diff --git a/tests/testthat/_snaps/constructor.md b/tests/testthat/_snaps/constructor.md
index 224bee13..eb98eff3 100644
--- a/tests/testthat/_snaps/constructor.md
+++ b/tests/testthat/_snaps/constructor.md
@@ -29,3 +29,27 @@
       Error in `new_model()`:
       ! `blueprint` must be a <hardhat_blueprint>, not the number 1.
 
+# `new_scalar()` must have elements
+
+    Code
+      new_scalar(list())
+    Condition
+      Error in `check_elems()`:
+      ! `elems` must be a list of length 1 or greater.
+
+# `new_scalar()` must have unique names
+
+    Code
+      new_scalar(list(x = 1, x = 2))
+    Condition
+      Error in `check_elems()`:
+      ! `elems` must have unique names.
+
+# `new_scalar()` must have no extra attributes
+
+    Code
+      new_scalar(x)
+    Condition
+      Error in `check_elems()`:
+      ! `elems` must have no attributes (apart from names).
+
diff --git a/tests/testthat/_snaps/forge-formula.md b/tests/testthat/_snaps/forge-formula.md
index b015313a..695a321d 100644
--- a/tests/testthat/_snaps/forge-formula.md
+++ b/tests/testthat/_snaps/forge-formula.md
@@ -1,3 +1,67 @@
+# asking for the outcome when it isn't there fails
+
+    Code
+      forge(example_train2, x1$blueprint, outcomes = TRUE)
+    Condition
+      Error in `validate_column_names()`:
+      ! The following required columns are missing: 'fac_1'.
+
+---
+
+    Code
+      forge(example_train2, x2$blueprint, outcomes = TRUE)
+    Condition
+      Error in `validate_column_names()`:
+      ! The following required columns are missing: 'fac_1'.
+
+# new_data can only be a data frame / matrix
+
+    Code
+      forge("hi", x1$blueprint)
+    Condition
+      Error in `forge()`:
+      ! The class of `new_data`, 'character', is not recognized.
+
+---
+
+    Code
+      forge("hi", x2$blueprint)
+    Condition
+      Error in `forge()`:
+      ! The class of `new_data`, 'character', is not recognized.
+
+# missing predictor columns fail appropriately
+
+    Code
+      forge(example_train[, 1, drop = FALSE], x1$blueprint)
+    Condition
+      Error in `validate_column_names()`:
+      ! The following required columns are missing: 'num_2'.
+
+---
+
+    Code
+      forge(example_train[, 1, drop = FALSE], x2$blueprint)
+    Condition
+      Error in `validate_column_names()`:
+      ! The following required columns are missing: 'num_2'.
+
+---
+
+    Code
+      forge(example_train[, 3, drop = FALSE], x1$blueprint)
+    Condition
+      Error in `validate_column_names()`:
+      ! The following required columns are missing: 'num_1', 'num_2'.
+
+---
+
+    Code
+      forge(example_train[, 3, drop = FALSE], x2$blueprint)
+    Condition
+      Error in `validate_column_names()`:
+      ! The following required columns are missing: 'num_1', 'num_2'.
+
 # novel predictor levels are caught
 
     Code
diff --git a/tests/testthat/_snaps/forge-recipe.md b/tests/testthat/_snaps/forge-recipe.md
index d7230cfa..fde9c1ca 100644
--- a/tests/testthat/_snaps/forge-recipe.md
+++ b/tests/testthat/_snaps/forge-recipe.md
@@ -1,3 +1,35 @@
+# asking for the outcome when it isn't there fails
+
+    Code
+      forge(iris2, x1$blueprint, outcomes = TRUE)
+    Condition
+      Error in `validate_column_names()`:
+      ! The following required columns are missing: 'Species'.
+
+---
+
+    Code
+      forge(iris2, x2$blueprint, outcomes = TRUE)
+    Condition
+      Error in `validate_column_names()`:
+      ! The following required columns are missing: 'Species'.
+
+# missing predictor columns fail appropriately
+
+    Code
+      forge(iris[, 1, drop = FALSE], x$blueprint)
+    Condition
+      Error in `validate_column_names()`:
+      ! The following required columns are missing: 'Sepal.Width'.
+
+---
+
+    Code
+      forge(iris[, 3, drop = FALSE], x$blueprint)
+    Condition
+      Error in `validate_column_names()`:
+      ! The following required columns are missing: 'Sepal.Length', 'Sepal.Width'.
+
 # novel predictor levels are caught
 
