrolled[i] <- runif(1) < 0.30
      tl_data$pages_viewed[i] <- rpois(1, 1) + 1
    }
  }
  
  tl_data$duration_sec <- pmax(tl_data$duration_sec, 5)
  
  # TL成功判定: 滞在60秒以上 + スクロール + 2ページ以上閲覧
  tl_data$is_tl_success <- with(tl_data, 
    duration_sec >= 60 & scrolled == TRUE & pages_viewed >= 2
  )
  
  tl_data
}

# データ生成
branding_data <- create_branding_data()
regional_data <- create_regional_data()
tl_data <- create_tl_data()

# ============================================================================
# UI定義
# ============================================================================

ui <- dashboardPage(
  skin = "blue",
  
  dashboardHeader(
    title = tags$span(
      tags$img(src = "", height = "30px", style = "margin-right: 10px;"),
      "Quality Growth Dashboard"
    ),
    titleWidth = 350
  ),
  
  dashboardSidebar(
    width = 280,
    sidebarMenu(
      id = "sidebar",
      menuItem("Executive Summary", tabName = "summary", icon = icon("chart-line")),
      menuItem("海外ブランディング分析", tabName = "branding", icon = icon("globe")),
      menuItem("地域別ROI分析", tabName = "regional", icon = icon("map")),
      menuItem("ソートリーダーシップ", tabName = "thought_leadership", icon = icon("lightbulb")),
      menuItem("統計モデル", tabName = "models", icon = icon("calculator")),
      
      hr(),
      
      tags$div(
        style = "padding: 15px; color: #b8c7ce; font-size: 12px;",
        tags$p(tags$strong("プロジェクト情報"))
      )
    )
  ),
  
  dashboardBody(
    tags$head(
      tags$style(HTML("
        .content-wrapper { background-color: #f4f6f9; }
        .box { border-radius: 8px; box-shadow: 0 2px 10px rgba(0,0,0,0.1); }
        .box-header { border-bottom: 2px solid #3c8dbc; }
        .info-box { border-radius: 8px; }
        .small-box { border-radius: 8px; }
        .nav-tabs-custom > .nav-tabs > li.active { border-top-color: #3c8dbc; }
        .skin-blue .main-header .logo { background-color: #1a3a5c; }
        .skin-blue .main-header .navbar { background-color: #1a3a5c; }
        .metric-card { 
          background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
          border-radius: 12px;
          padding: 20px;
          color: white;
          margin-bottom: 15px;
        }
        .metric-card h3 { margin: 0; font-size: 2em; }
        .metric-card p { margin: 5px 0 0; opacity: 0.9; }
      "))
    ),
    
    tabItems(
      # Executive Summary タブ
      tabItem(
        tabName = "summary",
        fluidRow(
          column(12,
            tags$div(
              style = "background: linear-gradient(135deg, #1a3a5c 0%, #2d5a87 100%); 
                       padding: 30px; border-radius: 12px; color: white; margin-bottom: 20px;",
              tags$h2("グローバルマーケティング戦略及びソートリーダーシップの高度化", 
                      style = "margin-top: 0;"),
              tags$p("データドリブンな経営ナラティブによる「Quality Growth」の実現", 
                     style = "font-size: 1.2em; opacity: 0.9;")
            )
          )
        ),
        
        fluidRow(
          valueBoxOutput("total_investment", width = 3),
          valueBoxOutput("avg_roi", width = 3),
          valueBoxOutput("tl_success_rate", width = 3),
          valueBoxOutput("top_region", width = 3)
        ),
        
        fluidRow(
          box(
            title = "Strategic Context: Quality Growthへの戦略的整合",
            status = "primary", solidHeader = TRUE, width = 6,
            tags$div(
              style = "display: flex; gap: 20px;",
              tags$div(
                style = "flex: 1; padding: 15px; background: #f8f9fa; border-radius: 8px;",
                tags$h4("Current Context (現状)", style = "color: #6c757d;"),
                tags$ul(
                  tags$li("Volume Expansion（量的拡大）"),
                  tags$li("Technology-focused Context（技術文脈）")
                ),
                tags$p("課題：経営層への認知・顧客化に最適化されていない", 
                       style = "color: #dc3545; font-size: 0.9em;")
              ),
              tags$div(
                style = "flex: 1; padding: 15px; background: #e8f4f8; border-radius: 8px;",
                tags$h4("Strategic Pivot (新方針)", style = "color: #007bff;"),
                tags$ul(
                  tags$li("Quality Growth（質的成長）"),
                  tags$li("Management-focused Narrative（経営ナラティブ）"),
                  tags$li("Key Themes: Powered & Next Gen Infrastructure")
                )
              )
            )
          ),
          
