diff --git a/pkg/dataobj/internal/dataset/column_builder.go b/pkg/dataobj/internal/dataset/column_builder.go
index a54d782054ef1..3c67724f44c4e 100644
--- a/pkg/dataobj/internal/dataset/column_builder.go
+++ b/pkg/dataobj/internal/dataset/column_builder.go
@@ -27,9 +27,19 @@ type BuilderOptions struct {
 	// CompressionOptions holds optional configuration for compression.
 	CompressionOptions CompressionOptions
 
+	// StatisticsOptions holds optional configuration for statistics.
+	Statistics StatisticsOptions
+}
+
+// StatisticsOptions customizes the collection of statistics for a column.
+type StatisticsOptions struct {
 	// StoreRangeStats indicates whether to store value range statistics for the
 	// column and pages.
 	StoreRangeStats bool
+
+	// StoreCardinalityStats indicates whether to store cardinality estimations,
+	// facilitated by hyperloglog
+	StoreCardinalityStats bool
 }
 
 // CompressionOptions customizes the compressor used when building pages.
@@ -48,8 +58,9 @@ type ColumnBuilder struct {
 
 	rows int // Total number of rows in the column.
 
-	pages   []*MemPage
-	builder *pageBuilder
+	pages        []*MemPage
+	statsBuilder *columnStatsBuilder
+	pageBuilder  *pageBuilder
 }
 
 // NewColumnBuilder creates a new ColumnBuilder from the optional name and
@@ -61,11 +72,17 @@ func NewColumnBuilder(name string, opts BuilderOptions) (*ColumnBuilder, error)
 		return nil, fmt.Errorf("creating page builder: %w", err)
 	}
 
+	statsBuilder, err := newColumnStatsBuilder(opts.Statistics)
+	if err != nil {
+		return nil, fmt.Errorf("creating stats builder: %w", err)
+	}
+
 	return &ColumnBuilder{
 		name: name,
 		opts: opts,
 
-		builder: builder,
+		pageBuilder:  builder,
+		statsBuilder: statsBuilder,
 	}, nil
 }
 
@@ -87,6 +104,7 @@ func (cb *ColumnBuilder) Append(row int, value Value) error {
 	for range 2 {
 		if cb.append(row, value) {
 			cb.rows = row + 1
+			cb.statsBuilder.Append(value)
 			return nil
 		}
 
@@ -107,7 +125,7 @@ func (cb *ColumnBuilder) EstimatedSize() int {
 	for _, p := range cb.pages {
 		size += p.Info.CompressedSize
 	}
-	size += cb.builder.EstimatedSize()
+	size += cb.pageBuilder.EstimatedSize()
 	return size
 }
 
@@ -132,7 +150,7 @@ func (cb *ColumnBuilder) Backfill(row int) {
 
 func (cb *ColumnBuilder) backfill(row int) bool {
 	for row > cb.rows {
-		if !cb.builder.AppendNull() {
+		if !cb.pageBuilder.AppendNull() {
 			return false
 		}
 		cb.rows++
@@ -146,7 +164,7 @@ func (cb *ColumnBuilder) append(row int, value Value) bool {
 	if !cb.backfill(row) {
 		return false
 	}
-	return cb.builder.Append(value)
+	return cb.pageBuilder.Append(value)
 }
 
 // Flush converts data in cb into a [MemColumn]. Afterwards, cb is reset to a
@@ -159,7 +177,7 @@ func (cb *ColumnBuilder) Flush() (*MemColumn, error) {
 		Type: cb.opts.Value,
 
 		Compression: cb.opts.Compression,
-		Statistics:  cb.buildStats(),
+		Statistics:  cb.statsBuilder.Flush(cb.pages),
 	}
 
 	for _, page := range cb.pages {
@@ -178,54 +196,12 @@ func (cb *ColumnBuilder) Flush() (*MemColumn, error) {
 	return column, nil
 }
 
-func (cb *ColumnBuilder) buildStats() *datasetmd.Statistics {
-	if !cb.opts.StoreRangeStats {
-		return nil
-	}
-
-	var stats datasetmd.Statistics
-
-	var minValue, maxValue Value
-
-	for i, page := range cb.pages {
-		if page.Info.Stats == nil {
-			// This should never hit; if cb.opts.StoreRangeStats is true, then
-			// page.Info.Stats will be populated.
-			panic("ColumnBuilder.buildStats: page missing stats")
-		}
-
-		var pageMin, pageMax Value
-
-		if err := pageMin.UnmarshalBinary(page.Info.Stats.MinValue); err != nil {
-			panic(fmt.Sprintf("ColumnBuilder.buildStats: failed to unmarshal min value: %s", err))
-		} else if err := pageMax.UnmarshalBinary(page.Info.Stats.MaxValue); err != nil {
-			panic(fmt.Sprintf("ColumnBuilder.buildStats: failed to unmarshal max value: %s", err))
-		}
-
-		if i == 0 || CompareValues(pageMin, minValue) < 0 {
-			minValue = pageMin
-		}
-		if i == 0 || CompareValues(pageMax, maxValue) > 0 {
-			maxValue = pageMax
-		}
-	}
-
-	var err error
-	if stats.MinValue, err = minValue.MarshalBinary(); err != nil {
-		panic(fmt.Sprintf("ColumnBuilder.buildStats: failed to marshal min value: %s", err))
-	}
-	if stats.MaxValue, err = maxValue.MarshalBinary(); err != nil {
-		panic(fmt.Sprintf("ColumnBuilder.buildStats: failed to marshal max value: %s", err))
-	}
-	return &stats
-}
-
 func (cb *ColumnBuilder) flushPage() {
-	if cb.builder.Rows() == 0 {
+	if cb.pageBuilder.Rows() == 0 {
 		return
 	}
 