     Code
diff --git a/tests/testthat/_snaps/forge-xy.md b/tests/testthat/_snaps/forge-xy.md
index a5a14f83..1a6ed11a 100644
--- a/tests/testthat/_snaps/forge-xy.md
+++ b/tests/testthat/_snaps/forge-xy.md
@@ -1,3 +1,79 @@
+# asking for the outcome is special cased for vector `y` values
+
+    Code
+      forge(iris, x1$blueprint, outcomes = TRUE)
+    Condition
+      Error in `validate_missing_name_isnt_.outcome()`:
+      ! The following required columns are missing: '.outcome'.
+      
+      (This indicates that `mold()` was called with a vector for `y`. When this is the case, and the outcome columns are requested in `forge()`, `new_data` must include a column with the automatically generated name, '.outcome', containing the outcome.)
+
+---
+
+    Code
+      forge(iris, x2$blueprint, outcomes = TRUE)
+    Condition
+      Error in `validate_missing_name_isnt_.outcome()`:
+      ! The following required columns are missing: '.outcome'.
+      
+      (This indicates that `mold()` was called with a vector for `y`. When this is the case, and the outcome columns are requested in `forge()`, `new_data` must include a column with the automatically generated name, '.outcome', containing the outcome.)
+
+# new_data can only be a data frame / matrix
+
+    Code
+      forge("hi", x1$blueprint)
+    Condition
+      Error in `forge()`:
+      ! The class of `new_data`, 'character', is not recognized.
+
+---
+
+    Code
+      forge("hi", x2$blueprint)
+    Condition
+      Error in `forge()`:
+      ! The class of `new_data`, 'character', is not recognized.
+
+---
+
+    Code
+      forge("hi", x3$blueprint)
+    Condition
+      Error in `forge()`:
+      ! The class of `new_data`, 'character', is not recognized.
+
+# missing predictor columns fail appropriately
+
+    Code
+      forge(iris[, 1, drop = FALSE], x1$blueprint)
+    Condition
+      Error in `validate_column_names()`:
+      ! The following required columns are missing: 'Sepal.Width'.
+
+---
+
+    Code
+      forge(iris[, 1, drop = FALSE], x2$blueprint)
+    Condition
+      Error in `validate_column_names()`:
+      ! The following required columns are missing: 'Sepal.Width'.
+
+---
+
+    Code
+      forge(iris[, 3, drop = FALSE], x1$blueprint)
+    Condition
+      Error in `validate_column_names()`:
+      ! The following required columns are missing: 'Sepal.Length', 'Sepal.Width'.
+
+---
+
+    Code
+      forge(iris[, 3, drop = FALSE], x2$blueprint)
+    Condition
+      Error in `validate_column_names()`:
+      ! The following required columns are missing: 'Sepal.Length', 'Sepal.Width'.
+
 # novel predictor levels are caught
 
     Code
diff --git a/tests/testthat/_snaps/levels.md b/tests/testthat/_snaps/levels.md
new file mode 100644
index 00000000..e8207e5e
--- /dev/null
+++ b/tests/testthat/_snaps/levels.md
@@ -0,0 +1,8 @@
+# Can extract levels from an outcome
+
+    Code
+      get_outcome_levels("a")
+    Condition
+      Error in `standardize()`:
+      ! `y` is of unknown type 'character'.
+
diff --git a/tests/testthat/_snaps/model-offset.md b/tests/testthat/_snaps/model-offset.md
new file mode 100644
index 00000000..cd93def8
--- /dev/null
+++ b/tests/testthat/_snaps/model-offset.md
@@ -0,0 +1,16 @@
+# Only numeric columns can be offsets
+
+    Code
+      mold(~ Sepal.Width + offset(Species), iris)
+    Condition
+      Error in `model_offset()`:
+      ! Column, 'offset(Species)', is tagged as an offset, but is not numeric. All offsets must be numeric.
+
+# offset columns are stored as predictors
+
+    Code
+      forge(iris2, x$blueprint)
+    Condition
+      Error in `validate_column_names()`:
+      ! The following required columns are missing: 'Sepal.Length'.
+
diff --git a/tests/testthat/_snaps/mold-formula.md b/tests/testthat/_snaps/mold-formula.md
index b2c96456..4dd7d4f3 100644
--- a/tests/testthat/_snaps/mold-formula.md
+++ b/tests/testthat/_snaps/mold-formula.md
@@ -121,6 +121,79 @@
       ! Functions involving factors or characters have been detected on the RHS of `formula`. These are not allowed when `indicators = "none"`.
       i Functions involving factors were detected for "fac_1" in `paste0(fac_1)`.
 