          box(
            title = "Project Scope: 2つの並行ワークストリーム",
            status = "primary", solidHeader = TRUE, width = 6,
            tags$div(
              style = "display: flex; gap: 20px;",
              tags$div(
                style = "flex: 1; padding: 15px; background: #fff3cd; border-radius: 8px;",
                tags$h4("Region A: 思想リーダーシップ", style = "color: #856404;"),
                tags$p("Focus: 基準の再定義と経営ナラティブ"),
                tags$ul(
                  tags$li("現状分析（回遊ログ・アクセス解析）"),
                  tags$li("競合ベンチマーク調査"),
                  tags$li("「質」を測る新KPIの定義")
                )
              ),
              tags$div(
                style = "flex: 1; padding: 15px; background: #d4edda; border-radius: 8px;",
                tags$h4("Region B: グローバルマーケティング", style = "color: #155724;"),
                tags$p("Focus: 標準化と投資対効果（ROI）"),
                tags$ul(
                  tags$li("拠点間横比較（市場規模 vs 成長率）"),
                  tags$li("ROI分析と投資効率の可視化"),
                  tags$li("レポートフォーマットの標準化")
                )
              )
            )
          )
        ),
        
        fluidRow(
          box(
            title = "Schedule & Roadmap",
            status = "info", solidHeader = TRUE, width = 12,
            plotlyOutput("timeline_chart", height = "250px")
          )
        )
      ),
      
      # 海外ブランディング分析 タブ
      tabItem(
        tabName = "branding",
        fluidRow(
          box(
            title = "海外ブランディング指標の推移（24ヶ月）",
            status = "primary", solidHeader = TRUE, width = 12,
            plotlyOutput("branding_timeseries", height = "400px")
          )
        ),
        
        fluidRow(
          box(
            title = "相互相関分析: Spend vs Revenue",
            status = "info", solidHeader = TRUE, width = 6,
            plotOutput("ccf_plot", height = "300px"),
            tags$p("タイムラグの特定：予算投下から効果が最大化するまでの期間を可視化", 
                   style = "font-size: 0.9em; color: #6c757d;")
          ),
          
          box(
            title = "回帰分析結果",
            status = "info", solidHeader = TRUE, width = 6,
            verbatimTextOutput("regression_summary"),
            tags$div(
              style = "margin-top: 15px; padding: 10px; background: #e8f4f8; border-radius: 8px;",
              tags$h5("解釈ポイント:"),
              tags$ul(
                tags$li("lag(Spend, 2)の係数: 2ヶ月前の投資が現在の売上に寄与"),
                tags$li("lag(Search, 1)の係数: 指名検索数が売上をドライブ"),
                tags$li("R-squared: モデルの説明力")
              )
            )
          )
        ),
        
        fluidRow(
          box(
            title = "投資効率サマリー",
            status = "success", solidHeader = TRUE, width = 12,
            fluidRow(
              column(4, 
                tags$div(class = "metric-card",
                  tags$h3(textOutput("total_spend_text")),
                  tags$p("総投資額")
                )
              ),
              column(4,
                tags$div(class = "metric-card", style = "background: linear-gradient(135deg, #11998e 0%, #38ef7d 100%);",
                  tags$h3(textOutput("total_revenue_text")),
                  tags$p("総売上額")
                )
              ),
              column(4,
                tags$div(class = "metric-card", style = "background: linear-gradient(135deg, #ee0979 0%, #ff6a00 100%);",
                  tags$h3(textOutput("simple_roi_text")),
                  tags$p("単純ROI")
                )
              )
            )
          )
        )
      ),
      
