diff --git a/CHANGELOG.md b/CHANGELOG.md
index 2d1baaf5c..856759c8a 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -47,6 +47,12 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/).
 
 ### Changed
 
+- BREAKING: `--input-csv` now requires `Term` and `PaymentOption` columns in
+  addition to `Service`, `Region`, `ResourceType` and `Count`. `Term` must be
+  `1yr` or `3yr` and `PaymentOption` must be `all-upfront`, `partial-upfront`
+  or `no-upfront` on every row. CSVs with only the older four columns, or with
+  a blank or unknown value, now fail to load with an error naming the missing
+  column or the row.
 - Split the CLI out of the CUDly monorepo into its own module,
   `github.com/LeanerCloud/cloud-commitments-cli`. The shared client and
   provider code it depends on now lives in
diff --git a/cmd/cancellation_prepurchase_test.go b/cmd/cancellation_prepurchase_test.go
index 9e5c8b7f1..80694e42c 100644
--- a/cmd/cancellation_prepurchase_test.go
+++ b/cmd/cancellation_prepurchase_test.go
@@ -51,7 +51,7 @@ func TestFilterAndAdjustRecommendations_CancellationAndDeadlineAreTerminal(t *te
 func TestPrepareCSVPurchaseRun_CanceledContextStopsBeforeConfirmationAndPurchase(t *testing.T) {
 	dir := t.TempDir()
 	csvPath := filepath.Join(dir, "in.csv")
-	require.NoError(t, os.WriteFile(csvPath, []byte("Service,Region,ResourceType,Count,Engine\nrds,us-east-1,db.t3.small,2,mysql\n"), 0o600))
+	require.NoError(t, os.WriteFile(csvPath, []byte("Service,Region,ResourceType,Count,Engine,Term,PaymentOption\nrds,us-east-1,db.t3.small,2,mysql,1yr,no-upfront\n"), 0o600))
 	cfg := Config{CSVInput: csvPath, AuditLog: filepath.Join(dir, "audit.jsonl")}
 
 	for _, dry := range []bool{true, false} {
diff --git a/cmd/multi_service_csv.go b/cmd/multi_service_csv.go
index ff4fb0175..ce2025d8a 100644
--- a/cmd/multi_service_csv.go
+++ b/cmd/multi_service_csv.go
@@ -13,6 +13,7 @@ import (
 	"strconv"
 	"strings"
 	"time"
+	"unicode/utf8"
 
 	"github.com/LeanerCloud/cloud-commitments-go/pkg/common"
 	"github.com/LeanerCloud/cloud-commitments-go/providers/aws/recommendations"
@@ -48,9 +49,14 @@ func loadRecommendationsFromCSV(csvPath string) ([]common.Recommendation, error)
 	if err != nil {
 		return nil, fmt.Errorf("failed to read CSV header: %w", err)
 	}
+	if err = validateCSVUTF8(header); err != nil {
+		return nil, fmt.Errorf("CSV header: %w", err)
+	}
 
-	// Build column index map
-	colIdx := buildColumnIndexMap(header)
+	colIdx, err := buildColumnIndexMap(header)
+	if err != nil {
+		return nil, err
+	}
 
 	// Parse all records
 	parsed, err := parseCSVRecords(reader, colIdx)
@@ -61,13 +67,60 @@ func loadRecommendationsFromCSV(csvPath string) ([]common.Recommendation, error)
 	return parsed, nil
 }
 
-// buildColumnIndexMap creates a map from column names to indices.
-func buildColumnIndexMap(header []string) map[string]int {
+// Minimal CSVs require these columns; service-specific columns stay optional.
+// Term and PaymentOption are required because the purchase path reads them per
+// row and must never guess a commitment's term or payment option.
+var requiredCSVColumns = []string{"Service", "Region", "ResourceType", "Count", "Term", "PaymentOption"}
+
+// Accepted Term values match fetchRecommendationsForRegion; accepted
+// PaymentOption values match validatePaymentAndTerm.
+var (
+	validCSVTerms          = []string{"1yr", "3yr"}
+	validCSVPaymentOptions = []string{"all-upfront", "partial-upfront", "no-upfront"}
+)
+
+// parseCSVEnum requires the cell to be exactly one of allowed.
+func parseCSVEnum(record []string, colIdx map[string]int, fieldName string, allowed []string) (string, error) {
+	col := colIdx[fieldName] + 1
+	value := getCSVField(record, colIdx, fieldName)
+	if strings.TrimSpace(value) == "" {
+		return "", fmt.Errorf("column %d %q: value is required (one of: %s)", col, fieldName, strings.Join(allowed, ", "))
+	}
+	for _, a := range allowed {
+		if value == a {
+			return value, nil
+		}
+	}
+	return "", fmt.Errorf("column %d %q: invalid value %q (must be one of: %s)", col, fieldName, value, strings.Join(allowed, ", "))
+}
+
+func validateCSVUTF8(fields []string) error {
+	for i, field := range fields {
+		if !utf8.ValidString(field) {
+			return fmt.Errorf("column %d: invalid UTF-8 encoding", i+1)
+		}
+	}
+	return nil
+}
+
+func buildColumnIndexMap(header []string) (map[string]int, error) {
+	if len(header) > 0 {
+		header[0] = strings.TrimPrefix(header[0], "\ufeff")
+	}
 	colIdx := make(map[string]int)
 	for i, col := range header {
 		colIdx[col] = i
 	}
-	return colIdx
+	var missing []string
+	for _, req := range requiredCSVColumns {
+		if _, ok := colIdx[req]; !ok {
+			missing = append(missing, req)
+		}
+	}
+	if len(missing) > 0 {
+		return nil, fmt.Errorf("CSV header missing required columns: %s", strings.Join(missing, ", "))
+	}
+	return colIdx, nil
 }
 
 // parseCSVRecords reads and parses all CSV records.
@@ -82,6 +135,10 @@ func parseCSVRecords(reader *csv.Reader, colIdx map[string]int) ([]common.Recomm
 		if err != nil {
 			return nil, fmt.Errorf("failed to read CSV record: %w", err)
 		}
+		if err = validateCSVUTF8(record); err != nil {
+			line, _ := reader.FieldPos(0)
+			return nil, fmt.Errorf("CSV line %d: %w", line, err)
+		}
 