+# formulas with non-existent columns are caught
+
+    Code
+      mold(fac_1 ~ y + z, example_train)
+    Condition
+      Error in `get_all_predictors()`:
+      ! The following predictors were not found in `data`: 'y', 'z'.
+
+---
+
+    Code
+      mold(fac_1 ~ y + z, example_train, blueprint = bp)
+    Condition
+      Error in `get_all_predictors()`:
+      ! The following predictors were not found in `data`: 'y', 'z'.
+
+---
+
+    Code
+      mold(y + z ~ fac_1, example_train)
+    Condition
+      Error in `get_all_outcomes()`:
+      ! The following outcomes were not found in `data`: 'y', 'z'.
+
+---
+
+    Code
+      mold(y + z ~ fac_1, example_train, blueprint = bp)
+    Condition
+      Error in `get_all_outcomes()`:
+      ! The following outcomes were not found in `data`: 'y', 'z'.
+
+# global environment variables cannot be used
+
+    Code
+      y <- 1
+      mold(fac_1 ~ y, example_train)
+    Condition
+      Error in `get_all_predictors()`:
+      ! The following predictors were not found in `data`: 'y'.
+
+# cannot manually remove intercept in the formula itself
+
+    Code
+      mold(fac_1 ~ y + 0, example_train)
+    Condition
+      Error in `mold_formula_default_clean()`:
+      ! `formula` must not contain the intercept removal term: `+ 0` or `0 +`.
+
+---
+
+    Code
+      mold(fac_1 ~ y + 0, example_train, blueprint = bp)
+    Condition
+      Error in `mold_formula_default_clean()`:
+      ! `formula` must not contain the intercept removal term: `+ 0` or `0 +`.
+
+---
+
+    Code
+      mold(fac_1 ~ 0 + y, example_train)
+    Condition
+      Error in `mold_formula_default_clean()`:
+      ! `formula` must not contain the intercept removal term: `+ 0` or `0 +`.
+
+---
+
+    Code
+      mold(fac_1 ~ y - 1, example_train)
+    Condition
+      Error in `mold_formula_default_clean()`:
+      ! `formula` must not contain the intercept removal term: `- 1`.
+
 # RHS with _only_ intercept related terms are caught
 
     Code
@@ -169,6 +242,22 @@
       Error in `mold_formula_default_clean()`:
       ! `formula` must not contain the intercept removal term, `0`.
 
+# `data` is validated
+
+    Code
+      mold(fac_1 ~ num_2, 1)
+    Condition
+      Error in `mold_formula_default_clean()`:
+      ! `data` must be a data frame or a matrix, not the number 1.
+
+---
+
+    Code
+      mold(fac_1 ~ num_2, 1, blueprint = bp)
+    Condition
+      Error in `mold_formula_default_clean()`:
+      ! `data` must be a data frame or a matrix, not the number 1.
+
 # LHS of the formula cannot contain interactions
 
     Code
@@ -223,6 +312,22 @@
       ! Interaction terms can't be specified on the LHS of `formula`.
       i The following interaction term was found: `num_1/num_2`.
 