-	page, err := cb.builder.Flush()
+	page, err := cb.pageBuilder.Flush()
 	if err != nil {
 		// Flush should only return an error when it's empty, which we already
 		// ensure it's not in the lines above.
@@ -238,5 +214,5 @@ func (cb *ColumnBuilder) flushPage() {
 func (cb *ColumnBuilder) Reset() {
 	cb.rows = 0
 	cb.pages = nil
-	cb.builder.Reset()
+	cb.pageBuilder.Reset()
 }
diff --git a/pkg/dataobj/internal/dataset/column_stats.go b/pkg/dataobj/internal/dataset/column_stats.go
new file mode 100644
index 0000000000000..94e6f94335378
--- /dev/null
+++ b/pkg/dataobj/internal/dataset/column_stats.go
@@ -0,0 +1,115 @@
+package dataset
+
+import (
+	"fmt"
+
+	"github.com/axiomhq/hyperloglog"
+
+	"github.com/grafana/loki/v3/pkg/dataobj/internal/metadata/datasetmd"
+)
+
+// NB: https://engineering.fb.com/2018/12/13/data-infrastructure/hyperloglog/
+// Standard error (SE) = 1.04/sqrt(2^n_registers)
+// so
+//
+// 65% of estimates will be within  ±1SE of true value
+// 95% of estimates will be within  ±2SE of true value
+// 99% of estimates will be within  ±3SE of true value
+//
+// e.g. given 2^12 registers
+// SE = 1.04/sqrt(2^12) = 0.01625
+// 65% of estimates will be within  ±1SE of true value (1.625%)
+// 95% of estimates will be within  ±2SE of true value (3.25%)
+// 99% of estimates will be within  ±3SE of true value (4.875%)
+// and with 8-bit registers, this is 2^12 = 4KB size.
+func newHyperLogLog() (*hyperloglog.Sketch, error) {
+	return hyperloglog.NewSketch(12, true)
+}
+
+type columnStatsBuilder struct {
+	opts StatisticsOptions
+
+	// for cardinality
+	hll *hyperloglog.Sketch
+}
+
+// ColumnStatsBuilder is for column-level statistics
+func newColumnStatsBuilder(opts StatisticsOptions) (*columnStatsBuilder, error) {
+	result := &columnStatsBuilder{
+		opts: opts,
+	}
+
+	if opts.StoreCardinalityStats {
+		var err error
+		if result.hll, err = newHyperLogLog(); err != nil {
+			return nil, fmt.Errorf("failed to create hll: %w", err)
+		}
+	}
+
+	return result, nil
+}
+
+func (csb *columnStatsBuilder) Append(value Value) {
+	if csb.opts.StoreCardinalityStats && !value.IsNil() && !value.IsZero() {
+		buf, err := value.MarshalBinary()
+		if err != nil {
+			panic(fmt.Sprintf(
+				"failed to marshal value for cardinality stats of type %s: %s",
+				value.Type(), err,
+			))
+		}
+
+		// TODO(owen-d): improve efficiency, ideally we don't need to marshal
+		// into an intermediate buffer.
+		csb.hll.Insert(buf)
+	}
+}
+
+// Flush builds the column-level stats both from the given pages and any internal
+// state
+func (csb *columnStatsBuilder) Flush(pages []*MemPage) *datasetmd.Statistics {
+	var dst datasetmd.Statistics
+	if csb.opts.StoreCardinalityStats {
+		dst.CardinalityCount = csb.hll.Estimate()
+	}
+	if csb.opts.StoreRangeStats {
+		csb.buildRangeStats(pages, &dst)
+	}
+
+	return &dst
+}
+
+func (csb *columnStatsBuilder) buildRangeStats(pages []*MemPage, dst *datasetmd.Statistics) {
+	var minValue, maxValue Value
+
+	for i, page := range pages {
+		if page.Info.Stats == nil {
+			// This should never hit; if cb.opts.StoreRangeStats is true, then
+			// page.Info.Stats will be populated.
+			panic("ColumnStatsBuilder.buildStats: page missing stats")
+		}
+
+		var pageMin, pageMax Value
+
+		if err := pageMin.UnmarshalBinary(page.Info.Stats.MinValue); err != nil {
+			panic(fmt.Sprintf("ColumnStatsBuilder.buildStats: failed to unmarshal min value: %s", err))
+		} else if err := pageMax.UnmarshalBinary(page.Info.Stats.MaxValue); err != nil {
+			panic(fmt.Sprintf("ColumnStatsBuilder.buildStats: failed to unmarshal max value: %s", err))
+		}
+
+		if i == 0 || CompareValues(pageMin, minValue) < 0 {
+			minValue = pageMin
+		}
+		if i == 0 || CompareValues(pageMax, maxValue) > 0 {
+			maxValue = pageMax
+		}
+	}
+
+	var err error
+	if dst.MinValue, err = minValue.MarshalBinary(); err != nil {
+		panic(fmt.Sprintf("ColumnStatsBuilder.buildStats: failed to marshal min value: %s", err))
+	}
+	if dst.MaxValue, err = maxValue.MarshalBinary(); err != nil {
+		panic(fmt.Sprintf("ColumnStatsBuilder.buildStats: failed to marshal max value: %s", err))
+	}
+}
diff --git a/pkg/dataobj/internal/dataset/column_test.go b/pkg/dataobj/internal/dataset/column_test.go
index de2f3f3549461..ca8c39cc9ae91 100644
--- a/pkg/dataobj/internal/dataset/column_test.go
+++ b/pkg/dataobj/internal/dataset/column_test.go
@@ -101,7 +101,9 @@ func TestColumnBuilder_MinMax(t *testing.T) {
 		Compression:  datasetmd.COMPRESSION_TYPE_NONE,
 		Encoding:     datasetmd.ENCODING_TYPE_PLAIN,
 
-		StoreRangeStats: true,
+		Statistics: StatisticsOptions{
+			StoreRangeStats: true,
+		},
 	}
 	b, err := NewColumnBuilder("", opts)
 	require.NoError(t, err)
@@ -132,6 +134,56 @@ func TestColumnBuilder_MinMax(t *testing.T) {
 	require.Equal(t, fString, page1Max.String())
 }
 