 		// Skip the trailing TOTAL summary row that writeMultiServiceCSVReport
 		// emits (label in the Service column). Without this, feeding the tool
@@ -113,14 +170,25 @@ func parseCSVRecord(record []string, colIdx map[string]int) (common.Recommendati
 	rec.ResourceType = getCSVField(record, colIdx, "ResourceType")
 	rec.Account = getCSVField(record, colIdx, "Account")
 	rec.AccountName = getCSVField(record, colIdx, "AccountName")
-	rec.Term = getCSVField(record, colIdx, "Term")
-	rec.PaymentOption = getCSVField(record, colIdx, "PaymentOption")
 
-	if err := parseCSVCount(record, colIdx, &rec.Count); err != nil {
+	term, err := parseCSVEnum(record, colIdx, "Term", validCSVTerms)
+	if err != nil {
+		return rec, err
+	}
+	rec.Term = term
+	payment, err := parseCSVEnum(record, colIdx, "PaymentOption", validCSVPaymentOptions)
+	if err != nil {
+		return rec, err
+	}
+	rec.PaymentOption = payment
+
+	err = parseCSVCount(record, colIdx, &rec.Count)
+	if err != nil {
 		return rec, err
 	}
 
-	if err := parseCSVFloat(record, colIdx, "EstimatedSavings", &rec.EstimatedSavings); err != nil {
+	err = parseCSVFloat(record, colIdx, "EstimatedSavings", &rec.EstimatedSavings)
+	if err != nil {
 		return rec, err
 	}
 
@@ -143,9 +211,7 @@ func getCSVField(record []string, colIdx map[string]int, fieldName string) strin
 	return ""
 }
 
-// parseCSVCount parses the required Count field as a whole non-negative
-// integer. It drives purchase quantities, so a blank, missing, fractional or
-// otherwise malformed cell is an error rather than a truncated or zero value.
+// Savings Plans exports use Count=1, preserving positive-count round trips.
 func parseCSVCount(record []string, colIdx map[string]int, target *int) error {
 	const fieldName = "Count"
 	idx, ok := colIdx[fieldName]
@@ -158,17 +224,16 @@ func parseCSVCount(record []string, colIdx map[string]int, target *int) error {
 	}
 	n, err := strconv.Atoi(value)
 	if err != nil {
-		return fmt.Errorf("column %d %q: invalid integer %q: %w", idx+1, fieldName, value, err)
+		return fmt.Errorf("column %d %q: invalid integer: %w", idx+1, fieldName, err)
 	}
-	if n < 0 {
-		return fmt.Errorf("column %d %q: must not be negative, got %d", idx+1, fieldName, n)
+	if n < 1 {
+		return fmt.Errorf("column %d %q: must be at least 1, got %d", idx+1, fieldName, n)
 	}
 	*target = n
 	return nil
 }
 