+# `.` notation fails on the LHS
+
+    Code
+      mold(. ~ fac_1, example_train)
+    Condition
+      Error in `get_all_outcomes()`:
+      ! The left hand side of the formula cannot contain `.`
+
+---
+
+    Code
+      mold(. ~ fac_1, example_train, blueprint = bp)
+    Condition
+      Error in `get_all_outcomes()`:
+      ! The left hand side of the formula cannot contain `.`
+
 # `blueprint` is validated
 
     Code
diff --git a/tests/testthat/_snaps/mold-recipe.md b/tests/testthat/_snaps/mold-recipe.md
new file mode 100644
index 00000000..28674f15
--- /dev/null
+++ b/tests/testthat/_snaps/mold-recipe.md
@@ -0,0 +1,8 @@
+# `data` is validated
+
+    Code
+      mold(recipes::recipe(Species ~ Sepal.Length, data = iris), 1)
+    Condition
+      Error in `mold_recipe_default_clean()`:
+      ! `data` must be a data frame or a matrix, not the number 1.
+
diff --git a/tests/testthat/_snaps/spruce.md b/tests/testthat/_snaps/spruce.md
index 2a345c45..cb6cc552 100644
--- a/tests/testthat/_snaps/spruce.md
+++ b/tests/testthat/_snaps/spruce.md
@@ -1,3 +1,77 @@
+# spruce - numeric
+
+    Code
+      spruce_numeric("hi")
+    Condition
+      Error in `spruce_numeric()`:
+      ! `pred` must be a numeric vector, not the string "hi".
+
+---
+
+    Code
+      spruce_numeric(matrix(1))
+    Condition
+      Error in `spruce_numeric()`:
+      ! `pred` must be a numeric vector, not a double matrix.
+
+# spruce - class
+
+    Code
+      spruce_class(1)
+    Condition
+      Error in `spruce_class()`:
+      ! `pred_class` must be a factor, not the number 1.
+
+---
+
+    Code
+      spruce_class("hi")
+    Condition
+      Error in `spruce_class()`:
+      ! `pred_class` must be a factor, not the string "hi".
+
+# spruce - prob
+
+    Code
+      spruce_prob(1, prob_matrix)
+    Condition
+      Error in `spruce_prob()`:
+      ! `pred_levels` must be a character vector, not the number 1.
+
+---
+
+    Code
+      spruce_prob(pred_levels, 1)
+    Condition
+      Error in `spruce_prob()`:
+      ! `prob_matrix` must be a numeric matrix, not the number 1.
+
+---
+
+    Code
+      spruce_prob("a", matrix("a"))
+    Condition
+      Error in `spruce_prob()`:
+      ! `prob_matrix` must be a numeric matrix, not a character matrix.
+
+---
+
+    Code
+      spruce_prob(c("a", "b"), matrix(1, ncol = 3))
+    Condition
+      Error in `spruce_prob()`:
+      ! The number of levels (2) must be
+      equal to the number of class probability columns (3).
+
+---
+
+    Code
+      spruce_prob(c("a"), matrix(1, ncol = 2))
+    Condition
+      Error in `spruce_prob()`:
+      ! The number of levels (1) must be
+      equal to the number of class probability columns (2).
+
 # spruce multiple helpers check input type
 
     Code
diff --git a/tests/testthat/_snaps/standardize.md b/tests/testthat/_snaps/standardize.md
index b1021630..23967906 100644
--- a/tests/testthat/_snaps/standardize.md
+++ b/tests/testthat/_snaps/standardize.md
@@ -38,3 +38,27 @@
       Error in `standardize()`:
       ! All columns of `y` must have unique names.
 
