Debugging the customer by shift mismatched

streamlit_app.py

@@ -91,6 +91,15 @@ LANG = {
         "sm_col_delivery":     "Delivery",
         "sm_col_partypack":    "Party Pack",
         "sm_chart_rev_split":  "Monthly Revenue by Channel",
+        "sm_recon_title":      "Customer count reconciliation",
+        "sm_recon_help":       "Rows where Summary Customers ≠ sum of round customers. "
+                               "Summary numbers come from the daily POS summary report; round "
+                               "numbers come from shift-level POS line items filtered to "
+                               "Adult + Kid packages. Gaps usually mean delivery / party-pack / "
+                               "comp customers counted in Summary but not assigned to a shift.",
+        "sm_recon_round_sum":  "Sum of Rounds",
+        "sm_recon_diff":       "Difference",
+        "sm_recon_no_gap":     "No discrepancy in this filter window — all rows match.",
         # Forecast tab
         "fc_month_title":     "This Month Forecast",
         "fc_month_customers": "Forecast Customers (this month)",
@@ -217,6 +226,15 @@ LANG = {
         "sm_col_delivery":     "เดลิเวอรี่",
         "sm_col_partypack":    "พาร์ตี้แพ็ค",
         "sm_chart_rev_split":  "รายได้รายเดือนตามช่องทาง",
+        "sm_recon_title":      "การกระทบยอดจำนวนลูกค้า",
+        "sm_recon_help":       "แถวที่ค่าลูกค้าจาก Summary ไม่ตรงกับผลรวมของลูกค้าตามรอบ "
+                               "ตัวเลข Summary มาจากรายงานสรุปประจำวันของ POS ส่วนตัวเลขรอบ "
+                               "มาจากรายการขายรายไอเทมระดับ Shift กรองเฉพาะแพ็คเกจ Adult + Kid "
+                               "ความต่างมักหมายถึงลูกค้าเดลิเวอรี่/พาร์ตี้แพ็ค/อภินันทนาการ "
+                               "ที่ถูกนับใน Summary แต่ไม่ได้ผูกกับรอบใด",
+        "sm_recon_round_sum":  "ผลรวมรอบ",
+        "sm_recon_diff":       "ผลต่าง",
+        "sm_recon_no_gap":     "ไม่พบความต่างในช่วงที่เลือก ทุกแถวตรงกัน",
         # Forecast tab
         "fc_month_title":     "พยากรณ์ของเดือนนี้",
         "fc_month_customers": "พยากรณ์จำนวนลูกค้า (เดือนนี้)",
@@ -1586,9 +1604,12 @@ with tab_summary:
             si_all = _summary_filter(fact_shift_items)
             if "Restaurant" in si_all.columns:
                 si_all = si_all[si_all["Restaurant"] == "Copper Buffet"]
+            # Only rows in the customer-paying tiers count toward the
+            # round customer total: Normal + Premium + Party Pack
+            # (Delivery and off-menu rows are excluded).
             si_cust = (
-                si_all[si_all["Group3"].isin(["Adult", "Kid"])]
-                if "Group3" in si_all.columns else si_all
+                si_all[si_all["SubType"].isin(["Normal", "Premium", "Party Pack"])]
+                if "SubType" in si_all.columns else si_all
             )
             if not si_cust.empty:
                 si_cust = si_cust.copy()
@@ -1878,6 +1899,102 @@ with tab_summary:
         else:
             st.caption(t("sm_no_monthly_rows"))
 