+func TestColumnBuilder_Cardinality(t *testing.T) {
+	var (
+		// We include the null string in the test to ensure that it's never
+		// considered in min/max ranges.
+		nullString = ""
+
+		aString = strings.Repeat("a", 100)
+		bString = strings.Repeat("b", 100)
+		cString = strings.Repeat("c", 100)
+	)
+
+	// We store nulls and duplicates (should not be counted in cardinality count)
+	in := []string{
+		nullString,
+
+		aString,
+
+		bString,
+		bString,
+		bString,
+
+		cString,
+	}
+
+	opts := BuilderOptions{
+		PageSizeHint: 301, // Slightly larger than the string length of 3 strings per page.
+		Value:        datasetmd.VALUE_TYPE_STRING,
+		Compression:  datasetmd.COMPRESSION_TYPE_NONE,
+		Encoding:     datasetmd.ENCODING_TYPE_PLAIN,
+
+		Statistics: StatisticsOptions{
+			StoreCardinalityStats: true,
+		},
+	}
+	b, err := NewColumnBuilder("", opts)
+	require.NoError(t, err)
+
+	for i, s := range in {
+		require.NoError(t, b.Append(i, StringValue(s)))
+	}
+
+	col, err := b.Flush()
+	require.NoError(t, err)
+	require.Equal(t, datasetmd.VALUE_TYPE_STRING, col.Info.Type)
+	require.NotNil(t, col.Info.Statistics)
+	// we use sparse hyperloglog reprs until a certain cardinality is reached,
+	// so this should not be approximate at low counts.
+	require.Equal(t, uint64(3), col.Info.Statistics.CardinalityCount)
+}
+
 func getMinMax(t *testing.T, stats *datasetmd.Statistics) (min, max Value) {
 	t.Helper()
 	require.NotNil(t, stats)
diff --git a/pkg/dataobj/internal/dataset/page_builder.go b/pkg/dataobj/internal/dataset/page_builder.go
index fa4963db1661b..809902587e84b 100644
--- a/pkg/dataobj/internal/dataset/page_builder.go
+++ b/pkg/dataobj/internal/dataset/page_builder.go
@@ -99,10 +99,10 @@ func (b *pageBuilder) Append(value Value) bool {
 		return false
 	}
 
-	// Update min/max values for stats. We only do this for non-NULL values,
+	// Update statistics. We only do this for non-NULL values,
 	// otherwise NULL would always be the min for columns that contain a single
 	// NULL.
-	b.updateMinMax(value)
+	b.accumulateStatistics(value)
 
 	// The following calls won't fail; they only return errors when the
 	// underlying writers fail, which ours cannot.
@@ -140,14 +140,16 @@ func (b *pageBuilder) AppendNull() bool {
 	return true
 }
 
-func (b *pageBuilder) updateMinMax(value Value) {
+func (b *pageBuilder) accumulateStatistics(value Value) {
 	// As a small optimization, we only update min/max values if we're intending
-	// on populating the stats. This avoids unnecessary comparisons for very
+	// on populating them in statistics. This avoids unnecessary comparisons for very
 	// large values.
-	if !b.opts.StoreRangeStats {
-		return
+	if b.opts.Statistics.StoreRangeStats {
+		b.updateMinMax(value)
 	}
+}
 
+func (b *pageBuilder) updateMinMax(value Value) {
 	// We'll init minValue/maxValue if this is our first non-NULL value (b.values == 0).
 	// This allows us to only avoid comparing against NULL values, which would lead to
 	// NULL always being the min.
@@ -259,9 +261,15 @@ func (b *pageBuilder) Flush() (*MemPage, error) {
 }
 
 func (b *pageBuilder) buildStats() *datasetmd.Statistics {
-	if !b.opts.StoreRangeStats {
-		return nil
+	var stats datasetmd.Statistics
+	if b.opts.Statistics.StoreRangeStats {
+		b.buildRangeStats(&stats)
+		return &stats
 	}
+	return nil
+}
+
+func (b *pageBuilder) buildRangeStats(dst *datasetmd.Statistics) {
 
 	minValueBytes, err := b.minValue.MarshalBinary()
 	if err != nil {
@@ -272,10 +280,8 @@ func (b *pageBuilder) buildStats() *datasetmd.Statistics {
 		panic(fmt.Sprintf("pageBuilder.buildStats: failed to marshal max value: %s", err))
 	}
 
-	return &datasetmd.Statistics{
-		MinValue: minValueBytes,
-		MaxValue: maxValueBytes,
-	}
+	dst.MinValue = minValueBytes
+	dst.MaxValue = maxValueBytes
 }
 
 // Reset resets the pageBuilder to a fresh state, allowing it to be reused.
diff --git a/pkg/dataobj/internal/metadata/datasetmd/datasetmd.pb.go b/pkg/dataobj/internal/metadata/datasetmd/datasetmd.pb.go
index 1920ffde67281..1ee081cd83e2b 100644
--- a/pkg/dataobj/internal/metadata/datasetmd/datasetmd.pb.go
+++ b/pkg/dataobj/internal/metadata/datasetmd/datasetmd.pb.go
@@ -268,6 +268,11 @@ type Statistics struct {
 	// Applications must not assume that an unset max_value means that the column
 	// is empty; check for values_count == 0 instead.
 	MaxValue []byte `protobuf:"bytes,2,opt,name=max_value,json=maxValue,proto3" json:"max_value,omitempty"`
+	// Estimated number of distinct values in the column.
+	//
+	// Applications must not assume that an unset cardinality_count means that
+	// the column has no distinct values; check for values_count == 0 instead.
+	CardinalityCount uint64 `protobuf:"varint,3,opt,name=cardinality_count,json=cardinalityCount,proto3" json:"cardinality_count,omitempty"`
 }
 
 func (m *Statistics) Reset()      { *m = Statistics{} }
@@ -316,6 +321,13 @@ func (m *Statistics) GetMaxValue() []byte {
 	return nil
 }
 
