diff --git a/apps/ops/src/modules/campaigns/pages/CampaignNewPage.vue b/apps/ops/src/modules/campaigns/pages/CampaignNewPage.vue
index 5c75d12..71710e0 100644
--- a/apps/ops/src/modules/campaigns/pages/CampaignNewPage.vue
+++ b/apps/ops/src/modules/campaigns/pages/CampaignNewPage.vue
@@ -373,6 +373,41 @@
           </q-list>
         </q-expansion-item>
 
+        <!-- ⚠ Lignes du Map CSV qui n'ont pas passé le parsing ─────────── -->
+        <q-expansion-item v-if="parseSkippedRows.length" expand-separator
+                          icon="filter_alt_off"
+                          :label="`${parseSkippedRows.length} ligne(s) du Map CSV non importée(s)`"
+                          header-class="bg-red-1 text-red-9"
+                          class="q-mt-sm">
+          <div class="q-pa-sm text-caption text-grey-7">
+            Chaque ligne ci-dessous a été <strong>droppée au parsing</strong> et n'apparaît pas
+            dans la liste d'envoi. Le numéro de ligne (#) correspond à la position dans le fichier
+            Excel original (header inclus).
+          </div>
+          <q-table
+            :rows="parseSkippedRows" :columns="skippedRowsColumns" row-key="source_row"
+            flat dense hide-bottom :pagination="{ rowsPerPage: 0 }"
+          >
+            <template v-slot:body-cell-source_row="props">
+              <q-td :props="props" class="text-grey-6">#{{ props.row.source_row }}</q-td>
+            </template>
+            <template v-slot:body-cell-reason="props">
+              <q-td :props="props">
+                <q-chip dense size="sm" :color="reasonColor(props.row.reason)" text-color="white"
+                        :label="reasonLabel(props.row.reason)" />
+              </q-td>
+            </template>
+            <template v-slot:body-cell-actions="props">
+              <q-td :props="props" class="text-right">
+                <q-btn v-if="props.row.reason !== 'no_email'" flat dense size="sm" color="primary"
+                       icon="person_add" @click="prefillManualFromSkipped(props.row)">
+                  <q-tooltip>Ouvrir la dialog "Ajouter manuellement" avec ces champs pré-remplis</q-tooltip>
+                </q-btn>
+              </q-td>
+            </template>
+          </q-table>
+        </q-expansion-item>
+
         <!-- Action row: preview + edit template are quick-access utilities,
              both non-destructive. The primary action is "Continuer" which
              moves to Step 3 (still NOT the send — Step 3 has its own
@@ -648,6 +683,9 @@ const unusedGifts = ref([])
 // exactly which rows didn't make it into the pairing (no email, duplicates,
 // multi-skip).
 const parseSkipped = ref(null)
+// The actual rows dropped (up to 200) — each with the raw source columns
+// + reason — for the inspectable expansion-item below the imbalance banner.
+const parseSkippedRows = ref([])
 
 // row_index (#1, #2, ...) is the source-CSV position — invaluable for the
 // user to cross-reference what they see here against the file they uploaded.
@@ -677,6 +715,52 @@ const unpairedColumns = [
   { name: 'postal_code', label: 'Code postal', field: 'postal_code', align: 'left' },
 ]
 