+---
+
+    Code
+      standardize(bad2)
+    Condition
+      Error in `validate_has_known_outcome_types()`:
+      ! Not all columns of `y` are known outcome types. These columns have unknown types: 'x'.
+
+# standardize - unknown
+
+    Code
+      standardize("hi")
+    Condition
+      Error in `standardize()`:
+      ! `y` is of unknown type 'character'.
+
+---
+
+    Code
+      standardize(Sys.time())
+    Condition
+      Error in `standardize()`:
+      ! `y` is of unknown type 'POSIXct'.
+
diff --git a/tests/testthat/_snaps/tune.md b/tests/testthat/_snaps/tune.md
new file mode 100644
index 00000000..66df0ba6
--- /dev/null
+++ b/tests/testthat/_snaps/tune.md
@@ -0,0 +1,24 @@
+# `id` is validated
+
+    Code
+      tune(1)
+    Condition
+      Error in `tune()`:
+      ! The `id` should be a single character string.
+
+---
+
+    Code
+      tune(c("x", "y"))
+    Condition
+      Error in `tune()`:
+      ! The `id` should be a single character string.
+
+---
+
+    Code
+      tune(NA_character_)
+    Condition
+      Error in `tune()`:
+      ! The `id` can't be missing.
+
diff --git a/tests/testthat/_snaps/validation.md b/tests/testthat/_snaps/validation.md
new file mode 100644
index 00000000..4c397ab9
--- /dev/null
+++ b/tests/testthat/_snaps/validation.md
@@ -0,0 +1,117 @@
+# validate_outcomes_are_univariate()
+
+    Code
+      validate_outcomes_are_univariate(iris)
+    Condition
+      Error in `validate_outcomes_are_univariate()`:
+      ! The outcome must be univariate, but 5 columns were found.
+
+# validate_outcomes_are_numeric()
+
+    Code
+      validate_outcomes_are_numeric(iris)
+    Condition
+      Error in `validate_outcomes_are_numeric()`:
+      ! All outcomes must be numeric, but the following are not:
+      'Species': 'factor'
+
+---
+
+    Code
+      validate_outcomes_are_numeric(x)
+    Condition
+      Error in `validate_outcomes_are_numeric()`:
+      ! All outcomes must be numeric, but the following are not:
+      'x': 'POSIXct', 'POSIXt'
+      'y': 'factor'
+
+# validate_no_formula_duplication()
+
+    Code
+      validate_no_formula_duplication(y ~ y)
+    Condition
+      Error in `validate_no_formula_duplication()`:
+      ! The following terms are duplicated on the left and right hand side of the `formula`: 'y'.
+
+---
+
+    Code
+      validate_no_formula_duplication(y ~ log(y), original = TRUE)
+    Condition
+      Error in `validate_no_formula_duplication()`:
+      ! The following terms are duplicated on the left and right hand side of the `formula`: 'y'.
+
+---
+
+    Code
+      validate_no_formula_duplication(y + x ~ y + x)
+    Condition
+      Error in `validate_no_formula_duplication()`:
+      ! The following terms are duplicated on the left and right hand side of the `formula`: 'y', 'x'.
+
+---
+
+    Code
+      validate_no_formula_duplication(y ~ . + y)
+    Condition
+      Error in `validate_no_formula_duplication()`:
+      ! The following terms are duplicated on the left and right hand side of the `formula`: 'y'.
+
+---
+
+    Code
+      validate_no_formula_duplication(y ~ offset(y), original = TRUE)
+    Condition
+      Error in `validate_no_formula_duplication()`:
+      ! The following terms are duplicated on the left and right hand side of the `formula`: 'y'.
+
+# validate_outcomes_are_factors()
+
+    Code
+      validate_outcomes_are_factors(x)
+    Condition
+      Error in `validate_outcomes_are_factors()`:
+      ! All outcomes must be factors, but the following are not:
+      'x': 'POSIXct', 'POSIXt'
+      'y': 'character'
+
+# validate_outcomes_are_binary()
+
+    Code
+      validate_outcomes_are_binary(iris)
+    Condition
+      Error in `validate_outcomes_are_binary()`:
+      ! The outcome must be binary, but the following number of levels were found:
+      'Sepal.Length': 0
+      'Sepal.Width': 0
+      'Petal.Length': 0
+      'Petal.Width': 0
+      'Species': 3
+
+# validate_predictors_are_numeric()
+
+    Code
+      validate_predictors_are_numeric(iris)
+    Condition
+      Error in `validate_predictors_are_numeric()`:
+      ! All predictors must be numeric, but the following are not:
+      'Species': 'factor'
+
+---
+
+    Code
+      validate_predictors_are_numeric(x)
+    Condition
+      Error in `validate_predictors_are_numeric()`:
+      ! All predictors must be numeric, but the following are not:
+      'x': 'POSIXct', 'POSIXt'
+      'y': 'factor'
+
+# validate_prediction_size()
+
+    Code
+      validate_prediction_size(mtcars[1:5, ], mtcars)
+    Condition
+      Error in `validate_prediction_size()`:
+      ! The size of `new_data` (32) must match the size of `pred` (5).
+
diff --git a/tests/testthat/test-constructor.R b/tests/testthat/test-constructor.R
index e1f0bcf7..bad5d5cc 100644
--- a/tests/testthat/test-constructor.R
+++ b/tests/testthat/test-constructor.R
@@ -32,9 +32,8 @@ test_that("can have custom elements", {
 })
 
