fix: improve analytics data quality and human-friendliness

- Add KPI stat cards to Overview tab (active users, sessions, pageviews, bounce rate)
- Filter technical paths from Pages and Paths tabs (/sw.js, /robots.txt, /.git/, /.env, etc.)
- Cap time_on_page at 30min to exclude outlier tabs left open
- Format time as human-readable (Xm Ys) instead of raw seconds
- Mask security tokens in unused pages list (/reset-password/*** etc.)
- Fix Polish labels (period display: "7 dni" instead of "week")
- Add percentages to logged/anonymous donut chart legend

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

blueprints/admin/routes_user_insights.py

@@ -682,6 +682,47 @@ def _tab_engagement(db, start_date, days):
 # TAB 3: PAGE MAP
 # ============================================================
 
+EXCLUDED_PATHS = (
+    '/sw.js', '/robots.txt', '/sitemap.xml', '/favicon.ico',
+    '/apple-touch-icon', '/.well-known/', '/.git/', '/.env',
+    '/check-verification-status', '/manifest.json',
+    '/sito/', '/wordpress/', '/wp-', '/xmlrpc',
+)
+
+
+def _is_technical_path(path):
+    """Check if path is a technical/system endpoint (not a user page)."""
+    if not path:
+        return True
+    for excl in EXCLUDED_PATHS:
+        if path.startswith(excl) or path == excl:
+            return True
+    return False
+
+
+def _mask_token_path(path):
+    """Mask security tokens in paths like /reset-password/<token>."""
+    import re
+    path = re.sub(r'(/reset-password/)[A-Za-z0-9_-]{10,}', r'\1***', path)
+    path = re.sub(r'(/verify-email/)[A-Za-z0-9_-]{10,}', r'\1***', path)
+    return path
+
+
+def _format_time(seconds):
+    """Format seconds to human-readable string."""
+    if seconds <= 0:
+        return '0s'
+    if seconds < 60:
+        return f'{seconds}s'
+    minutes = seconds // 60
+    secs = seconds % 60
+    if minutes < 60:
+        return f'{minutes}m {secs}s' if secs else f'{minutes}m'
+    hours = minutes // 60
+    mins = minutes % 60
+    return f'{hours}h {mins}m'
+
+
 def _tab_pages(db, start_date, days):
     """Page popularity map."""
     start_dt = datetime.combine(start_date, datetime.min.time())
@@ -713,21 +754,25 @@ def _tab_pages(db, start_date, days):
         if prefixes == ['/']:
             conditions = [PageView.path == '/']
 
+        # Cap time_on_page at 1800s (30 min) to exclude outliers (tabs left open)
+        capped_time = func.least(PageView.time_on_page_seconds, 1800)
         q = db.query(
             func.count(PageView.id).label('views'),
             func.count(func.distinct(PageView.user_id)).label('unique_users'),
-            func.avg(PageView.time_on_page_seconds).label('avg_time')
+            func.avg(capped_time).label('avg_time')
         ).join(UserSession, PageView.session_id == UserSession.id).filter(
             or_(*conditions),
             PageView.viewed_at >= start_dt,
             UserSession.is_bot == False
         ).first()
 
+        avg_time_val = int(q.avg_time or 0)
         sections.append({
             'name': name,
             'views': q.views or 0,
             'unique_users': q.unique_users or 0,
-            'avg_time': int(q.avg_time or 0),
+            'avg_time': avg_time_val,
+            'avg_time_fmt': _format_time(avg_time_val),
         })
 
     max_views = max((s['views'] for s in sections), default=1) or 1
@@ -735,36 +780,41 @@ def _tab_pages(db, start_date, days):
     for s in sections:
         s['intensity'] = min(100, int(s['views'] / max_views * 100))
 