+func (m *Statistics) GetCardinalityCount() uint64 {
+	if m != nil {
+		return m.CardinalityCount
+	}
+	return 0
+}
+
 // Page describes an individual page within a column.
 type PageInfo struct {
 	// Uncompressed size of the page within the data object.
@@ -448,51 +460,52 @@ func init() {
 }
 
 var fileDescriptor_7ab9d5b21b743868 = []byte{
-	// 699 bytes of a gzipped FileDescriptorProto
-	0x1f, 0x8b, 0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0xff, 0xac, 0x54, 0x41, 0x53, 0xda, 0x5a,
-	0x14, 0xe6, 0x02, 0xfa, 0xc8, 0x01, 0x35, 0xde, 0xa7, 0xcf, 0xf8, 0x78, 0xe6, 0x51, 0x3b, 0x53,
-	0xa9, 0xed, 0xc0, 0x14, 0x3b, 0xed, 0x1a, 0x21, 0x3a, 0x99, 0xd1, 0x90, 0x49, 0xd0, 0x19, 0xdd,
-	0x64, 0x62, 0xb8, 0xd0, 0x54, 0x92, 0x30, 0x24, 0x50, 0x75, 0xd5, 0x95, 0xeb, 0xfe, 0x8c, 0xfe,
-	0x94, 0x2e, 0x5d, 0xba, 0xac, 0x71, 0xa6, 0xd3, 0xa5, 0x3f, 0xa1, 0xc3, 0x85, 0x40, 0x04, 0xca,
-	0xb8, 0xe8, 0xee, 0xf2, 0x7d, 0xdf, 0x39, 0x27, 0x7c, 0xdf, 0xb9, 0x17, 0xde, 0xb7, 0xce, 0x1b,
-	0xf9, 0x9a, 0xee, 0xe9, 0xce, 0xd9, 0xc7, 0xbc, 0x69, 0x7b, 0xa4, 0x6d, 0xeb, 0xcd, 0xbc, 0x45,
-	0x3c, 0xbd, 0x07, 0x52, 0xc6, 0x25, 0x9e, 0x55, 0x1b, 0x9d, 0x72, 0xad, 0xb6, 0xe3, 0x39, 0x38,
-	0x3d, 0x28, 0xca, 0x05, 0xda, 0xdc, 0x40, 0x91, 0xeb, 0xbe, 0xd9, 0xfc, 0x11, 0x03, 0x28, 0x39,
-	0xcd, 0x8e, 0x65, 0x8b, 0x76, 0xdd, 0xc1, 0x18, 0xe2, 0xb6, 0x6e, 0x11, 0x0e, 0x65, 0x50, 0x96,
-	0x51, 0xe8, 0x19, 0x0b, 0x00, 0x5d, 0xbd, 0xd9, 0x21, 0x9a, 0x77, 0xd9, 0x22, 0x5c, 0x34, 0x83,
-	0xb2, 0x8b, 0x85, 0x17, 0xb9, 0x19, 0x4d, 0x73, 0xc7, 0x3d, 0x79, 0xf5, 0xb2, 0x45, 0x14, 0xa6,
-	0x1b, 0x1c, 0xf1, 0x06, 0x40, 0xdb, 0xf9, 0xe4, 0x6a, 0x86, 0xd3, 0xb1, 0x3d, 0x2e, 0x96, 0x41,
-	0xd9, 0xb8, 0xc2, 0xf4, 0x90, 0x52, 0x0f, 0xc0, 0x12, 0x24, 0x0d, 0xc7, 0x6a, 0xb5, 0x89, 0xeb,
-	0x9a, 0x8e, 0xcd, 0xc5, 0xe9, 0x98, 0xd7, 0x33, 0xc7, 0x94, 0x46, 0x7a, 0x3a, 0x2c, 0xdc, 0x00,
-	0xbf, 0x82, 0xe5, 0x8e, 0x1d, 0x00, 0xa4, 0xa6, 0xb9, 0xe6, 0x15, 0xe1, 0xe6, 0xe8, 0x54, 0x36,
-	0x4c, 0xa8, 0xe6, 0x15, 0xc1, 0x5b, 0xb0, 0x34, 0x2e, 0x9d, 0xa7, 0xd2, 0xc5, 0x49, 0x61, 0xf0,
-	0x25, 0x9a, 0x53, 0xaf, 0xbb, 0xc4, 0xe3, 0xfe, 0xea, 0x0b, 0x03, 0xb8, 0x42, 0x51, 0xfc, 0x1c,
-	0x16, 0x86, 0x42, 0xda, 0x2f, 0x41, 0x65, 0xa9, 0x00, 0xa4, 0xdd, 0xf6, 0x01, 0x5c, 0x4f, 0xf7,
-	0x4c, 0xd7, 0x33, 0x0d, 0x97, 0x63, 0x32, 0x28, 0x9b, 0x2c, 0x6c, 0xcd, 0xfc, 0xcb, 0xea, 0x50,
-	0xae, 0x84, 0x4a, 0xf1, 0x33, 0x48, 0x51, 0xa3, 0x03, 0x77, 0x81, 0x0e, 0x4b, 0xf6, 0x31, 0xea,
-	0xef, 0xe6, 0x1e, 0xc0, 0xa8, 0x18, 0xa7, 0x81, 0xb1, 0x4c, 0x5b, 0xa3, 0x02, 0x1a, 0x76, 0x4a,
-	0x49, 0x58, 0xa6, 0x4d, 0x83, 0xa3, 0xa4, 0x7e, 0x31, 0x20, 0xa3, 0x03, 0x52, 0xbf, 0xa0, 0xe4,
-	0xe6, 0x75, 0x0c, 0x12, 0xb2, 0xde, 0x20, 0x74, 0x5d, 0xa6, 0x9a, 0x8c, 0x9e, 0x6e, 0x72, 0x74,
-	0xaa, 0xc9, 0x2b, 0x30, 0x67, 0xb4, 0x8d, 0x9d, 0x02, 0x5d, 0x92, 0x05, 0xa5, 0xff, 0x63, 0x6c,
-	0x7f, 0xe2, 0xe3, 0xfb, 0x23, 0x40, 0x82, 0xd8, 0x86, 0x53, 0x33, 0xed, 0x06, 0x8d, 0x79, 0xb1,
-	0xf0, 0x72, 0xa6, 0x93, 0xc2, 0x40, 0x4c, 0x37, 0x67, 0x58, 0x8a, 0xff, 0x87, 0x64, 0x38, 0xdc,
-	0xfe, 0x16, 0x40, 0x28, 0xd8, 0x34, 0x30, 0xa3, 0x50, 0xfb, 0xd9, 0x27, 0x7e, 0x13, 0x68, 0xe2,
-	0xcf, 0x05, 0xca, 0x4c, 0x04, 0xba, 0x7d, 0x0e, 0xcc, 0xf0, 0x9e, 0xe1, 0x7f, 0xe1, 0x9f, 0xe3,
-	0xe2, 0xc1, 0x91, 0xa0, 0x55, 0x4f, 0x64, 0x41, 0x3b, 0x92, 0x54, 0x59, 0x28, 0x89, 0x7b, 0xa2,
-	0x50, 0x66, 0x23, 0x78, 0x05, 0xd8, 0x10, 0x27, 0x4a, 0xd5, 0x77, 0x6f, 0x59, 0x84, 0x57, 0x61,
-	0x39, 0x5c, 0xd1, 0x87, 0xa3, 0x63, 0xb0, 0x5a, 0x55, 0x44, 0x69, 0x9f, 0x8d, 0x6d, 0x5f, 0x23,
-	0x58, 0x1a, 0xbb, 0x6e, 0x38, 0x03, 0xff, 0x95, 0x2a, 0x87, 0xb2, 0x22, 0xa8, 0xaa, 0x58, 0x91,
-	0xa6, 0x4d, 0x5e, 0x87, 0xd5, 0x09, 0x85, 0x54, 0x91, 0x04, 0x16, 0xe1, 0x34, 0xac, 0x4d, 0x50,
-	0xaa, 0x54, 0x94, 0xe5, 0x13, 0x36, 0x3a, 0xb5, 0xee, 0x54, 0xad, 0x96, 0xd9, 0xd8, 0xf6, 0x25,
-	0xa4, 0xc2, 0xc9, 0xe1, 0x0d, 0x58, 0x17, 0xa4, 0x52, 0xa5, 0x2c, 0x4a, 0xfb, 0xd3, 0xbe, 0x60,
-	0x0d, 0xfe, 0x7e, 0x4c, 0xcb, 0x07, 0x45, 0x51, 0x62, 0xd1, 0x24, 0x51, 0x16, 0x0e, 0xaa, 0x45,
-	0x36, 0x8a, 0x39, 0x58, 0x79, 0x4c, 0xec, 0x8a, 0xd5, 0xc3, 0xa2, 0xcc, 0xc6, 0x76, 0x2f, 0x6e,
-	0xee, 0xf8, 0xc8, 0xed, 0x1d, 0x1f, 0x79, 0xb8, 0xe3, 0xd1, 0x67, 0x9f, 0x47, 0x5f, 0x7d, 0x1e,
-	0x7d, 0xf3, 0x79, 0x74, 0xe3, 0xf3, 0xe8, 0xbb, 0xcf, 0xa3, 0x9f, 0x3e, 0x1f, 0x79, 0xf0, 0x79,
-	0xf4, 0xe5, 0x9e, 0x8f, 0xdc, 0xdc, 0xf3, 0x91, 0xdb, 0x7b, 0x3e, 0x72, 0xba, 0xdb, 0x30, 0xbd,
-	0x0f, 0x9d, 0xb3, 0x9c, 0xe1, 0x58, 0xf9, 0x46, 0x5b, 0xaf, 0xeb, 0xb6, 0x9e, 0x6f, 0x3a, 0xe7,