      # 地域別ROI分析 タブ
      tabItem(
        tabName = "regional",
        fluidRow(
          box(
            title = "地域別ROI比較",
            status = "primary", solidHeader = TRUE, width = 6,
            plotlyOutput("regional_roi_chart", height = "350px")
          ),
          
          box(
            title = "Brand Search vs Revenue",
            status = "primary", solidHeader = TRUE, width = 6,
            plotlyOutput("search_revenue_scatter", height = "350px")
          )
        ),
        
        fluidRow(
          box(
            title = "定性データ: ブランド認知スコア（0-100）",
            status = "info", solidHeader = TRUE, width = 12,
            plotlyOutput("qualitative_heatmap", height = "300px"),
            tags$div(
              style = "margin-top: 15px; padding: 10px; background: #fff3cd; border-radius: 8px;",
              tags$h5("地域別インサイト:"),
              tags$ul(
                tags$li(tags$strong("Asia:"), " 技術力（Technology）が高評価 → 最新技術をアピール"),
                tags$li(tags$strong("Europe:"), " 信頼度（Trust）が重要 → 認証取得・伝統の訴求"),
                tags$li(tags$strong("US:"), " デザイン（Design）が先行指標 → クリエイティブ重視")
              )
            )
          )
        ),
        
        fluidRow(
          box(
            title = "地域別詳細データ",
            status = "success", solidHeader = TRUE, width = 12,
            DTOutput("regional_table")
          )
        )
      ),
      
      # ソートリーダーシップ タブ
      tabItem(
        tabName = "thought_leadership",
        fluidRow(
          box(
            title = "チャネル別TL成功率",
            status = "primary", solidHeader = TRUE, width = 6,
            plotlyOutput("tl_channel_chart", height = "350px"),
            tags$p("TL成功条件: 滞在60秒以上 + スクロール発生 + 2ページ以上閲覧", 
                   style = "font-size: 0.9em; color: #6c757d;")
          ),
          
          box(
            title = "コンテンツ別パフォーマンス",
            status = "primary", solidHeader = TRUE, width = 6,
            plotlyOutput("content_performance", height = "350px")
          )
        ),
        
        fluidRow(
          box(
            title = "チャネル別統計サマリー",
            status = "info", solidHeader = TRUE, width = 12,
            DTOutput("tl_summary_table")
          )
        ),
        
        fluidRow(
          box(
            title = "流入チャネルの「質」の評価",
            status = "warning", solidHeader = TRUE, width = 6,
            tags$div(
              style = "padding: 15px;",
              tags$h5("Referral / Organicの成功率が高い場合:"),
              tags$p("外部メディアでの紹介や検索意図が、レポートの内容と合致しており、信頼獲得に成功している。"),
              tags$hr(),
              tags$h5("Paid / Socialの成功率が低い場合:"),
              tags$p("「タイトル釣り」などで流入は稼げているが、中身が期待に沿っていない、あるいは「ながら読み」層が多い可能性がある。")
            )
          ),
          
          box(
            title = "次のアクション提案",
            status = "success", solidHeader = TRUE, width = 6,
            tags$div(
              style = "padding: 15px;",
              tags$ul(
                tags$li("スター・コンテンツ（High PV, High Success）の横展開"),
                tags$li("ニッチ・リーダーシップコンテンツの広報強化"),
                tags$li("Low Successコンテンツのトピック・ターゲット見直し"),
                tags$li("GA4実データへの移行とスコアリングカスタマイズ")
              )
            )
          )
        )
      ),
      
      # 統計モデル タブ
      tabItem(
        tabName = "models",
        fluidRow(
          box(
            title = "Analytical Methodology: 統計的アプローチの採用",
            status = "primary", solidHeader = TRUE, width = 12,
            tags$div(
              style = "display: flex; gap: 20px; flex-wrap: wrap;",
              tags$div(
                style = "flex: 1; min-width: 250px; padding: 20px; background: linear-gradient(135deg, #667eea 0%, #764ba2 100%); border-radius: 12px; color: white;",
                tags$h4("Regression Analysis（回帰分析）"),
                tags$p("成功に寄与している変数を統計的に切り分け、投資効率を算出する。")
              ),
              tags$div(
                style = "flex: 1; min-width: 250px; padding: 20px; background: linear-gradient(135deg, #11998e 0%, #38ef7d 100%); border-radius: 12px; color: white;",
                tags$h4("Weighted Average Models（加重平均モデル）"),
                tags$p("各国の市場環境の違いを考慮に入れた、公平な評価モデルの設計。")
              ),
              tags$div(
                style = "flex: 1; min-width: 250px; padding: 20px; background: linear-gradient(135deg, #ee0979 0%, #ff6a00 100%); border-radius: 12px; color: white;",
                tags$h4("Statistical Tools（分析ツール）"),
                tags$p("専門的な統計解析ソフトウェアを活用し、ブラックボックス化しない「説明可能なロジック」を構築。")
              )
            )
          )
        ),
        