 test_that("must use a valid blueprint", {
-  expect_error(
-    new_model(blueprint = default_xy_blueprint(), class = "custom"),
-    NA
+  expect_no_error(
+    new_model(blueprint = default_xy_blueprint(), class = "custom")
   )
 
   expect_snapshot(error = TRUE, {
@@ -43,15 +42,15 @@ test_that("must use a valid blueprint", {
 })
 
 test_that("`new_scalar()` must have elements", {
-  expect_error(new_scalar(list()), "must be a list of length 1 or greater")
+  expect_snapshot(error = TRUE, new_scalar(list()))
 })
 
 test_that("`new_scalar()` must have unique names", {
-  expect_error(new_scalar(list(x = 1, x = 2)), "must have unique names")
+  expect_snapshot(error = TRUE, new_scalar(list(x = 1, x = 2)))
 })
 
 test_that("`new_scalar()` must have no extra attributes", {
   x <- list(x = 1)
   attr(x, "extra") <- 1
-  expect_error(new_scalar(x), "must have no attributes")
+  expect_snapshot(error = TRUE, new_scalar(x))
 })
diff --git a/tests/testthat/test-forge-formula.R b/tests/testthat/test-forge-formula.R
index 93fb59ef..dfeebff0 100644
--- a/tests/testthat/test-forge-formula.R
+++ b/tests/testthat/test-forge-formula.R
@@ -131,13 +131,13 @@ test_that("asking for the outcome when it isn't there fails", {
   example_train2 <- example_train
   example_train2$fac_1 <- NULL
 
-  expect_error(
-    forge(example_train2, x1$blueprint, outcomes = TRUE),
-    "The following required columns"
+  expect_snapshot(
+    error = TRUE,
+    forge(example_train2, x1$blueprint, outcomes = TRUE)
   )
-  expect_error(
-    forge(example_train2, x2$blueprint, outcomes = TRUE),
-    "The following required columns"
+  expect_snapshot(
+    error = TRUE,
+    forge(example_train2, x2$blueprint, outcomes = TRUE)
   )
 })
 
@@ -178,14 +178,8 @@ test_that("new_data can be a matrix", {
   x2 <- mold(fac_1 ~ num_1, example_train, blueprint = bp)
   example_train_mat <- as.matrix(example_train[, "num_1", drop = FALSE])
 
-  expect_error(
-    xx1 <- forge(example_train_mat, x1$blueprint),
-    NA
-  )
-  expect_error(
-    xx2 <- forge(example_train_mat, x2$blueprint),
-    NA
-  )
+  expect_no_error(xx1 <- forge(example_train_mat, x1$blueprint))
+  expect_no_error(xx2 <- forge(example_train_mat, x2$blueprint))
 
   sep_len <- example_train$num_1
 
@@ -204,13 +198,13 @@ test_that("new_data can only be a data frame / matrix", {
   x1 <- mold(fac_1 ~ num_1, example_train)
   x2 <- mold(fac_1 ~ num_1, example_train, blueprint = bp)
 
-  expect_error(
-    forge("hi", x1$blueprint),
-    "The class of `new_data`, 'character'"
+  expect_snapshot(
+    error = TRUE,
+    forge("hi", x1$blueprint)
   )
-  expect_error(
-    forge("hi", x2$blueprint),
-    "The class of `new_data`, 'character'"
+  expect_snapshot(
+    error = TRUE,
+    forge("hi", x2$blueprint)
   )
 })
 
@@ -219,22 +213,22 @@ test_that("missing predictor columns fail appropriately", {
   x1 <- mold(fac_1 ~ num_1 + num_2, example_train)
   x2 <- mold(fac_1 ~ num_1 + num_2, example_train, blueprint = bp)
 
-  expect_error(
-    forge(example_train[, 1, drop = FALSE], x1$blueprint),
-    "num_2"
+  expect_snapshot(
+    error = TRUE,
+    forge(example_train[, 1, drop = FALSE], x1$blueprint)
   )
-  expect_error(
-    forge(example_train[, 1, drop = FALSE], x2$blueprint),
-    "num_2"
+  expect_snapshot(
+    error = TRUE,
+    forge(example_train[, 1, drop = FALSE], x2$blueprint)
   )
 