-	0x66, 0xbe, 0xbb, 0x93, 0x7f, 0xe2, 0x8b, 0x7e, 0x36, 0x4f, 0x1f, 0xf2, 0x9d, 0x5f, 0x01, 0x00,
-	0x00, 0xff, 0xff, 0xe6, 0x72, 0xec, 0xb1, 0x03, 0x06, 0x00, 0x00,
+	// 714 bytes of a gzipped FileDescriptorProto
+	0x1f, 0x8b, 0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0xff, 0xac, 0x54, 0x41, 0x53, 0xda, 0x5c,
+	0x14, 0xe5, 0x01, 0xfa, 0x91, 0x0b, 0x6a, 0x7c, 0x9f, 0xd6, 0x58, 0x6a, 0x4a, 0xed, 0x4c, 0xa5,
+	0xda, 0x81, 0x29, 0x76, 0xda, 0x35, 0x42, 0xea, 0x64, 0x46, 0x43, 0x26, 0x41, 0x67, 0x74, 0xc3,
+	0x3c, 0x43, 0xa0, 0xa9, 0x24, 0x61, 0x48, 0xa0, 0xe2, 0xaa, 0x2b, 0xd7, 0xfd, 0x19, 0xfd, 0x29,
+	0x5d, 0xba, 0x74, 0x59, 0x71, 0xa6, 0xd3, 0xa5, 0x3f, 0xa1, 0xc3, 0x0b, 0x81, 0x08, 0x94, 0x71,
+	0xd1, 0xdd, 0xe3, 0x9c, 0x73, 0xef, 0x7d, 0xdc, 0x73, 0xf2, 0xe0, 0x43, 0xf3, 0xbc, 0x9e, 0xad,
+	0x12, 0x97, 0xd8, 0x67, 0x9f, 0xb3, 0x86, 0xe5, 0xea, 0x2d, 0x8b, 0x34, 0xb2, 0xa6, 0xee, 0x92,
+	0x3e, 0x48, 0x19, 0x47, 0x77, 0xcd, 0xea, 0xe8, 0x94, 0x69, 0xb6, 0x6c, 0xd7, 0xc6, 0xc9, 0x41,
+	0x51, 0xc6, 0xd7, 0x66, 0x06, 0x8a, 0x4c, 0xe7, 0xed, 0xe6, 0xaf, 0x08, 0x40, 0xc1, 0x6e, 0xb4,
+	0x4d, 0x4b, 0xb4, 0x6a, 0x36, 0xc6, 0x10, 0xb5, 0x88, 0xa9, 0x73, 0x28, 0x85, 0xd2, 0x8c, 0x42,
+	0xcf, 0x58, 0x00, 0xe8, 0x90, 0x46, 0x5b, 0xaf, 0xb8, 0xdd, 0xa6, 0xce, 0x85, 0x53, 0x28, 0xbd,
+	0x98, 0x7b, 0x95, 0x99, 0xd1, 0x34, 0x73, 0xdc, 0x97, 0x97, 0xbb, 0x4d, 0x5d, 0x61, 0x3a, 0xfe,
+	0x11, 0x6f, 0x00, 0xb4, 0xec, 0x2f, 0x4e, 0x45, 0xb3, 0xdb, 0x96, 0xcb, 0x45, 0x52, 0x28, 0x1d,
+	0x55, 0x98, 0x3e, 0x52, 0xe8, 0x03, 0x58, 0x82, 0xb8, 0x66, 0x9b, 0xcd, 0x96, 0xee, 0x38, 0x86,
+	0x6d, 0x71, 0x51, 0x3a, 0xe6, 0xcd, 0xcc, 0x31, 0x85, 0x91, 0x9e, 0x0e, 0x0b, 0x36, 0xc0, 0x3b,
+	0xb0, 0xdc, 0xb6, 0x7c, 0x40, 0xaf, 0x56, 0x1c, 0xe3, 0x52, 0xe7, 0xe6, 0xe8, 0x54, 0x36, 0x48,
+	0xa8, 0xc6, 0xa5, 0x8e, 0xb7, 0x60, 0x69, 0x5c, 0x3a, 0x4f, 0xa5, 0x8b, 0x93, 0x42, 0xff, 0x26,
+	0x15, 0xbb, 0x56, 0x73, 0x74, 0x97, 0xfb, 0xcf, 0x13, 0xfa, 0x70, 0x89, 0xa2, 0xf8, 0x25, 0x2c,
+	0x0c, 0x85, 0xb4, 0x5f, 0x8c, 0xca, 0x12, 0x3e, 0x48, 0xbb, 0xed, 0x03, 0x38, 0x2e, 0x71, 0x0d,
+	0xc7, 0x35, 0x34, 0x87, 0x63, 0x52, 0x28, 0x1d, 0xcf, 0x6d, 0xcd, 0xfc, 0xcb, 0xea, 0x50, 0xae,
+	0x04, 0x4a, 0xf1, 0x0b, 0x48, 0xd0, 0x45, 0xfb, 0xdb, 0x05, 0x3a, 0x2c, 0xee, 0x61, 0x74, 0xbf,
+	0x9b, 0x0e, 0xc0, 0xa8, 0x18, 0x27, 0x81, 0x31, 0x0d, 0xab, 0x42, 0x05, 0xd4, 0xec, 0x84, 0x12,
+	0x33, 0x0d, 0x8b, 0x1a, 0x47, 0x49, 0x72, 0x31, 0x20, 0xc3, 0x03, 0x92, 0x5c, 0x78, 0xe4, 0x0e,
+	0x2c, 0x6b, 0xa4, 0x55, 0x35, 0x2c, 0xd2, 0x30, 0xdc, 0xee, 0x03, 0x37, 0xd9, 0x00, 0xe1, 0x0d,
+	0xbd, 0x8a, 0x40, 0x4c, 0x26, 0x75, 0x9d, 0x66, 0x6b, 0xaa, 0x23, 0xe8, 0xf1, 0x8e, 0x84, 0xa7,
+	0x3a, 0xb2, 0x02, 0x73, 0x5a, 0x4b, 0xdb, 0xcd, 0xd1, 0x3b, 0x2c, 0x28, 0xde, 0x8f, 0xb1, 0xb0,
+	0x45, 0xc7, 0xc3, 0x26, 0x40, 0x4c, 0xb7, 0x34, 0xbb, 0x6a, 0x58, 0x75, 0x9a, 0x89, 0xc5, 0xdc,
+	0xeb, 0x99, 0x6b, 0x17, 0x06, 0x62, 0x1a, 0xb3, 0x61, 0x29, 0x7e, 0x0e, 0xf1, 0x60, 0x12, 0xbc,
+	0xc8, 0x40, 0x20, 0x05, 0x49, 0x60, 0x46, 0x09, 0xf0, 0x82, 0x12, 0xfb, 0x8b, 0xfb, 0xb1, 0x7f,
+	0xe7, 0x3e, 0x33, 0xe1, 0xfe, 0xf6, 0x39, 0x30, 0xc3, 0x8f, 0x12, 0x3f, 0x85, 0x27, 0xc7, 0xf9,
+	0x83, 0x23, 0xa1, 0x52, 0x3e, 0x91, 0x85, 0xca, 0x91, 0xa4, 0xca, 0x42, 0x41, 0xfc, 0x28, 0x0a,
+	0x45, 0x36, 0x84, 0x57, 0x80, 0x0d, 0x70, 0xa2, 0x54, 0x7e, 0xff, 0x8e, 0x45, 0x78, 0x15, 0x96,
+	0x83, 0x15, 0x1e, 0x1c, 0x1e, 0x83, 0xd5, 0xb2, 0x22, 0x4a, 0xfb, 0x6c, 0x64, 0xfb, 0x0a, 0xc1,
+	0xd2, 0xd8, 0xb7, 0x89, 0x53, 0xf0, 0xac, 0x50, 0x3a, 0x94, 0x15, 0x41, 0x55, 0xc5, 0x92, 0x34,
+	0x6d, 0xf2, 0x3a, 0xac, 0x4e, 0x28, 0xa4, 0x92, 0x24, 0xb0, 0x08, 0x27, 0x61, 0x6d, 0x82, 0x52,
+	0xa5, 0xbc, 0x2c, 0x9f, 0xb0, 0xe1, 0xa9, 0x75, 0xa7, 0x6a, 0xb9, 0xc8, 0x46, 0xb6, 0xbb, 0x90,
+	0x08, 0x3a, 0x87, 0x37, 0x60, 0x5d, 0x90, 0x0a, 0xa5, 0xa2, 0x28, 0xed, 0x4f, 0xbb, 0xc1, 0x1a,
+	0xfc, 0xff, 0x90, 0x96, 0x0f, 0xf2, 0xa2, 0xc4, 0xa2, 0x49, 0xa2, 0x28, 0x1c, 0x94, 0xf3, 0x6c,
+	0x18, 0x73, 0xb0, 0xf2, 0x90, 0xd8, 0x13, 0xcb, 0x87, 0x79, 0x99, 0x8d, 0xec, 0x5d, 0x5c, 0xdf,
+	0xf2, 0xa1, 0x9b, 0x5b, 0x3e, 0x74, 0x7f, 0xcb, 0xa3, 0xaf, 0x3d, 0x1e, 0x7d, 0xef, 0xf1, 0xe8,
+	0x47, 0x8f, 0x47, 0xd7, 0x3d, 0x1e, 0xfd, 0xec, 0xf1, 0xe8, 0x77, 0x8f, 0x0f, 0xdd, 0xf7, 0x78,
+	0xf4, 0xed, 0x8e, 0x0f, 0x5d, 0xdf, 0xf1, 0xa1, 0x9b, 0x3b, 0x3e, 0x74, 0xba, 0x57, 0x37, 0xdc,
+	0x4f, 0xed, 0xb3, 0x8c, 0x66, 0x9b, 0xd9, 0x7a, 0x8b, 0xd4, 0x88, 0x45, 0xb2, 0x0d, 0xfb, 0xdc,
+	0xc8, 0x76, 0x76, 0xb3, 0x8f, 0x7c, 0xfe, 0xcf, 0xe6, 0xe9, 0xab, 0xbf, 0xfb, 0x27, 0x00, 0x00,
+	0xff, 0xff, 0xbf, 0x4d, 0x54, 0xfd, 0x30, 0x06, 0x00, 0x00,
 }
 