-// parseCSVFloat parses a finite float field from a CSV record. A blank or
-// absent cell leaves target untouched (see requireRankingSignal).
+// Blank savings stay absent-as-zero for requireRankingSignal.
 func parseCSVFloat(record []string, colIdx map[string]int, fieldName string, target *float64) error {
 	value := strings.TrimSpace(getCSVField(record, colIdx, fieldName))
 	if value == "" {
@@ -177,11 +242,14 @@ func parseCSVFloat(record []string, colIdx map[string]int, fieldName string, tar
 	col := colIdx[fieldName] + 1
 	f, err := strconv.ParseFloat(value, 64)
 	if err != nil {
-		return fmt.Errorf("column %d %q: invalid number %q: %w", col, fieldName, value, err)
+		return fmt.Errorf("column %d %q: invalid number: %w", col, fieldName, err)
 	}
 	if math.IsNaN(f) || math.IsInf(f, 0) {
 		return fmt.Errorf("column %d %q: invalid number %q: must be finite", col, fieldName, value)
 	}
+	if f < 0 {
+		return fmt.Errorf("column %d %q: must not be negative, got %g", col, fieldName, f)
+	}
 	*target = f
 	return nil
 }
diff --git a/cmd/multi_service_csv_header_test.go b/cmd/multi_service_csv_header_test.go
new file mode 100644
index 000000000..8adfd7fd9
--- /dev/null
+++ b/cmd/multi_service_csv_header_test.go
@@ -0,0 +1,151 @@
+package main
+
+import (
+	"encoding/binary"
+	"testing"
+	"unicode/utf16"
+
+	"github.com/LeanerCloud/cloud-commitments-go/pkg/common"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestLoadRecommendationsFromCSV_HeaderValidation_1327(t *testing.T) {
+	tests := []struct {
+		name        string
+		header      string
+		row         string
+		errContains string
+	}{
+		{"missing Service", "Region,ResourceType,Count,Term,PaymentOption\n", "us-east-1,db.t3.micro,2,3yr,partial-upfront\n", "Service"},
+		{"missing Region", "Service,ResourceType,Count,Term,PaymentOption\n", "rds,db.t3.micro,2,3yr,partial-upfront\n", "Region"},
+		{"missing ResourceType", "Service,Region,Count,Term,PaymentOption\n", "rds,us-east-1,2,3yr,partial-upfront\n", "ResourceType"},
+		{"missing Count", "Service,Region,ResourceType,Term,PaymentOption\n", "rds,us-east-1,db.t3.micro,3yr,partial-upfront\n", "Count"},
+		{"missing Term", "Service,Region,ResourceType,Count,PaymentOption\n", "rds,us-east-1,db.t3.micro,2,partial-upfront\n", "Term"},
+		{"missing PaymentOption", "Service,Region,ResourceType,Count,Term\n", "rds,us-east-1,db.t3.micro,2,3yr\n", "PaymentOption"},
+		{"legacy 4-column header rejected", "Service,Region,ResourceType,Count\n", "rds,us-east-1,db.t3.micro,2\n", "Term, PaymentOption"},
+		{"multiple missing", "Service,Count\n", "rds,2\n", "Region, ResourceType, Term, PaymentOption"},
+		{
+			"legacy TEST-02 headers rejected",
+			"Service,Region,Instance Type,Instance Count\n",
+			"rds,us-east-1,db.t3.micro,2\n",
+			"ResourceType, Count, Term, PaymentOption",
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			err := loadCSVContent(t, tt.header+tt.row)
+			require.Error(t, err)
+			assert.Contains(t, err.Error(), "CSV header missing required columns")
+			assert.Contains(t, err.Error(), tt.errContains)
+		})
+	}
+}
+
+func TestLoadRecommendationsFromCSV_BOMPrefixed_1327(t *testing.T) {
+	recs, err := loadRecommendationsFromCSV(writeTestRecommendationsCSV(t,
+		"\ufeffService,Region,ResourceType,Count,Term,PaymentOption\nrds,us-east-1,db.t3.micro,2,3yr,partial-upfront\n"))
+	require.NoError(t, err)
+	require.Len(t, recs, 1)
+	assert.Equal(t, common.ServiceRDS, recs[0].Service)
+	assert.Equal(t, "us-east-1", recs[0].Region)
+	assert.Equal(t, 2, recs[0].Count)
+}
+
+func TestLoadRecommendationsFromCSV_Encoding_1327(t *testing.T) {
+	t.Run("UTF-16", func(t *testing.T) {
+		units := utf16.Encode([]rune("Service,Region,ResourceType,Count,Term,PaymentOption\nrds,us-east-1,db.t3.micro,2,3yr,partial-upfront\n"))
+		for _, tt := range []struct {
+			name  string
+			bom   []byte
+			order binary.ByteOrder
+		}{
+			{"little endian", []byte{0xff, 0xfe}, binary.LittleEndian},
+			{"big endian", []byte{0xfe, 0xff}, binary.BigEndian},
+		} {
+			t.Run(tt.name, func(t *testing.T) {
+				encoded := make([]byte, 2+2*len(units))
+				copy(encoded, tt.bom)
+				for i, unit := range units {
+					tt.order.PutUint16(encoded[2+2*i:], unit)
+				}
+				err := loadCSVContent(t, string(encoded))
+				require.Error(t, err)
+				assert.EqualError(t, err, "CSV header: column 1: invalid UTF-8 encoding")
+			})
+		}
+	})
+
+	t.Run("Latin-1 optional header", func(t *testing.T) {
+		err := loadCSVContent(t, "Service,Region,ResourceType,Count,Caf\xe9,Term,PaymentOption\n"+
+			"rds,us-east-1,db.t3.micro,2,prod,3yr,partial-upfront\n")
+		require.Error(t, err)
+		assert.EqualError(t, err, "CSV header: column 5: invalid UTF-8 encoding")
+	})
+
+	t.Run("Latin-1 AccountName", func(t *testing.T) {
+		err := loadCSVContent(t, "Service,Region,ResourceType,Count,AccountName,Term,PaymentOption\n"+
+			"rds,us-east-1,db.t3.micro,2,Caf\xe9,3yr,partial-upfront\n")
+		require.Error(t, err)
+		assert.EqualError(t, err, "CSV line 2: column 5: invalid UTF-8 encoding")
+	})
+
+	t.Run("invalid ignored field in TOTAL after multiline record", func(t *testing.T) {
+		err := loadCSVContent(t, "Service,Region,ResourceType,Count,Ignored,Term,PaymentOption\n"+
+			"rds,us-east-1,db.t3.micro,2,\"prod\naccount\",3yr,partial-upfront\n"+
+			"TOTAL,,,2,\xff,,\n")
+		require.Error(t, err)
+		assert.EqualError(t, err, "CSV line 4: column 5: invalid UTF-8 encoding")
+	})
+
+	t.Run("valid non-ASCII UTF-8", func(t *testing.T) {
+		recs, err := loadRecommendationsFromCSV(writeTestRecommendationsCSV(t,
+			"Service,Region,ResourceType,Count,AccountName,Term,PaymentOption\n"+
+				"rds,us-east-1,db.t3.micro,2,Café 東京,3yr,partial-upfront\n"))
+		require.NoError(t, err)
+		require.Len(t, recs, 1)
+		assert.Equal(t, "Café 東京", recs[0].AccountName)
+	})
+}
+
+// Lock in encoding/csv behaviors the purchase path relies on.
+func TestLoadRecommendationsFromCSV_EdgeCases_1327(t *testing.T) {
+	t.Run("CRLF line endings", func(t *testing.T) {
+		recs, err := loadRecommendationsFromCSV(writeTestRecommendationsCSV(t,
+			"Service,Region,ResourceType,Count,Term,PaymentOption\r\nrds,us-east-1,db.t3.micro,2,3yr,partial-upfront\r\n"))
+		require.NoError(t, err)
+		require.Len(t, recs, 1)
+		assert.Equal(t, 2, recs[0].Count)
+	})
+
+	t.Run("quoted commas and newlines inside fields", func(t *testing.T) {
+		recs, err := loadRecommendationsFromCSV(writeTestRecommendationsCSV(t,
+			"Service,Region,ResourceType,Count,AccountName,Term,PaymentOption\n"+
+				"rds,us-east-1,\"db.t3.micro, burstable\",2,\"prod\naccount\",3yr,partial-upfront\n"))
+		require.NoError(t, err)
+		require.Len(t, recs, 1)
+		assert.Equal(t, "db.t3.micro, burstable", recs[0].ResourceType)
+		assert.Equal(t, "prod\naccount", recs[0].AccountName)
+	})
+
+	t.Run("headers only loads zero recs", func(t *testing.T) {
+		recs, err := loadRecommendationsFromCSV(writeTestRecommendationsCSV(t,
+			"Service,Region,ResourceType,Count,Term,PaymentOption\n"))
+		require.NoError(t, err)
+		assert.Empty(t, recs)
+	})
+
+	t.Run("wrong field count on a data row", func(t *testing.T) {
+		err := loadCSVContent(t,
+			"Service,Region,ResourceType,Count,Term,PaymentOption\nrds,us-east-1,db.t3.micro\n")
+		require.Error(t, err)
+		assert.Contains(t, err.Error(), "failed to read CSV record")
+	})
+
+	t.Run("TOTAL row still skipped with full header", func(t *testing.T) {
+		recs, err := loadRecommendationsFromCSV(writeTestRecommendationsCSV(t,
+			"Service,Region,ResourceType,Count,Term,PaymentOption\nrds,us-east-1,db.t3.micro,2,3yr,partial-upfront\nTOTAL,,,2,,\n"))
+		require.NoError(t, err)
+		require.Len(t, recs, 1)
+	})
+}
diff --git a/cmd/multi_service_csv_strict_test.go b/cmd/multi_service_csv_strict_test.go
index ac0ed3fba..62ced1c8c 100644
--- a/cmd/multi_service_csv_strict_test.go
+++ b/cmd/multi_service_csv_strict_test.go
@@ -26,15 +26,16 @@ func TestLoadRecommendationsFromCSV_StrictCount(t *testing.T) {
 		{"trailing garbage", "3abc"},
 		{"trailing unit", "12 units"},
 		{"negative", "-1"},
+		{"zero", "0"},
 		{"blank", ""},
 		{"whitespace only", "  "},
 		{"overflows int64", "99999999999999999999"},
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			content := "Service,Region,ResourceType,Count\n" +
-				"rds,us-east-1,db.t3.micro,1\n" +
-				"rds,us-east-1,db.t3.small,\"" + tt.cell + "\"\n"
+			content := "Service,Region,ResourceType,Count,Term,PaymentOption\n" +
+				"rds,us-east-1,db.t3.micro,1,3yr,partial-upfront\n" +
+				"rds,us-east-1,db.t3.small,\"" + tt.cell + "\",3yr,partial-upfront\n"
 			err := loadCSVContent(t, content)
 			require.Error(t, err)
 			assert.Contains(t, err.Error(), "line 3")
@@ -44,7 +45,7 @@ func TestLoadRecommendationsFromCSV_StrictCount(t *testing.T) {
 }
 
 func TestLoadRecommendationsFromCSV_CountTrimmed(t *testing.T) {
-	csvPath := writeTestRecommendationsCSV(t, "Service,Region,ResourceType,Count\nrds,us-east-1,db.t3.micro,\" 3 \"\n")
+	csvPath := writeTestRecommendationsCSV(t, "Service,Region,ResourceType,Count,Term,PaymentOption\nrds,us-east-1,db.t3.micro,\" 3 \",3yr,partial-upfront\n")
 	recs, err := loadRecommendationsFromCSV(csvPath)
 	require.NoError(t, err)
 	require.Len(t, recs, 1)
@@ -52,7 +53,7 @@ func TestLoadRecommendationsFromCSV_CountTrimmed(t *testing.T) {
 }
 
 func TestLoadRecommendationsFromCSV_CountColumnRequired(t *testing.T) {
-	err := loadCSVContent(t, "Service,Region,ResourceType\nrds,us-east-1,db.t3.micro\n")
+	err := loadCSVContent(t, "Service,Region,ResourceType,Term,PaymentOption\nrds,us-east-1,db.t3.micro,3yr,partial-upfront\n")
 	require.Error(t, err)
 	assert.Contains(t, err.Error(), "Count")
 }
@@ -62,6 +63,7 @@ func TestLoadRecommendationsFromCSV_StrictEstimatedSavings(t *testing.T) {
 		name string
 		cell string
 	}{
+		{"negative", "-100"},
 		{"trailing currency", "1000 USD"},
 		{"trailing garbage", "12.5abc"},
 		{"NaN", "NaN"},
@@ -70,8 +72,8 @@ func TestLoadRecommendationsFromCSV_StrictEstimatedSavings(t *testing.T) {
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			content := "Service,Region,ResourceType,Count,EstimatedSavings\n" +
-				"rds,us-east-1,db.t3.micro,2,\"" + tt.cell + "\"\n"
+			content := "Service,Region,ResourceType,Count,EstimatedSavings,Term,PaymentOption\n" +
+				"rds,us-east-1,db.t3.micro,2,\"" + tt.cell + "\",3yr,partial-upfront\n"
 			err := loadCSVContent(t, content)
 			require.Error(t, err)
 			assert.Contains(t, err.Error(), "line 2")
@@ -83,7 +85,7 @@ func TestLoadRecommendationsFromCSV_StrictEstimatedSavings(t *testing.T) {
 // A blank EstimatedSavings cell stays absent-as-zero: requireRankingSignal
 // relies on it to refuse a binding --max-instances cap.
 func TestLoadRecommendationsFromCSV_BlankSavingsStillZero(t *testing.T) {
-	csvPath := writeTestRecommendationsCSV(t, "Service,Region,ResourceType,Count,EstimatedSavings\nrds,us-east-1,db.t3.micro,2, \n")
+	csvPath := writeTestRecommendationsCSV(t, "Service,Region,ResourceType,Count,EstimatedSavings,Term,PaymentOption\nrds,us-east-1,db.t3.micro,2, ,3yr,partial-upfront\n")
 	recs, err := loadRecommendationsFromCSV(csvPath)
 	require.NoError(t, err)
 	require.Len(t, recs, 1)
@@ -107,8 +109,8 @@ func TestRunToolFromCSV_RejectsMalformedNumbers(t *testing.T) {
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
 			cfg := Config{
-				CSVInput: writeTestRecommendationsCSV(t, "Service,Region,ResourceType,Count,EstimatedSavings\n"+
-					"rds,us-east-1,db.t3.micro,"+tt.count+","+tt.savings+"\n"),
+				CSVInput: writeTestRecommendationsCSV(t, "Service,Region,ResourceType,Count,EstimatedSavings,Term,PaymentOption\n"+
+					"rds,us-east-1,db.t3.micro,"+tt.count+","+tt.savings+",3yr,partial-upfront\n"),
 				CSVOutput:      filepath.Join(t.TempDir(), "report.csv"),
 				AuditLog:       filepath.Join(t.TempDir(), "audit.jsonl"),
 				ActualPurchase: false,
@@ -122,3 +124,51 @@ func TestRunToolFromCSV_RejectsMalformedNumbers(t *testing.T) {
 		})
 	}
 }
+
+// Term and PaymentOption cells must be exactly what the purchase path accepts
+// (1yr/3yr, all-upfront/partial-upfront/no-upfront); a blank or unknown value
+// fails the load with the row number instead of reaching the purchase loop.
+func TestLoadRecommendationsFromCSV_TermAndPaymentValues(t *testing.T) {
+	tests := []struct {
+		name    string
+		term    string
+		payment string
+		column  string
+	}{
+		{"empty Term", "", "no-upfront", `column 5 "Term"`},
+		{"whitespace Term", " ", "no-upfront", `column 5 "Term"`},
+		{"unknown Term", "5yr", "no-upfront", `column 5 "Term"`},
+		{"numeric Term", "3", "no-upfront", `column 5 "Term"`},
+		{"uppercase Term", "3YR", "no-upfront", `column 5 "Term"`},
+		{"empty PaymentOption", "1yr", "", `column 6 "PaymentOption"`},
+		{"unknown PaymentOption", "1yr", "monthly", `column 6 "PaymentOption"`},
+		{"display-form PaymentOption", "1yr", "All Upfront", `column 6 "PaymentOption"`},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			content := "Service,Region,ResourceType,Count,Term,PaymentOption\n" +
+				"rds,us-east-1,db.t3.micro,1,1yr,all-upfront\n" +
+				"rds,us-east-1,db.t3.small,2," + tt.term + "," + tt.payment + "\n"
+			err := loadCSVContent(t, content)
+			require.Error(t, err)
+			assert.Contains(t, err.Error(), "line 3")
+			assert.Contains(t, err.Error(), tt.column)
+		})
+	}
+}
+
+func TestLoadRecommendationsFromCSV_TermAndPaymentAccepted(t *testing.T) {
+	for _, term := range []string{"1yr", "3yr"} {
+		for _, payment := range []string{"all-upfront", "partial-upfront", "no-upfront"} {
+			t.Run(term+"/"+payment, func(t *testing.T) {
+				recs, err := loadRecommendationsFromCSV(writeTestRecommendationsCSV(t,
+					"Service,Region,ResourceType,Count,Term,PaymentOption\n"+
+						"rds,us-east-1,db.t3.micro,2,"+term+","+payment+"\n"))
+				require.NoError(t, err)
+				require.Len(t, recs, 1)
+				assert.Equal(t, term, recs[0].Term)
+				assert.Equal(t, payment, recs[0].PaymentOption)
+			})
+		}
+	}
+}
diff --git a/cmd/multi_service_csv_test.go b/cmd/multi_service_csv_test.go
index 8209914c2..4ec85e604 100644
--- a/cmd/multi_service_csv_test.go
+++ b/cmd/multi_service_csv_test.go
@@ -614,8 +614,8 @@ ec2,us-west-2,t3.medium,10,123456789012,Development,1yr,all-upfront,2000.75`,
 		},
 		{
 			name: "Valid CSV with minimal fields",
-			csvContent: `Service,Region,ResourceType,Count
-elasticache,eu-west-1,cache.t3.micro,3`,
+			csvContent: `Service,Region,ResourceType,Count,Term,PaymentOption
+elasticache,eu-west-1,cache.t3.micro,3,3yr,partial-upfront`,
 			wantErr: false,
 			validate: func(t *testing.T, recs []common.Recommendation) {
 				require.Len(t, recs, 1)
@@ -627,8 +627,8 @@ elasticache,eu-west-1,cache.t3.micro,3`,
 		},
 		{
 			name: "Valid CSV with empty optional Account field",
-			csvContent: `Service,Region,ResourceType,Count,Account
-rds,us-east-1,db.t3.micro,2,`,
+			csvContent: `Service,Region,ResourceType,Count,Account,Term,PaymentOption
+rds,us-east-1,db.t3.micro,2,,3yr,partial-upfront`,
 			wantErr: false,
 			validate: func(t *testing.T, recs []common.Recommendation) {
 				require.Len(t, recs, 1)
@@ -699,8 +699,8 @@ TOTAL,,,,,3,,`,
 		},
 		{
 			name: "CSV with different column order",
-			csvContent: `Count,Service,Region,ResourceType
-7,rds,ap-south-1,db.r5.large`,
+			csvContent: `Count,Service,Region,ResourceType,Term,PaymentOption
+7,rds,ap-south-1,db.r5.large,3yr,partial-upfront`,
 			wantErr: false,
 			validate: func(t *testing.T, recs []common.Recommendation) {
 				require.Len(t, recs, 1)
@@ -721,25 +721,25 @@ TOTAL,,,,,3,,`,
 		},
 		{
 			name: "Invalid Count value - non-numeric",
-			csvContent: `Service,Region,ResourceType,Count
-rds,us-east-1,db.t3.micro,abc`,
+			csvContent: `Service,Region,ResourceType,Count,Term,PaymentOption
+rds,us-east-1,db.t3.micro,abc,3yr,partial-upfront`,
 			wantErr:     true,
 			errContains: `column 4 "Count": invalid integer`,
 		},
 		{
 			name: "Invalid EstimatedSavings value - non-numeric",
-			csvContent: `Service,Region,ResourceType,Count,EstimatedSavings
-rds,us-east-1,db.t3.micro,5,invalid`,
+			csvContent: `Service,Region,ResourceType,Count,EstimatedSavings,Term,PaymentOption
+rds,us-east-1,db.t3.micro,5,invalid,3yr,partial-upfront`,
 			wantErr:     true,
 			errContains: `column 5 "EstimatedSavings": invalid number`,
 		},
 		{
 			name: "Multiple rows with various services",
-			csvContent: `Service,Region,ResourceType,Count,EstimatedSavings
-rds,us-east-1,db.t3.micro,5,100.00
-ec2,us-west-2,t3.medium,10,200.50
-elasticache,eu-west-1,cache.t3.micro,3,50.25
-opensearch,ap-southeast-1,t3.small.search,2,75.00`,
+			csvContent: `Service,Region,ResourceType,Count,EstimatedSavings,Term,PaymentOption
+rds,us-east-1,db.t3.micro,5,100.00,3yr,partial-upfront
+ec2,us-west-2,t3.medium,10,200.50,3yr,partial-upfront
+elasticache,eu-west-1,cache.t3.micro,3,50.25,3yr,partial-upfront
+opensearch,ap-southeast-1,t3.small.search,2,75.00,3yr,partial-upfront`,
 			wantErr: false,
 			validate: func(t *testing.T, recs []common.Recommendation) {
 				require.Len(t, recs, 4)
@@ -751,8 +751,8 @@ opensearch,ap-southeast-1,t3.small.search,2,75.00`,
 		},
 		{
 			name: "CSV with large Count values",
-			csvContent: `Service,Region,ResourceType,Count
-ec2,us-east-1,t3.large,1000`,
+			csvContent: `Service,Region,ResourceType,Count,Term,PaymentOption
+ec2,us-east-1,t3.large,1000,3yr,partial-upfront`,
 			wantErr: false,
 			validate: func(t *testing.T, recs []common.Recommendation) {
 				require.Len(t, recs, 1)
@@ -761,8 +761,8 @@ ec2,us-east-1,t3.large,1000`,
 		},
 		{
 			name: "CSV with decimal savings",
-			csvContent: `Service,Region,ResourceType,Count,EstimatedSavings
-rds,us-east-1,db.t3.micro,5,1234.5678`,
+			csvContent: `Service,Region,ResourceType,Count,EstimatedSavings,Term,PaymentOption
+rds,us-east-1,db.t3.micro,5,1234.5678,3yr,partial-upfront`,
 			wantErr: false,
 			validate: func(t *testing.T, recs []common.Recommendation) {
 				require.Len(t, recs, 1)
@@ -770,13 +770,13 @@ rds,us-east-1,db.t3.micro,5,1234.5678`,
 			},
 		},
 		{
-			name: "CSV with zero values",
-			csvContent: `Service,Region,ResourceType,Count,EstimatedSavings
-rds,us-east-1,db.t3.micro,0,0`,
+			name: "CSV with zero EstimatedSavings stays valid",
+			csvContent: `Service,Region,ResourceType,Count,EstimatedSavings,Term,PaymentOption
+rds,us-east-1,db.t3.micro,5,0,3yr,partial-upfront`,
 			wantErr: false,
 			validate: func(t *testing.T, recs []common.Recommendation) {
 				require.Len(t, recs, 1)
-				assert.Equal(t, 0, recs[0].Count)
+				assert.Equal(t, 5, recs[0].Count)
 				assert.Equal(t, float64(0), recs[0].EstimatedSavings)
 			},
 		},
diff --git a/cmd/multi_service_max_instances_test.go b/cmd/multi_service_max_instances_test.go
index 4ec6144b3..8e52d4bd3 100644
--- a/cmd/multi_service_max_instances_test.go
+++ b/cmd/multi_service_max_instances_test.go
@@ -484,9 +484,9 @@ func TestCSVCapKeepsHighestSavingsNotFileOrder(t *testing.T) {
 		countPerRow  = 6
 	)
 
