diff --git a/NEWS.md b/NEWS.md
index 397b6711..08c1d82c 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -10,6 +10,8 @@
   ISO 8061 convention where positive offset denotes time zone east of 
   Greenwich. (#946)
 
+* `varchar()` helper now returns number of bytes not characters (#960)
+
 # odbc 1.6.3
 
 * Addressed a compiler warning on `r-devel-linux-x86_64-fedora-clang` (#941).
diff --git a/R/aaa-odbc-data-type.R b/R/aaa-odbc-data-type.R
index 9d46fa2f..9018d17d 100644
--- a/R/aaa-odbc-data-type.R
+++ b/R/aaa-odbc-data-type.R
@@ -118,7 +118,7 @@ is_blob <- function(obj) {
 
 varchar <- function(x, type = "varchar") {
   # at least 255 characters, use max if more than 8000:
-  max_length <- max(c(255, nchar(as.character(x))), na.rm = TRUE)
+  max_length <- max(c(255, nchar(as.character(x), type = "bytes")), na.rm = TRUE)
 
   if (max_length > 8000) {
     max_length <- "max"
diff --git a/tests/testthat/test-utils.R b/tests/testthat/test-utils.R
index 7827298f..80c24c51 100644
--- a/tests/testthat/test-utils.R
+++ b/tests/testthat/test-utils.R
@@ -383,3 +383,9 @@ test_that("configure_unixodbc_simba() writes reasonable entries", {
     action = "warn"
   ))
 })
+
+test_that("varchar() uses byte length (avoids truncation with multibyte)", {
+  # U+2019 RIGHT SINGLE QUOTATION MARK in UTF-8 is 3 bytes
+  string <- paste0("\xe2\x80\x99", paste(rep("a", 255), collapse = ""))
+  expect_equal(varchar(string, type = "varchar"), "varchar(258)")
+})