 func (x ValueType) String() string {
@@ -592,6 +605,9 @@ func (this *Statistics) Equal(that interface{}) bool {
 	if !bytes.Equal(this.MaxValue, that1.MaxValue) {
 		return false
 	}
+	if this.CardinalityCount != that1.CardinalityCount {
+		return false
+	}
 	return true
 }
 func (this *PageInfo) Equal(that interface{}) bool {
@@ -667,10 +683,11 @@ func (this *Statistics) GoString() string {
 	if this == nil {
 		return "nil"
 	}
-	s := make([]string, 0, 6)
+	s := make([]string, 0, 7)
 	s = append(s, "&datasetmd.Statistics{")
 	s = append(s, "MinValue: "+fmt.Sprintf("%#v", this.MinValue)+",\n")
 	s = append(s, "MaxValue: "+fmt.Sprintf("%#v", this.MaxValue)+",\n")
+	s = append(s, "CardinalityCount: "+fmt.Sprintf("%#v", this.CardinalityCount)+",\n")
 	s = append(s, "}")
 	return strings.Join(s, "")
 }
@@ -804,6 +821,11 @@ func (m *Statistics) MarshalToSizedBuffer(dAtA []byte) (int, error) {
 	_ = i
 	var l int
 	_ = l
+	if m.CardinalityCount != 0 {
+		i = encodeVarintDatasetmd(dAtA, i, uint64(m.CardinalityCount))
+		i--
+		dAtA[i] = 0x18
+	}
 	if len(m.MaxValue) > 0 {
 		i -= len(m.MaxValue)
 		copy(dAtA[i:], m.MaxValue)
@@ -962,6 +984,9 @@ func (m *Statistics) Size() (n int) {
 	if l > 0 {
 		n += 1 + l + sovDatasetmd(uint64(l))
 	}
+	if m.CardinalityCount != 0 {
+		n += 1 + sovDatasetmd(uint64(m.CardinalityCount))
+	}
 	return n
 }
 