-	recs := csvRecsFrom(t, `rds,us-east-1,db.t3.small,postgres,6,10.00,1yr,All Upfront,123456789012
-rds,us-east-1,db.t3.medium,postgres,6,500.00,1yr,All Upfront,123456789012
-rds,us-east-1,db.t3.large,postgres,6,100.00,1yr,All Upfront,123456789012
+	recs := csvRecsFrom(t, `rds,us-east-1,db.t3.small,postgres,6,10.00,1yr,all-upfront,123456789012
+rds,us-east-1,db.t3.medium,postgres,6,500.00,1yr,all-upfront,123456789012
+rds,us-east-1,db.t3.large,postgres,6,100.00,1yr,all-upfront,123456789012
 `)
 
 	var (
@@ -537,8 +537,8 @@ func TestCSVCapDropsTruncationBelowMinCount(t *testing.T) {
 		countPerRow  = 6
 	)
 
-	recs := csvRecsFrom(t, `rds,us-east-1,db.t3.large,postgres,6,100.00,1yr,All Upfront,123456789012
-rds,us-east-1,db.t3.medium,postgres,6,500.00,1yr,All Upfront,123456789012
+	recs := csvRecsFrom(t, `rds,us-east-1,db.t3.large,postgres,6,100.00,1yr,all-upfront,123456789012
+rds,us-east-1,db.t3.medium,postgres,6,500.00,1yr,all-upfront,123456789012
 `)
 
 	var (
@@ -571,8 +571,8 @@ func TestCSVMinCountDropsRowsUnderTheFloor(t *testing.T) {
 	isolateAWSEnv(t)
 	const minCount = 5
 
-	recs := csvRecsFrom(t, `rds,us-east-1,db.t3.small,postgres,2,500.00,1yr,All Upfront,123456789012
-rds,us-east-1,db.t3.medium,postgres,6,100.00,1yr,All Upfront,123456789012
+	recs := csvRecsFrom(t, `rds,us-east-1,db.t3.small,postgres,2,500.00,1yr,all-upfront,123456789012
+rds,us-east-1,db.t3.medium,postgres,6,100.00,1yr,all-upfront,123456789012
 `)
 
 	var (
@@ -602,9 +602,9 @@ func TestRunToolFromCSVEnforcesMinCountAndCap(t *testing.T) {
 
 	// Worst-value row first, so file order and savings order disagree.
 	csvPath := writeTestRecommendationsCSV(t, csvFixtureHeader+
-		`rds,us-east-1,db.t3.small,postgres,6,10.00,1yr,All Upfront,123456789012
-rds,us-east-1,db.t3.medium,postgres,6,500.00,1yr,All Upfront,123456789012
-rds,us-east-1,db.t3.large,postgres,6,100.00,1yr,All Upfront,123456789012
+		`rds,us-east-1,db.t3.small,postgres,6,10.00,1yr,all-upfront,123456789012
+rds,us-east-1,db.t3.medium,postgres,6,500.00,1yr,all-upfront,123456789012
+rds,us-east-1,db.t3.large,postgres,6,100.00,1yr,all-upfront,123456789012
 `)
 
 	tests := []struct {
@@ -688,8 +688,8 @@ func TestCSVCapRanksBySavingsPerInstance(t *testing.T) {
 	isolateAWSEnv(t)
 	const maxInstances = 10
 
