Update app.py

app.py

@@ -12,8 +12,8 @@ warnings.filterwarnings('ignore')
 
 # ========== PAGE CONFIGURATION ==========
 st.set_page_config(
-    page_title="Getaround Delay Analysis",
-    page_icon="🚗",
+    page_title="Car Rental Delay Analysis",
+    #page_icon="🚗",
     layout="wide",
     initial_sidebar_state="expanded"
 )
@@ -292,6 +292,111 @@ if selected == "Overview & Problems":
     </div>
     """, unsafe_allow_html=True)
     
+    # Direct answer to PM's questions
+    st.markdown('<div class="section-header">How Often Are Drivers Late & Impact on Next Driver</div>', unsafe_allow_html=True)
+    
+    # Calculate late return frequency and impact
+    df_late_analysis = df[df["has_previous_rental"]].copy()
+    
+    # Join with previous rental delay data
+    prev_rental_data = df[["rental_id", "delay_at_checkout_in_minutes"]].rename(
+        columns={"rental_id": "previous_ended_rental_id", 
+                "delay_at_checkout_in_minutes": "previous_delay"}
+    )
+    df_late_analysis = df_late_analysis.merge(prev_rental_data, on="previous_ended_rental_id", how="left")
+    
+    # Calculate late return frequency
+    df_late_analysis["previous_delay_clean"] = df_late_analysis["previous_delay"].clip(-720, 720)
+    df_late_analysis["previous_was_late"] = (
+        df_late_analysis["previous_delay"].notnull() & 
+        (df_late_analysis["previous_delay_clean"] > 0)
+    )
+    
+    # Calculate impact on next driver
+    df_late_analysis["causes_problem"] = (
+        df_late_analysis["previous_delay"].notnull() & 
+        (df_late_analysis["previous_delay_clean"] > df_late_analysis["time_delta_with_previous_rental_in_minutes"])
+    )
+    df_late_analysis["wait_time"] = np.maximum(
+        0, 
+        df_late_analysis["previous_delay_clean"] - df_late_analysis["time_delta_with_previous_rental_in_minutes"]
+    ).fillna(0)
+    
+    late_returns = df_late_analysis[df_late_analysis["previous_was_late"]]
+    impacted_next_drivers = df_late_analysis[df_late_analysis["causes_problem"]]
+    
+    col1, col2 = st.columns(2)
+    
+    with col1:
+        st.markdown("#### Late Return Frequency")
+        total_with_delay_data = df_late_analysis[df_late_analysis["previous_delay"].notnull()]
+        late_frequency = (len(late_returns) / len(total_with_delay_data)) * 100 if len(total_with_delay_data) > 0 else 0
+        
+        st.metric("Late Returns", f"{len(late_returns):,}")
+        st.metric("Late Return Rate", f"{late_frequency:.1f}%",
+                 help="Percentage of returns that are late (delay > 0 minutes)")
+        
+        # Late return severity
+        if len(late_returns) > 0:
+            avg_late_delay = late_returns["previous_delay_clean"].mean()
+            st.metric("Average Late Delay", f"{avg_late_delay:.1f} min")
+    
+    with col2:
+        st.markdown("#### Impact on Next Driver")
+        impact_rate = (len(impacted_next_drivers) / len(df_late_analysis)) * 100 if len(df_late_analysis) > 0 else 0
+        
+        st.metric("Next Drivers Impacted", f"{len(impacted_next_drivers):,}")
+        st.metric("Impact Rate", f"{impact_rate:.1f}%",
+                 help="Percentage of consecutive rentals where late return causes waiting")
+        
+        if len(impacted_next_drivers) > 0:
+            avg_wait = impacted_next_drivers["wait_time"].mean()
+            st.metric("Average Wait Time", f"{avg_wait:.1f} min",
+                     help="Average additional wait time when impacted")
+    
+    # Visual analysis of the relationship
+    col1, col2 = st.columns(2)
+    
+    with col1:
+        # Late return distribution
+        if len(late_returns) > 0:
+            late_delays = late_returns["previous_delay_clean"]
+            late_filtered = late_delays[late_delays <= 300]  # Cap at 5 hours for visualization
+            
+            fig_late = px.histogram(
+                late_filtered, 
+                nbins=20,
+                title="Distribution of Late Return Delays",
+                labels={"value": "Delay (minutes)", "count": "Number of Late Returns"}
+            )
+            st.plotly_chart(fig_late, use_container_width=True)
+    
+    with col2:
+        # Wait time impact distribution
+        if len(impacted_next_drivers) > 0:
+            wait_times = impacted_next_drivers[impacted_next_drivers["wait_time"] > 0]["wait_time"]
+            
+            fig_wait = px.histogram(
+                wait_times, 
+                nbins=20,
+                title="Wait Time Distribution for Impacted Next Drivers",
+                labels={"value": "Wait Time (minutes)", "count": "Number of Impacted Drivers"}
+            )
+            st.plotly_chart(fig_wait, use_container_width=True)
+    
+    # Key insight summary
+    late_to_impact_ratio = (len(impacted_next_drivers) / len(late_returns)) * 100 if len(late_returns) > 0 else 0
+    
+    st.markdown(f"""
+    <div class="insight-box">
+    <strong>Key Insights:</strong><br>
+    • <strong>{late_frequency:.1f}%</strong> of returns are late (drivers return after scheduled time)<br>
+    • <strong>{impact_rate:.1f}%</strong> of consecutive rentals are negatively impacted by previous late returns<br>
+    • <strong>{late_to_impact_ratio:.1f}%</strong> of late returns actually cause problems for the next driver<br>
+    • When problems occur, next drivers wait an average of <strong>{avg_wait:.1f} minutes</strong>
+    </div>
+    """, unsafe_allow_html=True)
+    
     # Visual analysis
     st.markdown('<div class="section-header">Delay Patterns</div>', unsafe_allow_html=True)