@@ -1034,6 +1059,7 @@ func (this *Statistics) String() string {
 	s := strings.Join([]string{`&Statistics{`,
 		`MinValue:` + fmt.Sprintf("%v", this.MinValue) + `,`,
 		`MaxValue:` + fmt.Sprintf("%v", this.MaxValue) + `,`,
+		`CardinalityCount:` + fmt.Sprintf("%v", this.CardinalityCount) + `,`,
 		`}`,
 	}, "")
 	return s
@@ -1434,6 +1460,25 @@ func (m *Statistics) Unmarshal(dAtA []byte) error {
 				m.MaxValue = []byte{}
 			}
 			iNdEx = postIndex
+		case 3:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field CardinalityCount", wireType)
+			}
+			m.CardinalityCount = 0
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowDatasetmd
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := dAtA[iNdEx]
+				iNdEx++
+				m.CardinalityCount |= uint64(b&0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
 		default:
 			iNdEx = preIndex
 			skippy, err := skipDatasetmd(dAtA[iNdEx:])
diff --git a/pkg/dataobj/internal/metadata/datasetmd/datasetmd.proto b/pkg/dataobj/internal/metadata/datasetmd/datasetmd.proto
index 450b5c241bcf7..c03e10e041c25 100644
--- a/pkg/dataobj/internal/metadata/datasetmd/datasetmd.proto
+++ b/pkg/dataobj/internal/metadata/datasetmd/datasetmd.proto
@@ -86,6 +86,12 @@ message Statistics {
   // Applications must not assume that an unset max_value means that the column
   // is empty; check for values_count == 0 instead.
   bytes max_value = 2;
+
+  // Estimated number of distinct values in the column.
+  //
+  // Applications must not assume that an unset cardinality_count means that
+  // the column has no distinct values; check for values_count == 0 instead.
+  uint64 cardinality_count = 3;
 }
 
 // Page describes an individual page within a column.
diff --git a/pkg/dataobj/internal/sections/logs/table.go b/pkg/dataobj/internal/sections/logs/table.go
index 82658e967fa40..a2427bda59351 100644
--- a/pkg/dataobj/internal/sections/logs/table.go
+++ b/pkg/dataobj/internal/sections/logs/table.go
@@ -116,11 +116,13 @@ func (b *tableBuffer) StreamID(pageSize int) *dataset.ColumnBuilder {
 	}
 
 	col, err := dataset.NewColumnBuilder("", dataset.BuilderOptions{
-		PageSizeHint:    pageSize,
-		Value:           datasetmd.VALUE_TYPE_INT64,
-		Encoding:        datasetmd.ENCODING_TYPE_DELTA,
-		Compression:     datasetmd.COMPRESSION_TYPE_NONE,
-		StoreRangeStats: true,
+		PageSizeHint: pageSize,
+		Value:        datasetmd.VALUE_TYPE_INT64,
+		Encoding:     datasetmd.ENCODING_TYPE_DELTA,
+		Compression:  datasetmd.COMPRESSION_TYPE_NONE,
+		Statistics: dataset.StatisticsOptions{
+			StoreRangeStats: true,
+		},
 	})
 	if err != nil {
 		// We control the Value/Encoding tuple so this can't fail; if it does,
@@ -140,11 +142,13 @@ func (b *tableBuffer) Timestamp(pageSize int) *dataset.ColumnBuilder {
 	}
 
 	col, err := dataset.NewColumnBuilder("", dataset.BuilderOptions{
-		PageSizeHint:    pageSize,
-		Value:           datasetmd.VALUE_TYPE_INT64,
-		Encoding:        datasetmd.ENCODING_TYPE_DELTA,
-		Compression:     datasetmd.COMPRESSION_TYPE_NONE,
-		StoreRangeStats: true,
+		PageSizeHint: pageSize,
+		Value:        datasetmd.VALUE_TYPE_INT64,
+		Encoding:     datasetmd.ENCODING_TYPE_DELTA,
+		Compression:  datasetmd.COMPRESSION_TYPE_NONE,
+		Statistics: dataset.StatisticsOptions{
+			StoreRangeStats: true,
+		},
 	})
 	if err != nil {
 		// We control the Value/Encoding tuple so this can't fail; if it does,
@@ -177,7 +181,10 @@ func (b *tableBuffer) Metadata(key string, pageSize int, compressionOpts dataset
 		Encoding:           datasetmd.ENCODING_TYPE_PLAIN,
 		Compression:        datasetmd.COMPRESSION_TYPE_ZSTD,
 		CompressionOptions: compressionOpts,
-		StoreRangeStats:    true,
+		Statistics: dataset.StatisticsOptions{
+			StoreRangeStats:       true,
+			StoreCardinalityStats: true,
+		},
 	})
 	if err != nil {
 		// We control the Value/Encoding tuple so this can't fail; if it does,
@@ -213,7 +220,9 @@ func (b *tableBuffer) Message(pageSize int, compressionOpts dataset.CompressionO
 		//
 		// A "min log line" and "max log line" isn't very valuable, and since log
 		// lines can be quite long, it would consume a fair amount of metadata.
-		StoreRangeStats: false,
+		Statistics: dataset.StatisticsOptions{
+			StoreRangeStats: false,
+		},
 	})
 	if err != nil {
 		// We control the Value/Encoding tuple so this can't fail; if it does,
diff --git a/pkg/dataobj/internal/sections/streams/streams.go b/pkg/dataobj/internal/sections/streams/streams.go
index 4d208fb1676d5..05002fb77e345 100644
--- a/pkg/dataobj/internal/sections/streams/streams.go
+++ b/pkg/dataobj/internal/sections/streams/streams.go
@@ -251,11 +251,13 @@ func (s *Streams) EncodeTo(enc *encoding.Encoder) error {
 		}
 
 		builder, err := dataset.NewColumnBuilder(name, dataset.BuilderOptions{
-			PageSizeHint:    s.pageSize,
-			Value:           datasetmd.VALUE_TYPE_STRING,
-			Encoding:        datasetmd.ENCODING_TYPE_PLAIN,
-			Compression:     datasetmd.COMPRESSION_TYPE_ZSTD,
-			StoreRangeStats: true,
+			PageSizeHint: s.pageSize,
+			Value:        datasetmd.VALUE_TYPE_STRING,
+			Encoding:     datasetmd.ENCODING_TYPE_PLAIN,
+			Compression:  datasetmd.COMPRESSION_TYPE_ZSTD,
+			Statistics: dataset.StatisticsOptions{
+				StoreRangeStats: true,
+			},
 		})
 		if err != nil {
 			return nil, fmt.Errorf("creating label column: %w", err)
@@ -323,11 +325,13 @@ func (s *Streams) EncodeTo(enc *encoding.Encoder) error {
 
 func numberColumnBuilder(pageSize int) (*dataset.ColumnBuilder, error) {
 	return dataset.NewColumnBuilder("", dataset.BuilderOptions{
-		PageSizeHint:    pageSize,
-		Value:           datasetmd.VALUE_TYPE_INT64,
-		Encoding:        datasetmd.ENCODING_TYPE_DELTA,
-		Compression:     datasetmd.COMPRESSION_TYPE_NONE,
-		StoreRangeStats: true,
+		PageSizeHint: pageSize,
+		Value:        datasetmd.VALUE_TYPE_INT64,
+		Encoding:     datasetmd.ENCODING_TYPE_DELTA,
+		Compression:  datasetmd.COMPRESSION_TYPE_NONE,
+		Statistics: dataset.StatisticsOptions{
+			StoreRangeStats: true,
+		},
 	})
 }