-	recs := csvRecsFrom(t, `rds,us-east-1,db.bulk.large,postgres,100,600.00,1yr,All Upfront,123456789012
-rds,us-east-1,db.rich.large,postgres,6,500.00,1yr,All Upfront,123456789012
+	recs := csvRecsFrom(t, `rds,us-east-1,db.bulk.large,postgres,100,600.00,1yr,all-upfront,123456789012
+rds,us-east-1,db.rich.large,postgres,6,500.00,1yr,all-upfront,123456789012
 `)
 
 	var (
@@ -737,8 +737,8 @@ func TestCSVCapRefusesRowsWithoutARankingSignal(t *testing.T) {
 		// Worst alphabetically first, so a name-ordered cap is visible: it
 		// keeps db.aaa.large and drops db.zzz.large regardless of file order.
 		recs := csvRecsFromHeader(t, csvNoSavingsHeader,
-			`rds,us-east-1,db.zzz.large,postgres,6,1yr,All Upfront,123456789012
-rds,us-east-1,db.aaa.large,postgres,6,1yr,All Upfront,123456789012
+			`rds,us-east-1,db.zzz.large,postgres,6,1yr,all-upfront,123456789012
+rds,us-east-1,db.aaa.large,postgres,6,1yr,all-upfront,123456789012
 `)
 
 		var err error
@@ -756,8 +756,8 @@ rds,us-east-1,db.aaa.large,postgres,6,1yr,All Upfront,123456789012
 		// The partial case is the worse one: the blank row loads as $0, ranks
 		// last, and is dropped first, with nothing saying the value was absent
 		// rather than genuinely zero.
-		recs := csvRecsFrom(t, `rds,us-east-1,db.t3.medium,postgres,6,500.00,1yr,All Upfront,123456789012
-rds,us-east-1,db.t3.small,postgres,6,,1yr,All Upfront,123456789012
+		recs := csvRecsFrom(t, `rds,us-east-1,db.t3.medium,postgres,6,500.00,1yr,all-upfront,123456789012
+rds,us-east-1,db.t3.small,postgres,6,,1yr,all-upfront,123456789012
 `)
 
 		var err error
@@ -774,8 +774,8 @@ rds,us-east-1,db.t3.small,postgres,6,,1yr,All Upfront,123456789012
 		// Nothing has to be chosen between, so nothing has to be ranked. A
 		// guard that over-blocks is as wrong as one that never fires.
 		recs := csvRecsFromHeader(t, csvNoSavingsHeader,
-			`rds,us-east-1,db.zzz.large,postgres,6,1yr,All Upfront,123456789012
-rds,us-east-1,db.aaa.large,postgres,6,1yr,All Upfront,123456789012
+			`rds,us-east-1,db.zzz.large,postgres,6,1yr,all-upfront,123456789012
+rds,us-east-1,db.aaa.large,postgres,6,1yr,all-upfront,123456789012
 `)
 
 		var (
@@ -792,8 +792,8 @@ rds,us-east-1,db.aaa.large,postgres,6,1yr,All Upfront,123456789012
 	})
 
 	t.Run("populated savings column still ranks and caps", func(t *testing.T) {
-		recs := csvRecsFrom(t, `rds,us-east-1,db.t3.small,postgres,6,10.00,1yr,All Upfront,123456789012
-rds,us-east-1,db.t3.medium,postgres,6,500.00,1yr,All Upfront,123456789012
+		recs := csvRecsFrom(t, `rds,us-east-1,db.t3.small,postgres,6,10.00,1yr,all-upfront,123456789012
+rds,us-east-1,db.t3.medium,postgres,6,500.00,1yr,all-upfront,123456789012
 `)
 
 		var (
@@ -873,8 +873,8 @@ func TestCSVCapOrderIsIndependentOfFileOrder(t *testing.T) {
 	// Two rows at the identical rate ($100 for 6 instances), listed
 	// worst-alphabetically-first.
 	rows := []string{
-		"rds,us-east-1,db.t3.zulu,postgres,6,100.00,1yr,All Upfront,123456789012\n",
-		"rds,us-east-1,db.t3.alpha,postgres,6,100.00,1yr,All Upfront,123456789012\n",
+		"rds,us-east-1,db.t3.zulu,postgres,6,100.00,1yr,all-upfront,123456789012\n",
+		"rds,us-east-1,db.t3.alpha,postgres,6,100.00,1yr,all-upfront,123456789012\n",
 	}
 
 	selection := func(t *testing.T, rows []string) []string {
diff --git a/cmd/multi_service_test.go b/cmd/multi_service_test.go
index fadc4c585..99644bf9b 100644
--- a/cmd/multi_service_test.go
+++ b/cmd/multi_service_test.go
@@ -1700,8 +1700,8 @@ func TestRunToolFromCSV(t *testing.T) {
 	// silently ignored, so every row loaded with Count=0 and the whole run
 	// processed nothing while the NotPanics assertion stayed green.
 	csvPath := writeTestRecommendationsCSV(t, `Service,Region,ResourceType,Engine,Count,Term,PaymentOption,Account
-rds,us-east-1,db.t3.small,postgres,2,1yr,All Upfront,111111111111
-elasticache,us-west-2,cache.t3.micro,redis,1,1yr,All Upfront,111111111111
+rds,us-east-1,db.t3.small,postgres,2,1yr,all-upfront,111111111111
+elasticache,us-west-2,cache.t3.micro,redis,1,1yr,all-upfront,111111111111
 `)
 
 	tests := []struct {
@@ -1768,7 +1768,7 @@ func TestRunToolFromCSV_DuplicateCheckFailureRefusesToPurchase(t *testing.T) {
 	isolateAWSEnv(t)
 
 	csvPath := writeTestRecommendationsCSV(t, `Service,Region,ResourceType,Engine,Count,Term,PaymentOption,Account
-rds,us-east-1,db.t3.small,postgres,2,1yr,All Upfront,111111111111
+rds,us-east-1,db.t3.small,postgres,2,1yr,all-upfront,111111111111
 `)
 	reportPath := filepath.Join(t.TempDir(), "report.csv")
 	auditPath := filepath.Join(t.TempDir(), "audit.jsonl")
@@ -1836,9 +1836,9 @@ func TestRunToolFromCSV_WithMaxInstances(t *testing.T) {
 	// rather than capped by name; that case is covered by
 	// TestCSVCapRefusesRowsWithoutARankingSignal.
 	csvPath := writeTestRecommendationsCSV(t, `Service,Region,ResourceType,Engine,Count,EstimatedSavings,Term,PaymentOption,Account
-rds,us-east-1,db.t3.small,postgres,10,100.00,1yr,All Upfront,123456789012
-rds,us-east-1,db.t3.medium,mysql,10,200.00,1yr,All Upfront,123456789012
-rds,us-east-1,db.t3.large,postgres,10,300.00,1yr,All Upfront,123456789012
+rds,us-east-1,db.t3.small,postgres,10,100.00,1yr,all-upfront,123456789012
+rds,us-east-1,db.t3.medium,mysql,10,200.00,1yr,all-upfront,123456789012
+rds,us-east-1,db.t3.large,postgres,10,300.00,1yr,all-upfront,123456789012
 `)
 