-    # Top 50 pages (exclude bots)
-    top_pages = db.query(
+    # Top 50 pages (exclude bots + technical paths)
+    capped_time = func.least(PageView.time_on_page_seconds, 1800)
+    top_pages_raw = db.query(
         PageView.path,
         func.count(PageView.id).label('views'),
         func.count(func.distinct(PageView.user_id)).label('unique_users'),
-        func.avg(PageView.time_on_page_seconds).label('avg_time'),
+        func.avg(capped_time).label('avg_time'),
         func.avg(PageView.scroll_depth_percent).label('avg_scroll'),
         func.avg(PageView.load_time_ms).label('avg_load'),
     ).join(UserSession, PageView.session_id == UserSession.id).filter(
         PageView.viewed_at >= start_dt,
         UserSession.is_bot == False
-    ).group_by(PageView.path).order_by(desc('views')).limit(50).all()
+    ).group_by(PageView.path).order_by(desc('views')).limit(80).all()
 
+    # Filter out technical paths after query (more flexible than SQL LIKE)
+    top_pages = [p for p in top_pages_raw if not _is_technical_path(p.path)][:50]
     max_page_views = top_pages[0].views if top_pages else 1
 
     pages_list = []
     for p in top_pages:
+        avg_time_val = int(p.avg_time or 0)
         pages_list.append({
             'path': p.path,
             'views': p.views,
             'unique_users': p.unique_users,
-            'avg_time': int(p.avg_time or 0),
+            'avg_time': avg_time_val,
+            'avg_time_fmt': _format_time(avg_time_val),
             'avg_scroll': int(p.avg_scroll or 0),
             'avg_load': int(p.avg_load or 0),
             'bar_pct': int(p.views / max_page_views * 100),
         })
 
-    # Ignored pages (< 5 views in 30d, exclude bots)
+    # Ignored pages (< 5 views in 30d, exclude bots + technical paths)
     start_30d = datetime.combine(date.today() - timedelta(days=30), datetime.min.time())
-    ignored = db.query(
+    ignored_raw = db.query(
         PageView.path,
         func.count(PageView.id).label('views'),
     ).join(UserSession, PageView.session_id == UserSession.id).filter(
@@ -772,7 +822,8 @@ def _tab_pages(db, start_date, days):
         UserSession.is_bot == False
     ).group_by(PageView.path).having(
         func.count(PageView.id) < 5
-    ).order_by('views').limit(30).all()
+    ).order_by('views').limit(60).all()
+    ignored = [p for p in ignored_raw if not _is_technical_path(p.path)][:30]
 
     # Top searches
     search_query = db.query(
@@ -795,7 +846,7 @@ def _tab_pages(db, start_date, days):
     return {
         'sections': sections,
         'top_pages': pages_list,
-        'ignored_pages': [{'path': p.path, 'views': p.views} for p in ignored],
+        'ignored_pages': [{'path': _mask_token_path(p.path), 'views': p.views} for p in ignored],
         'top_searches': search_query,
         'searches_no_results': no_results_query,
     }
@@ -820,7 +871,7 @@ def _tab_paths(db, start_date, days):
         )
         SELECT path, COUNT(*) as cnt
         FROM first_pages
-        GROUP BY path ORDER BY cnt DESC LIMIT 10
+        GROUP BY path ORDER BY cnt DESC LIMIT 25
     """)
     entry_pages = db.execute(entry_sql, {'start_dt': start_dt}).fetchall()
 
@@ -835,7 +886,7 @@ def _tab_paths(db, start_date, days):
         )
         SELECT path, COUNT(*) as cnt
         FROM last_pages
-        GROUP BY path ORDER BY cnt DESC LIMIT 10
+        GROUP BY path ORDER BY cnt DESC LIMIT 25
     """)
     exit_pages = db.execute(exit_sql, {'start_dt': start_dt}).fetchall()
 
@@ -854,7 +905,7 @@ def _tab_paths(db, start_date, days):
         SELECT path, next_path, COUNT(*) as cnt
         FROM ordered
         WHERE next_path IS NOT NULL AND path != next_path
-        GROUP BY path, next_path ORDER BY cnt DESC LIMIT 30
+        GROUP BY path, next_path ORDER BY cnt DESC LIMIT 60
     """)
     transitions = db.execute(transitions_sql, {'start_dt': start_dt}).fetchall()
 
@@ -911,11 +962,20 @@ def _tab_paths(db, start_date, days):
     session_lengths = db.execute(session_length_sql, {'start_dt': start_dt}).fetchall()
     max_sl = max((r.cnt for r in session_lengths), default=1) or 1
 
+    # Filter out technical paths from results
+    entry_filtered = [r for r in entry_pages if not _is_technical_path(r.path)][:10]
+    exit_filtered = [r for r in exit_pages if not _is_technical_path(r.path)][:10]
+    transitions_filtered = [r for r in transitions if not _is_technical_path(r.path) and not _is_technical_path(r.next_path)][:30]
+    dropoff_filtered = [r for r in dropoff if not _is_technical_path(r.path)][:20]
+
+    max_entry_f = entry_filtered[0].cnt if entry_filtered else 1
+    max_exit_f = exit_filtered[0].cnt if exit_filtered else 1
+
     return {
-        'entry_pages': [{'path': r.path, 'count': r.cnt, 'bar_pct': int(r.cnt / max_entry * 100)} for r in entry_pages],
-        'exit_pages': [{'path': r.path, 'count': r.cnt, 'bar_pct': int(r.cnt / max_exit * 100)} for r in exit_pages],
-        'transitions': [{'from': r.path, 'to': r.next_path, 'count': r.cnt} for r in transitions],
-        'dropoff': [{'path': r.path, 'views': r.views, 'exits': r.exits, 'exit_rate': float(r.exit_rate)} for r in dropoff],
+        'entry_pages': [{'path': r.path, 'count': r.cnt, 'bar_pct': int(r.cnt / max_entry_f * 100)} for r in entry_filtered],
+        'exit_pages': [{'path': r.path, 'count': r.cnt, 'bar_pct': int(r.cnt / max_exit_f * 100)} for r in exit_filtered],
+        'transitions': [{'from': r.path, 'to': r.next_path, 'count': r.cnt} for r in transitions_filtered],
+        'dropoff': [{'path': r.path, 'views': r.views, 'exits': r.exits, 'exit_rate': float(r.exit_rate)} for r in dropoff_filtered],
         'session_lengths': [{'bucket': r.bucket, 'count': r.cnt, 'bar_pct': int(r.cnt / max_sl * 100)} for r in session_lengths],
     }
 
@@ -930,6 +990,38 @@ def _tab_overview(db, start_date, days):
     start_dt = datetime.combine(start_date, datetime.min.time())
     start_30d = datetime.combine(date.today() - timedelta(days=30), datetime.min.time())
 
+    # KPI stat cards (period-based, bot-filtered)
+    active_users = db.query(func.count(func.distinct(UserSession.user_id))).filter(
+        UserSession.started_at >= start_dt,
+        UserSession.user_id.isnot(None),
+        UserSession.is_bot == False
+    ).scalar() or 0
+
+    total_sessions = db.query(func.count(UserSession.id)).filter(
+        UserSession.started_at >= start_dt,
+        UserSession.is_bot == False
+    ).scalar() or 0
+
+    total_pageviews = db.query(func.count(PageView.id)).join(
+        UserSession, PageView.session_id == UserSession.id
+    ).filter(
+        PageView.viewed_at >= start_dt,
+        UserSession.is_bot == False
+    ).scalar() or 0
+
+    # Bounce rate: sessions with only 1 page view / total sessions
+    single_pv_sessions = db.query(func.count()).select_from(
+        db.query(PageView.session_id).join(
+            UserSession, PageView.session_id == UserSession.id
+        ).filter(
+            PageView.viewed_at >= start_dt,
+            UserSession.is_bot == False
+        ).group_by(PageView.session_id).having(
+            func.count(PageView.id) == 1
+        ).subquery()
+    ).scalar() or 0
+    bounce_rate = round(single_pv_sessions / total_sessions * 100) if total_sessions > 0 else 0
+
     # Daily sessions from analytics_daily (already bot-filtered after migration)
     daily_data = db.query(AnalyticsDaily).filter(
         AnalyticsDaily.date >= date.today() - timedelta(days=30)
@@ -1033,6 +1125,12 @@ def _tab_overview(db, start_date, days):
 