+// Columns for the "dropped at parsing" expansion table. Reads from the raw
+// CSV columns (raw_full_name, raw_email, etc.) since the row never made it
+// through the contact-building pipeline.
+const skippedRowsColumns = [
+  { name: 'source_row', label: 'Ligne #', field: 'source_row', align: 'left' },
+  { name: 'reason',     label: 'Raison',  field: 'reason',     align: 'left' },
+  { name: 'raw_full_name', label: 'Nom au CSV', field: 'raw_full_name', align: 'left' },
+  { name: 'raw_email',     label: 'Email au CSV', field: 'raw_email',  align: 'left' },
+  { name: 'raw_address',   label: 'Adresse',      field: 'raw_address', align: 'left' },
+  { name: 'raw_postal',    label: 'Code postal',  field: 'raw_postal',  align: 'left' },
+  { name: 'actions',       label: '',             field: '',            align: 'right' },
+]
+
+function reasonColor (reason) {
+  return {
+    no_email:   'red-7',
+    duplicate:  'orange-7',
+    multi_skip: 'amber-8',
+  }[reason] || 'grey-7'
+}
+function reasonLabel (reason) {
+  return {
+    no_email:   'Sans email valide',
+    duplicate:  'Email en double',
+    multi_skip: 'Couple ignoré (réglage multi)',
+  }[reason] || reason
+}
+
+// Open the manual-add dialog pre-filled with the dropped row's data so the
+// operator can recover it in one click. The duplicate/multi_skip cases are
+// the ones worth recovering (the email IS there — we just intentionally
+// skipped it); no_email cases can't be recovered without an email address.
+function prefillManualFromSkipped (row) {
+  const [first, ...rest] = (row.raw_full_name || '').trim().split(/\s+/)
+  manualRow.value = {
+    ...emptyManualRow(),
+    firstname:     first || '',
+    lastname:      rest.join(' ') || '',
+    email:         (row.raw_email || '').split(/[;,\s]+/).filter(e => e.includes('@'))[0] || '',
+    phone:         row.raw_phone || '',
+    civic_address: row.raw_address || '',
+    postal_code:   row.raw_postal || '',
+  }
+  manualOpen.value = true
+}
+
 const matchedCount = computed(() => recipients.value.filter(r => r.customer_id).length)
 const unmatchedCount = computed(() => recipients.value.filter(r => !r.customer_id).length)
 const excludedCount = computed(() => recipients.value.filter(r => r.excluded).length)
@@ -848,6 +932,7 @@ async function goPreview () {
     unpairedContacts.value = r.unpaired_contacts || []
     unusedGifts.value      = r.unused_gifts || []
     parseSkipped.value     = r.skipped || null
+    parseSkippedRows.value = r.skipped_rows || []
     step.value = 2
   } catch (e) {
     $q.notify({ type: 'negative', message: 'Erreur de parsing: ' + e.message })
diff --git a/services/targo-hub/lib/campaigns.js b/services/targo-hub/lib/campaigns.js
index 46d5622..80efdb7 100644
--- a/services/targo-hub/lib/campaigns.js
+++ b/services/targo-hub/lib/campaigns.js
@@ -391,20 +391,34 @@ function parseMapCsv (text, multi = 'first') {
   const contacts = []
   const seen = new Set()
   let skippedNoEmail = 0, skippedNoName = 0, skippedDuplicate = 0, skippedMultiSkip = 0
+  // Capture the actual rows that were dropped so the wizard can show
+  // operators exactly WHO got lost, not just a count. Each entry includes
+  // the reason and just enough source-CSV columns to identify the human.
+  const skippedRows = []
+  const snapshot = (row, reason, idx) => ({
+    source_row:   idx + 2,   // +1 for 0-index, +1 for header line — easier to cross-reference in Excel
+    reason,
+    raw_full_name: (row['nom au compte'] || row["nom à l'adresse"] || '').trim(),
+    raw_email:     (row['email au compte'] || row["email à l'adresse"] || '').trim(),
+    raw_phone:     (row['telephone au compte'] || row["telephone à l'adresse"] || '').trim(),
+    raw_address:   (row['adresse dans F'] || '').trim(),
+    raw_postal:    (row['code postal au compte'] || row["code postal à l'adresse"] || '').trim(),
+  })
 
-  for (const r of rows) {
+  for (let rowIdx = 0; rowIdx < rows.length; rowIdx++) {
+    const r = rows[rowIdx]
     // Pull emails from either source column
     let rawEmails = (r['email au compte'] || r["email à l'adresse"] || '').trim()
-    if (!rawEmails) { skippedNoEmail++; continue }
+    if (!rawEmails) { skippedNoEmail++; skippedRows.push(snapshot(r, 'no_email', rowIdx)); continue }
     const emails = rawEmails.split(EMAIL_SPLIT)
                             .map(e => normalizeEmail(e))
                             .filter(e => e.includes('@') && e.split('@')[1]?.includes('.'))
-    if (!emails.length) { skippedNoEmail++; continue }
+    if (!emails.length) { skippedNoEmail++; skippedRows.push(snapshot(r, 'no_email', rowIdx)); continue }
 
     const sendEmails = multi === 'split' ? emails
                      : multi === 'skip'  ? (emails.length > 1 ? [] : emails)
                      : emails.slice(0, 1)
-    if (!sendEmails.length) { skippedMultiSkip++; continue }
+    if (!sendEmails.length) { skippedMultiSkip++; skippedRows.push(snapshot(r, 'multi_skip', rowIdx)); continue }
 
     // We used to skip rows without a name here. That dropped the contact AND
     // wasted its paired Giftbit shortlink — the worker already defaults
@@ -434,7 +448,13 @@ function parseMapCsv (text, multi = 'first') {
     const phone         = normalizePhone(r['telephone au compte'] || r["telephone à l'adresse"] || '')
 
     for (const em of sendEmails) {
-      if (seen.has(em)) { skippedDuplicate++; continue }
+      if (seen.has(em)) {
+        skippedDuplicate++
+        // Capture the duplicate row with its specific email — useful when one
+        // raw row had multiple emails and only some were dupes (multi='split').
+        skippedRows.push({ ...snapshot(r, 'duplicate', rowIdx), raw_email: em })
+        continue
+      }
       seen.add(em)
       contacts.push({
         firstname, lastname,
@@ -457,6 +477,9 @@ function parseMapCsv (text, multi = 'first') {
       multi_skip: skippedMultiSkip,
       total_rows: rows.length,
     },
+    // Cap at 200 to keep response size reasonable on very dirty exports;
+    // the count in `skipped` always reflects the true total.
+    skipped_rows: skippedRows.slice(0, 200),
   }
 }
 
@@ -1308,7 +1331,7 @@ async function handle (req, res, method, path) {
     if (!map_csv || !giftbit_csv) {
       return json(res, 400, { error: 'map_csv and giftbit_csv required' })
     }
-    const { contacts, skipped } = parseMapCsv(map_csv, multi || 'first')
+    const { contacts, skipped, skipped_rows } = parseMapCsv(map_csv, multi || 'first')
     const gifts = parseGiftbitCsv(giftbit_csv)
 
     // Match contacts ↔ gifts (by row order, fallback to email when Giftbit
@@ -1352,6 +1375,7 @@ async function handle (req, res, method, path) {
       leftover_gifts:    unused_gifts.length,
       leftover_contacts: unpaired_contacts.length,
       skipped,
+      skipped_rows,
     })
   }