        fluidRow(
          box(
            title = "地域別回帰分析結果",
            status = "info", solidHeader = TRUE, width = 12,
            selectInput("region_select", "地域を選択:", 
                        choices = c("US", "Europe", "Asia", "LATAM", "MEA"),
                        selected = "US"),
            verbatimTextOutput("regional_regression")
          )
        ),
        
        fluidRow(
          box(
            title = "モデル診断プロット",
            status = "warning", solidHeader = TRUE, width = 6,
            plotOutput("residual_plot", height = "300px")
          ),
          
          box(
            title = "予測 vs 実績",
            status = "warning", solidHeader = TRUE, width = 6,
            plotlyOutput("predicted_vs_actual", height = "300px")
          )
        )
      )
    )
  )
)

# ============================================================================
# Server定義
# ============================================================================

server <- function(input, output, session) {
  
  # --- Value Boxes ---
  output$total_investment <- renderValueBox({
    valueBox(
      "¥10,500,000",
      "Total Investment (税別)",
      icon = icon("yen-sign"),
      color = "blue"
    )
  })
  
  output$avg_roi <- renderValueBox({
    roi_summary <- regional_data %>%
      group_by(Region) %>%
      summarise(ROI = sum(Revenue) / sum(Spend)) %>%
      summarise(Avg_ROI = mean(ROI))
    
    valueBox(
      sprintf("%.2fx", roi_summary$Avg_ROI),
      "平均ROI（全地域）",
      icon = icon("chart-line"),
      color = "green"
    )
  })
  
  output$tl_success_rate <- renderValueBox({
    success_rate <- mean(tl_data$is_tl_success) * 100
    valueBox(
      sprintf("%.1f%%", success_rate),
      "TL成功率",
      icon = icon("lightbulb"),
      color = "yellow"
    )
  })
  
  output$top_region <- renderValueBox({
    top <- regional_data %>%
      group_by(Region) %>%
      summarise(ROI = sum(Revenue) / sum(Spend)) %>%
      arrange(desc(ROI)) %>%
      slice(1)
    
    valueBox(
      top$Region,
      sprintf("Top ROI Region (%.2fx)", top$ROI),
      icon = icon("trophy"),
      color = "purple"
    )
  })
  
  # --- Timeline Chart ---
  output$timeline_chart <- renderPlotly({
    phases <- data.frame(
      Phase = c("Phase 1: Mobilization", "Phase 2: Deep Analysis", "Phase 3: Reporting", "Phase 4: Close"),
      Start = as.Date(c("2026-01-19", "2026-02-02", "2026-02-23", "2026-03-02")),
      End = as.Date(c("2026-02-01", "2026-02-22", "2026-03-01", "2026-03-15")),
      Color = c("#3498db", "#e74c3c", "#2ecc71", "#9b59b6")
    )
    
    plot_ly() %>%
      add_segments(
        data = phases,
        x = ~Start, xend = ~End,
        y = ~Phase, yend = ~Phase,
        color = ~Phase,
        colors = phases$Color,
        line = list(width = 20),
        showlegend = FALSE
      ) %>%
      layout(
        xaxis = list(title = "", tickformat = "%b %d"),
        yaxis = list(title = "", categoryorder = "array", 
                     categoryarray = rev(phases$Phase)),
        margin = list(l = 150)
      )
  })
  
  # --- Branding Time Series ---
  output$branding_timeseries <- renderPlotly({
    df_long <- branding_data %>%
      pivot_longer(cols = c(Spend, Search, Revenue), 
                   names_to = "Metric", values_to = "Value")
    
    plot_ly(df_long, x = ~Date, y = ~Value, color = ~Metric, 
            type = 'scatter', mode = 'lines+markers',
            colors = c("Revenue" = "#2ecc71", "Search" = "#e74c3c", "Spend" = "#3498db")) %>%
      layout(
        xaxis = list(title = ""),
        yaxis = list(title = "Value (Standardized Units)"),
        legend = list(orientation = "h", y = -0.1),
        hovermode = "x unified"
      )
  })
  
  # --- CCF Plot ---
  output$ccf_plot <- renderPlot({
    ccf_result <- ccf(branding_data$Spend, branding_data$Revenue, 
                      lag.max = 10, plot = FALSE)
    
    df_ccf <- data.frame(
      Lag = ccf_result$lag,
      ACF = ccf_result$acf
    )
    
    ggplot(df_ccf, aes(x = Lag, y = ACF)) +
      geom_hline(yintercept = 0, color = "gray50") +
      geom_hline(yintercept = c(-0.4, 0.4), linetype = "dashed", color = "blue", alpha = 0.5) +
      geom_segment(aes(xend = Lag, yend = 0), color = "#3498db", size = 1) +
      geom_point(color = "#3498db", size = 3) +
      labs(title = "Cross-Correlation: Spend vs Revenue",
           x = "Lag (months)", y = "ACF") +
      theme_minimal(base_size = 14) +
      theme(plot.title = element_text(hjust = 0.5, face = "bold"))
  })
  
  # --- Regression Summary ---
  output$regression_summary <- renderPrint({
    df <- branding_data
    model <- lm(Revenue ~ lag(Spend, 2) + lag(Search, 1), data = df)
    summary(model)
  })
  
  # --- Metric Cards ---
  output$total_spend_text <- renderText({
    paste0("¥", format(sum(branding_data$Spend) * 1000, big.mark = ","))
  })
  
  output$total_revenue_text <- renderText({
    paste0("¥", format(sum(branding_data$Revenue) * 1000, big.mark = ","))
  })
  
  output$simple_roi_text <- renderText({
    roi <- sum(branding_data$Revenue) / sum(branding_data$Spend)
    sprintf("%.2fx", roi)
  })
  
  # --- Regional ROI Chart ---
  output$regional_roi_chart <- renderPlotly({
    roi_data <- regional_data %>%
      group_by(Region) %>%
      summarise(ROI = sum(Revenue) / sum(Spend)) %>%
      arrange(desc(ROI))
    
    plot_ly(roi_data, x = ~reorder(Region, ROI), y = ~ROI, type = 'bar',
            marker = list(color = c("#2ecc71", "#3498db", "#e74c3c", "#9b59b6", "#f39c12"))) %>%
      layout(
        xaxis = list(title = "Region"),
        yaxis = list(title = "ROI (Revenue / Spend)"),
        title = list(text = "Revenue per 1 Unit of Spend")
      )
  })
  
  # --- Search vs Revenue Scatter ---
  output$search_revenue_scatter <- renderPlotly({
    agg_data <- regional_data %>%
      group_by(Region) %>%
      summarise(
        Avg_Search = mean(Search),
        Avg_Revenue = mean(Revenue)
      )
    
    plot_ly(agg_data, x = ~Avg_Search, y = ~Avg_Revenue, color = ~Region,
            type = 'scatter', mode = 'markers',
            marker = list(size = 15),
            text = ~Region) %>%
      layout(
        xaxis = list(title = "Brand Search Volume (Avg)"),
        yaxis = list(title = "Revenue (Avg)")
      )
  })
  
  # --- Qualitative Heatmap ---
  output$qualitative_heatmap <- renderPlotly({
    qual_data <- regional_data %>%
      group_by(Region) %>%
      summarise(
        Technology = mean(Tech_Score),
        Design = mean(Design_Score),
        Trust = mean(Trust_Score)
      ) %>%
      pivot_longer(cols = c(Technology, Design, Trust),
                   names_to = "Attribute", values_to = "Score")
    
    plot_ly(qual_data, x = ~Attribute, y = ~Region, z = ~Score,
            type = 'heatmap',
            colorscale = list(c(0, "#f7fbff"), c(0.5, "#6baed6"), c(1, "#08306b")),
            text = ~round(Score, 1),
            texttemplate = "%{text}",
            showscale = TRUE) %>%
      layout(
        xaxis = list(title = ""),
        yaxis = list(title = "")
      )
  })
  
  # --- Regional Table ---
  output$regional_table <- renderDT({
    summary_data <- regional_data %>%
      group_by(Region) %>%
      summarise(
        `Total Spend` = sum(Spend),
        `Total Revenue` = sum(Revenue),
        `ROI` = round(sum(Revenue) / sum(Spend), 2),
        `Avg Tech Score` = round(mean(Tech_Score), 1),
        `Avg Design Score` = round(mean(Design_Score), 1),
        `Avg Trust Score` = round(mean(Trust_Score), 1)
      )
    
    datatable(summary_data, 
              options = list(pageLength = 10, dom = 't'),
              rownames = FALSE) %>%
      formatCurrency(c('Total Spend', 'Total Revenue'), currency = "¥", digits = 0)
  })
  