+        # ── Customer count reconciliation (Copper Buffet only) ───────
+        # Surfaces rows where the Summary-based Customers column doesn't
+        # equal the sum of round columns from fact_shift_items, so the
+        # operations team can spot which months / branches drive the gap
+        # and decide whether it's expected (delivery customers etc.) or
+        # actually broken in the source data.
+        if (restaurant_name == "Copper Buffet"
+                and not m.empty
+                and "Customers" in m.columns
+                and round_cols):
+            with st.expander(f"🔍 {t('sm_recon_title')}", expanded=True):
+                st.caption(t("sm_recon_help"))
+                _rec = m.copy()
+                _rec["_RoundSum"] = _rec[round_cols].fillna(0).sum(axis=1)
+
+                # Compute a *raw* Qty total per (Year, Month, Branch)
+                # directly from si_cust before pivoting. If RawQty
+                # matches RoundSum but neither matches Customers, the
+                # gap is between data sources (Summary vs shift POS) or
+                # caused by a Year/Month/Branch merge mismatch. If
+                # RawQty ≠ RoundSum, the pivot itself is dropping data.
+                _raw = pd.DataFrame(columns=["Year", "Month", "Branch", "_RawQty"])
+                try:
+                    _si = _summary_filter(fact_shift_items)
+                    if "Restaurant" in _si.columns:
+                        _si = _si[_si["Restaurant"] == "Copper Buffet"]
+                    if "SubType" in _si.columns:
+                        _si = _si[_si["SubType"].isin(["Normal", "Premium", "Party Pack"])]
+                    if not _si.empty and "Date" in _si.columns:
+                        _si = _si.copy()
+                        _si["Year"]  = pd.to_datetime(_si["Date"], errors="coerce").dt.year
+                        _si["Month"] = pd.to_datetime(_si["Date"], errors="coerce").dt.month
+                        _raw = (
+                            _si.groupby(["Year", "Month", "Branch"], as_index=False)["Qty"]
+                               .sum().rename(columns={"Qty": "_RawQty"})
+                        )
+                except Exception as e:
+                    st.caption(f"(raw-qty debug failed: {e})")
+
+                # Normalize the merge keys to int so a float/int
+                # dtype mismatch can't silently drop rows.
+                if not _raw.empty:
+                    _raw["Year"]  = pd.to_numeric(_raw["Year"], errors="coerce").astype("Int64")
+                    _raw["Month"] = pd.to_numeric(_raw["Month"], errors="coerce").astype("Int64")
+                _rec["Year"]  = pd.to_numeric(_rec["Year"],  errors="coerce").astype("Int64")
+                _rec["Month"] = pd.to_numeric(_rec["Month"], errors="coerce").astype("Int64")
+                _rec = _rec.merge(_raw, on=["Year", "Month", "Branch"], how="left")
+                _rec["_RawQty"] = _rec["_RawQty"].fillna(0)
+                _rec["_Diff"]   = _rec["Customers"].fillna(0) - _rec["_RoundSum"]
+                _rec["_Diff2"]  = _rec["_RawQty"]  - _rec["_RoundSum"]
+
+                _gap = _rec[
+                    (_rec["_Diff"].abs() > 0.5) | (_rec["_Diff2"].abs() > 0.5)
+                ]
+                if _gap.empty:
+                    st.caption(t("sm_recon_no_gap"))
+                else:
+                    _gap = _gap.sort_values(
+                        ["Year", "Month", "Branch"], ascending=[True, True, True]
+                    )
+                    _gcols = ["Year", "Month", "Branch",
+                              "Customers", "_RoundSum", "_RawQty",
+                              "_Diff", "_Diff2"]
+                    _gdisp = _gap[_gcols].copy()
+                    _gdisp["Customers"] = _gdisp["Customers"].map(fmt_num)
+                    _gdisp["_RoundSum"] = _gdisp["_RoundSum"].map(fmt_num)
+                    _gdisp["_RawQty"]   = _gdisp["_RawQty"].map(fmt_num)
+                    _gdisp["_Diff"]     = _gdisp["_Diff"].map(
+                        lambda v: "—" if pd.isna(v) else f"{v:+,.0f}"
+                    )
+                    _gdisp["_Diff2"] = _gdisp["_Diff2"].map(
+                        lambda v: "—" if pd.isna(v) else f"{v:+,.0f}"
+                    )
+                    _gdisp = _gdisp.rename(columns={
+                        "_RoundSum": t("sm_recon_round_sum"),
+                        "_RawQty":   "Raw Qty",
+                        "_Diff":     t("sm_recon_diff"),
+                        "_Diff2":    "Raw − Rounds",
+                    })
+                    st.dataframe(
+                        _gdisp,
+                        use_container_width=True, hide_index=True,
+                        column_config={
+                            "Year":  st.column_config.NumberColumn(format="%d"),
+                            "Month": st.column_config.NumberColumn(format="%d"),
+                        },
+                    )
+                    st.caption(
+                        "Interpretation: if **Raw − Rounds** is non-zero, "
+                        "the pivot is dropping data (likely a Round/Shift "
+                        "value falling outside the expected set). If **"
+                        "Difference** is non-zero but Raw − Rounds is "
+                        "zero, the gap is between data sources or a "
+                        "Branch/Year-Month merge mismatch."
+                    )
+
         # ── Daily detail (collapsed by default — can be long) ────────────
         # Mirror the Monthly summary column layout, just with Date instead
         # of Year / Month. Adds %Cap, %Premium and per-round customer
@@ -1967,9 +2084,11 @@ with tab_summary:
                     si_all = _summary_filter(fact_shift_items)
                     if "Restaurant" in si_all.columns:
                         si_all = si_all[si_all["Restaurant"] == "Copper Buffet"]
+                    # Only Normal + Premium + Party Pack rows count as
+                    # customers (matches the monthly logic above).
                     si_cust = (
-                        si_all[si_all["Group3"].isin(["Adult", "Kid"])]
-                        if "Group3" in si_all.columns else si_all
+                        si_all[si_all["SubType"].isin(["Normal", "Premium", "Party Pack"])]
+                        if "SubType" in si_all.columns else si_all
                     )
 
                     if not si_cust.empty: