diff --git a/examples/io_odbc.rs b/examples/io_odbc.rs
index 996fa7130c6..9305fab6e24 100644
--- a/examples/io_odbc.rs
+++ b/examples/io_odbc.rs
@@ -1,4 +1,4 @@
-//! Example showing how to write to, and read from, an ODBC connector
+//! Demo of how to write to, and read from, an ODBC connector
 //!
 //! On an Ubuntu, you need to run the following (to install the driver):
 //! ```bash
@@ -28,13 +28,14 @@ fn main() -> Result<()> {
     let query = "INSERT INTO example (c1, c2) VALUES (?, ?)";
     let prepared = connection.prepare(query).unwrap();
 
-    // first, initialize buffers from odbc-api
+    // secondly, we initialize buffers from odbc-api
     let fields = vec![
         // (for now) the types here must match the tables' schema
         Field::new("unused", DataType::Int32, true),
         Field::new("unused", DataType::LargeUtf8, true),
     ];
 
+    // third, we initialize the writer
     let mut writer = write::Writer::try_new(prepared, fields)?;
 
     // say we have (or receive from a channel) a chunk:
diff --git a/guide/src/io/README.md b/guide/src/io/README.md
index f8e8bb8ca64..5a8cd477949 100644
--- a/guide/src/io/README.md
+++ b/guide/src/io/README.md
@@ -7,5 +7,6 @@ This crate offers optional features that enable interoperability with different
 * Parquet (`io_parquet`)
 * JSON and NDJSON (`io_json`)
 * Avro (`io_avro` and `io_avro_async`)
+* ODBC-compliant databases (`io_odbc`)
 
 In this section you can find a guide and examples for each one of them.
diff --git a/guide/src/io/odbc.md b/guide/src/io/odbc.md
new file mode 100644
index 00000000000..7e362daf7c6
--- /dev/null
+++ b/guide/src/io/odbc.md
@@ -0,0 +1,8 @@
+# ODBC
+
+When compiled with feature `io_odbc`, this crate can be used to read from, and write to
+any [ODBC](https://en.wikipedia.org/wiki/Open_Database_Connectivity) interface:
+
+```rust
+{{#include ../../../examples/odbc.rs}}
+```
diff --git a/src/io/odbc/write/mod.rs b/src/io/odbc/write/mod.rs
index 18d2ebdbdaf..245f2455bb8 100644
--- a/src/io/odbc/write/mod.rs
+++ b/src/io/odbc/write/mod.rs
@@ -9,6 +9,8 @@ pub use schema::infer_descriptions;
 pub use serialize::serialize;
 
 /// Creates a [`api::buffers::ColumnarBuffer`] from [`api::ColumnDescription`]s.
+///
+/// This is useful when separating the serialization (CPU-bounded) to writing to the DB (IO-bounded).
 pub fn buffer_from_description(
     descriptions: Vec<api::ColumnDescription>,
     capacity: usize,
@@ -23,7 +25,10 @@ pub fn buffer_from_description(
     api::buffers::buffer_from_description(capacity, descs)
 }
 
-/// A writer of [`Chunk`] to an ODBC prepared statement.
+/// A writer of [`Chunk`]s to an ODBC [`api::Prepared`] statement.
+/// # Implementation
+/// This struct mixes CPU-bounded and IO-bounded tasks and is not ideal
+/// for an `async` context.
 pub struct Writer<'a> {
     fields: Vec<Field>,
     buffer: api::buffers::ColumnarBuffer<api::buffers::AnyColumnBuffer>,
@@ -31,7 +36,9 @@ pub struct Writer<'a> {
 }
 
 impl<'a> Writer<'a> {
-    /// Creates a new [`Writer`]
+    /// Creates a new [`Writer`].
+    /// # Errors
+    /// Errors iff any of the types from [`Field`] is not supported.
     pub fn try_new(prepared: api::Prepared<'a>, fields: Vec<Field>) -> Result<Self> {
         let buffer = buffer_from_description(infer_descriptions(&fields)?, 0);
         Ok(Self {
@@ -41,7 +48,9 @@ impl<'a> Writer<'a> {
         })
     }
 
-    /// Writes a chunk to the writter.
+    /// Writes a chunk to the writer.
+    /// # Errors
+    /// Errors iff the execution of the statement fails.
     pub fn write<A: AsRef<dyn Array>>(&mut self, chunk: &Chunk<A>) -> Result<()> {
         if chunk.len() > self.buffer.num_rows() {
             // if the chunk is larger, we re-allocate new buffers to hold it
diff --git a/src/io/odbc/write/schema.rs b/src/io/odbc/write/schema.rs
index 546c6d2b427..9e4b61f704e 100644
--- a/src/io/odbc/write/schema.rs
+++ b/src/io/odbc/write/schema.rs
@@ -3,7 +3,7 @@ use super::super::api;
 use crate::datatypes::{DataType, Field};
 use crate::error::{ArrowError, Result};
 
-/// Infers the [`ColumnDescription`] from the fields
+/// Infers the [`api::ColumnDescription`] from the fields
 pub fn infer_descriptions(fields: &[Field]) -> Result<Vec<api::ColumnDescription>> {
     fields
         .iter()
diff --git a/src/io/odbc/write/serialize.rs b/src/io/odbc/write/serialize.rs
index f3a68fc1de2..3128ceb964b 100644
--- a/src/io/odbc/write/serialize.rs
+++ b/src/io/odbc/write/serialize.rs
@@ -121,12 +121,13 @@ fn bool(array: &BooleanArray, values: &mut [api::Bit]) {
 }
 
 fn bool_optional(array: &BooleanArray, values: &mut NullableSliceMut<api::Bit>) {
+    let (values, indicators) = values.raw_values();
     array
         .values()
         .iter()
-        .zip(values.values().iter_mut())
+        .zip(values.iter_mut())
         .for_each(|(from, to)| *to = api::Bit(from as u8));
-    write_validity(array.validity(), values.indicators());
+    write_validity(array.validity(), indicators);
 }
 
 fn primitive<T: NativeType>(array: &PrimitiveArray<T>, values: &mut [T]) {
@@ -145,8 +146,9 @@ fn write_validity(validity: Option<&Bitmap>, indicators: &mut [isize]) {
 }
 
 fn primitive_optional<T: NativeType>(array: &PrimitiveArray<T>, values: &mut NullableSliceMut<T>) {
-    values.values().copy_from_slice(array.values());
-    write_validity(array.validity(), values.indicators());
+    let (values, indicators) = values.raw_values();
+    values.copy_from_slice(array.values());
+    write_validity(array.validity(), indicators);
 }
 
 fn fixed_binary(array: &FixedSizeBinaryArray, writer: &mut BinColumnWriter) {