 	reportPath := filepath.Join(t.TempDir(), "report.csv")
@@ -1871,8 +1871,8 @@ func TestRunToolFromCSV_WithOverrideCount(t *testing.T) {
 	isolateAWSEnv(t)
 
 	csvPath := writeTestRecommendationsCSV(t, `Service,Region,ResourceType,Engine,Count,Term,PaymentOption,Account
-rds,us-east-1,db.t3.small,postgres,10,1yr,All Upfront,123456789012
-rds,us-east-1,db.t3.medium,mysql,5,1yr,All Upfront,123456789012
+rds,us-east-1,db.t3.small,postgres,10,1yr,all-upfront,123456789012
+rds,us-east-1,db.t3.medium,mysql,5,1yr,all-upfront,123456789012
 `)
 
 	reportPath := filepath.Join(t.TempDir(), "report.csv")
@@ -1917,7 +1917,7 @@ func TestRunToolFromCSV_WritesAuditRecordsOnDryRun(t *testing.T) {
 	isolateAWSEnv(t)
 
 	csvPath := writeTestRecommendationsCSV(t, `Service,Region,ResourceType,Engine,Count,Term,PaymentOption,Account
-rds,us-east-1,db.t3.small,postgres,2,1yr,All Upfront,111111111111
+rds,us-east-1,db.t3.small,postgres,2,1yr,all-upfront,111111111111
 `)
 	auditPath := filepath.Join(t.TempDir(), "audit.jsonl")
 
@@ -2025,7 +2025,7 @@ func TestRunToolFromCSV_ChecksAuditLogWritability(t *testing.T) {
 	defer func() { toolCfg = origCfg }()
 
 	csvPath := writeTestRecommendationsCSV(t, `Service,Region,ResourceType,Engine,Count,Term,PaymentOption,Account
-rds,us-east-1,db.t3.small,postgres,2,1yr,All Upfront,111111111111
+rds,us-east-1,db.t3.small,postgres,2,1yr,all-upfront,111111111111
 `)
 
 	toolCfg.CSVInput = csvPath
@@ -2060,8 +2060,8 @@ func TestRunToolFromCSV_ConfirmsOnceBeforeAnyRegion(t *testing.T) {
 	t.Cleanup(func() { log.SetOutput(oldWriter) })
 
 	csvPath := writeTestRecommendationsCSV(t, `Service,Region,ResourceType,Engine,Count,Term,PaymentOption,Account
-rds,us-east-1,db.t3.small,postgres,2,1yr,All Upfront,111111111111
-rds,us-west-2,db.t3.medium,postgres,3,1yr,All Upfront,123456789012
+rds,us-east-1,db.t3.small,postgres,2,1yr,all-upfront,111111111111
+rds,us-west-2,db.t3.medium,postgres,3,1yr,all-upfront,123456789012
 `)
 	reportPath := filepath.Join(t.TempDir(), "report.csv")
 	auditPath := filepath.Join(t.TempDir(), "audit.jsonl")
diff --git a/cmd/savings_plan_purchase_test.go b/cmd/savings_plan_purchase_test.go
index 5a4c5820d..dbf095228 100644
--- a/cmd/savings_plan_purchase_test.go
+++ b/cmd/savings_plan_purchase_test.go
@@ -331,6 +331,10 @@ func TestCSVRoundTrip_SavingsPlanAndEC2DetailsSurviveWriteThenRead(t *testing.T)
 		Details: &common.DatabaseDetails{Engine: "Aurora MySQL", AZConfig: "multi-az", InstanceClass: "db.r6g.large"}}
 
 	in := []common.Recommendation{compute, ec2sp, ec2, rds}
+	for i := range in {
+		// The CSV loader requires Term and PaymentOption on every row.
+		in[i].Term, in[i].PaymentOption = "1yr", "no-upfront"
+	}
 	results := make([]common.PurchaseResult, 0, len(in))
 	for _, r := range in {
 		results = append(results, common.PurchaseResult{Recommendation: r, Success: true, Timestamp: time.Now()})
@@ -375,7 +379,7 @@ func TestCSVWriter_AppendsDetailColumnsWithoutReorderingExistingOnes(t *testing.
 
 func TestCSVReader_MalformedHourlyCommitmentRejected(t *testing.T) {
 	path := filepath.Join(t.TempDir(), "in.csv")
-	content := "Service,Region,Count,PlanType,HourlyCommitment\nsavings-plans-compute,,1,Compute,abc\n"
+	content := "Service,Region,ResourceType,Count,PlanType,HourlyCommitment,Term,PaymentOption\nsavings-plans-compute,,Compute,1,Compute,abc,1yr,no-upfront\n"
 	require.NoError(t, os.WriteFile(path, []byte(content), 0o600))
 	_, err := loadRecommendationsFromCSV(path)
 	require.Error(t, err)
diff --git a/docs/cli/README.md b/docs/cli/README.md
index 17f69e539..b853b757f 100644
--- a/docs/cli/README.md
+++ b/docs/cli/README.md
@@ -88,7 +88,7 @@ All flags belong to the root command unless noted otherwise.
 | Flag | Short | Default | Description |
 |------|-------|---------|-------------|
 | `--output` | `-o` | auto-generated | Output CSV file path. Auto-name format: `ri-helper-dryrun-YYYYMMDD-HHMMSS.csv` or `ri-helper-purchase-YYYYMMDD-HHMMSS.csv`. |
-| `--input-csv` | `-i` | (none) | Read recommendations from an existing CSV instead of fetching from cloud APIs. The file must have a `.csv` extension and exist at the path given. |
+| `--input-csv` | `-i` | (none) | Read recommendations from an existing CSV instead of fetching from cloud APIs. The file must have a `.csv` extension and exist at the path given. The header must contain `Service`, `Region`, `ResourceType`, `Count`, `Term` and `PaymentOption`; a missing column is a load error that names it. `Term` must be `1yr` or `3yr` and `PaymentOption` must be `all-upfront`, `partial-upfront` or `no-upfront` on every row; a blank or unknown value fails the load with the row number. CSVs with only the older four columns are rejected. |
 
 ### Authentication