  # --- TL Channel Chart ---
  output$tl_channel_chart <- renderPlotly({
    channel_summary <- tl_data %>%
      group_by(channel) %>%
      summarise(
        sessions = n(),
        success_count = sum(is_tl_success),
        success_rate = mean(is_tl_success)
      ) %>%
      arrange(desc(success_rate))
    
    plot_ly(channel_summary, y = ~reorder(channel, success_rate), x = ~success_rate,
            type = 'bar', orientation = 'h',
            marker = list(color = c("#00bcd4", "#8bc34a", "#ff9800", "#e91e63", "#9c27b0"))) %>%
      layout(
        xaxis = list(title = "Success Rate", tickformat = ".0%"),
        yaxis = list(title = "")
      )
  })
  
  # --- Content Performance ---
  output$content_performance <- renderPlotly({
    content_summary <- tl_data %>%
      group_by(report) %>%
      summarise(
        sessions = n(),
        success_rate = mean(is_tl_success),
        avg_duration = mean(duration_sec)
      )
    
    plot_ly(content_summary, x = ~sessions, y = ~success_rate, 
            size = ~avg_duration, color = ~report,
            type = 'scatter', mode = 'markers',
            marker = list(sizemode = 'diameter', sizeref = 2),
            text = ~paste(report, "<br>Sessions:", sessions, 
                          "<br>Success Rate:", scales::percent(success_rate),
                          "<br>Avg Duration:", round(avg_duration), "sec")) %>%
      layout(
        xaxis = list(title = "Sessions (PV)"),
        yaxis = list(title = "TL Success Rate", tickformat = ".0%")
      )
  })
  
  # --- TL Summary Table ---
  output$tl_summary_table <- renderDT({
    tl_data %>%
      group_by(channel) %>%
      summarise(
        `Sessions` = n(),
        `Success Count` = sum(is_tl_success),
        `Success Rate` = scales::percent(mean(is_tl_success), accuracy = 0.1),
        `Avg Duration (sec)` = round(mean(duration_sec), 1),
        `Avg Pages Viewed` = round(mean(pages_viewed), 2)
      ) %>%
      arrange(desc(`Success Count`)) %>%
      datatable(options = list(pageLength = 10, dom = 't'), rownames = FALSE)
  })
  
  # --- Regional Regression ---
  output$regional_regression <- renderPrint({
    region_df <- regional_data %>% filter(Region == input$region_select)
    model <- lm(Revenue ~ Tech_Score + Design_Score + Trust_Score, data = region_df)
    summary(model)
  })
  
  # --- Residual Plot ---
  output$residual_plot <- renderPlot({
    df <- branding_data %>% filter(!is.na(lag(Spend, 2)) & !is.na(lag(Search, 1)))
    model <- lm(Revenue ~ lag(Spend, 2) + lag(Search, 1), data = branding_data)
    
    residuals_df <- data.frame(
      Fitted = fitted(model),
      Residuals = residuals(model)
    )
    
    ggplot(residuals_df, aes(x = Fitted, y = Residuals)) +
      geom_point(color = "#3498db", size = 3, alpha = 0.7) +
      geom_hline(yintercept = 0, linetype = "dashed", color = "red") +
      geom_smooth(method = "loess", se = FALSE, color = "#e74c3c") +
      labs(title = "Residuals vs Fitted", x = "Fitted Values", y = "Residuals") +
      theme_minimal(base_size = 14)
  })
  
  # --- Predicted vs Actual ---
  output$predicted_vs_actual <- renderPlotly({
    model <- lm(Revenue ~ lag(Spend, 2) + lag(Search, 1), data = branding_data)
    
    pred_df <- data.frame(
      Actual = branding_data$Revenue[3:24],
      Predicted = predict(model, branding_data)[3:24]
    )
    
    plot_ly(pred_df) %>%
      add_trace(x = ~Actual, y = ~Predicted, type = 'scatter', mode = 'markers',
                marker = list(color = "#3498db", size = 10),
                name = "Data Points") %>%
      add_trace(x = c(min(pred_df$Actual), max(pred_df$Actual)),
                y = c(min(pred_df$Actual), max(pred_df$Actual)),
                type = 'scatter', mode = 'lines',
                line = list(color = "red", dash = "dash"),
                name = "Perfect Fit") %>%
      layout(
        xaxis = list(title = "Actual Revenue"),
        yaxis = list(title = "Predicted Revenue")
      )
  })
}

# ============================================================================
# アプリ起動
# ============================================================================

shinyApp(ui = ui, server = server)