     return {
         'filter_type': filter_type,
+        'kpi': {
+            'active_users': active_users,
+            'sessions': total_sessions,
+            'pageviews': total_pageviews,
+            'bounce_rate': bounce_rate,
+        },
         'chart_data': {
             'labels': chart_labels,
             'sessions': chart_sessions,

templates/admin/user_insights.html

@@ -436,7 +436,7 @@
     <div class="stats-grid">
         <div class="stat-card success">
             <div class="stat-value">{{ data.active_7d }}</div>
-            <div class="stat-label">Aktywni ({{ period }})</div>
+            <div class="stat-label">Aktywni ({% if period == 'day' %}dziś{% elif period == 'week' %}7 dni{% else %}30 dni{% endif %})</div>
         </div>
         <div class="stat-card warning">
             <div class="stat-value">{{ data.at_risk }}</div>
@@ -534,7 +534,7 @@
             <div class="section-tile" style="background: rgba(34, 197, 94, {{ s.intensity / 100 * 0.3 + 0.05 }});">
                 <h3>{{ s.name }}</h3>
                 <div class="stat-value" style="font-size: var(--font-size-lg);">{{ s.views }}</div>
-                <div class="metric">{{ s.unique_users }} unikalnych &middot; {{ s.avg_time }}s śr.</div>
+                <div class="metric">{{ s.unique_users }} unikalnych &middot; {{ s.avg_time_fmt }} śr.</div>
             </div>
             {% endfor %}
         </div>
@@ -568,7 +568,7 @@
                             </div>
                         </td>
                         <td>{{ p.unique_users }}</td>
-                        <td>{{ p.avg_time }}s</td>
+                        <td>{{ p.avg_time_fmt }}</td>
                         <td>{{ p.avg_scroll }}%</td>
                         <td>
                             <span {% if p.avg_load > 3000 %}style="color: var(--error); font-weight: 600;"{% elif p.avg_load > 1500 %}style="color: #d97706;"{% endif %}>
@@ -770,6 +770,26 @@
     <!-- ============================================================ -->
     {% elif tab == 'overview' %}
 
+    <!-- KPI Stat Cards -->
+    <div class="stats-grid">
+        <div class="stat-card success">
+            <div class="stat-value">{{ data.kpi.active_users }}</div>
+            <div class="stat-label">Aktywni użytkownicy</div>
+        </div>
+        <div class="stat-card info">
+            <div class="stat-value">{{ data.kpi.sessions }}</div>
+            <div class="stat-label">Sesje</div>
+        </div>
+        <div class="stat-card info">
+            <div class="stat-value">{{ data.kpi.pageviews }}</div>
+            <div class="stat-label">Odsłony</div>
+        </div>
+        <div class="stat-card {% if data.kpi.bounce_rate > 70 %}error{% elif data.kpi.bounce_rate > 50 %}warning{% else %}success{% endif %}">
+            <div class="stat-value">{{ data.kpi.bounce_rate }}%</div>
+            <div class="stat-label">Współczynnik odrzuceń</div>
+        </div>
+    </div>
+
     <!-- Filter -->
     <div style="margin-bottom: var(--spacing-lg);">
         <div class="filter-group">
@@ -1026,10 +1046,13 @@
     // Auth doughnut
     const authData = {{ data.logged_vs_anon|tojson|safe }};
     const authCtx = document.getElementById('authChart').getContext('2d');
+    const authTotal = authData.logged + authData.anonymous;
+    const authPctLogged = authTotal > 0 ? Math.round(authData.logged / authTotal * 100) : 0;
+    const authPctAnon = authTotal > 0 ? Math.round(authData.anonymous / authTotal * 100) : 0;
     new Chart(authCtx, {
         type: 'doughnut',
         data: {
-            labels: ['Zalogowani', 'Anonimowi'],
+            labels: ['Zalogowani (' + authPctLogged + '%)', 'Anonimowi (' + authPctAnon + '%)'],
             datasets: [{
                 data: [authData.logged, authData.anonymous],
                 backgroundColor: ['#6366f1', '#d1d5db'],