-  expect_error(
-    forge(example_train[, 3, drop = FALSE], x1$blueprint),
-    "'num_1', 'num_2'"
+  expect_snapshot(
+    error = TRUE,
+    forge(example_train[, 3, drop = FALSE], x1$blueprint)
   )
-  expect_error(
-    forge(example_train[, 3, drop = FALSE], x2$blueprint),
-    "'num_1', 'num_2'"
+  expect_snapshot(
+    error = TRUE,
+    forge(example_train[, 3, drop = FALSE], x2$blueprint)
   )
 })
 
@@ -770,9 +764,8 @@ test_that("new data classes are caught", {
   x <- mold(num_1 ~ fac_1, example_train, blueprint = default_formula_blueprint(indicators = "none"))
 
   # Silently recover character -> factor
-  expect_error(
-    x_example_train2 <- forge(example_train2, x$blueprint),
-    NA
+  expect_no_error(
+    x_example_train2 <- forge(example_train2, x$blueprint)
   )
 
   expect_s3_class(
@@ -782,9 +775,8 @@ test_that("new data classes are caught", {
 
   xx <- mold(fac_1 ~ num_1, example_train)
 
-  expect_error(
-    xx_example_train2 <- forge(example_train2, xx$blueprint, outcomes = TRUE),
-    NA
+  expect_no_error(
+    xx_example_train2 <- forge(example_train2, xx$blueprint, outcomes = TRUE)
   )
 
   expect_s3_class(
@@ -801,25 +793,17 @@ test_that("new data classes can interchange integer/numeric", {
   x1 <- mold(fac_1 ~ num_1, example_train)
   x2 <- mold(fac_1 ~ num_1, example_train, blueprint = bp)
 
-  expect_error(
-    forge(example_train2, x1$blueprint),
-    NA
-  )
-  expect_error(
-    forge(example_train2, x2$blueprint),
-    NA
-  )
+  expect_no_error(forge(example_train2, x1$blueprint))
+  expect_no_error(forge(example_train2, x2$blueprint))
 
   x3 <- mold(num_1 ~ fac_1, example_train)
   x4 <- mold(num_1 ~ fac_1, example_train, blueprint = bp)
 
-  expect_error(
-    forge(example_train2, x3$blueprint, outcomes = TRUE),
-    NA
+  expect_no_error(
+    forge(example_train2, x3$blueprint, outcomes = TRUE)
   )
-  expect_error(
-    forge(example_train2, x4$blueprint, outcomes = TRUE),
-    NA
+  expect_no_error(
+    forge(example_train2, x4$blueprint, outcomes = TRUE)
   )
 })
 
diff --git a/tests/testthat/test-forge-recipe.R b/tests/testthat/test-forge-recipe.R
index aa16eead..66e23998 100644
--- a/tests/testthat/test-forge-recipe.R
+++ b/tests/testthat/test-forge-recipe.R
@@ -98,14 +98,14 @@ test_that("asking for the outcome when it isn't there fails", {
   iris2 <- iris
   iris2$Species <- NULL
 
-  expect_error(
-    forge(iris2, x1$blueprint, outcomes = TRUE),
-    "The following required columns"
+  expect_snapshot(
+    error = TRUE,
+    forge(iris2, x1$blueprint, outcomes = TRUE)
   )
 
-  expect_error(
-    forge(iris2, x2$blueprint, outcomes = TRUE),
-    "The following required columns"
+  expect_snapshot(
+    error = TRUE,
+    forge(iris2, x2$blueprint, outcomes = TRUE)
   )
 })
 
@@ -141,14 +141,14 @@ test_that("missing predictor columns fail appropriately", {
     iris
   )
 
-  expect_error(
-    forge(iris[, 1, drop = FALSE], x$blueprint),
-    "Sepal.Width"
+  expect_snapshot(
+    error = TRUE,
+    forge(iris[, 1, drop = FALSE], x$blueprint)
   )
 
-  expect_error(
-    forge(iris[, 3, drop = FALSE], x$blueprint),
-    "'Sepal.Length', 'Sepal.Width'"
+  expect_snapshot(
+    error = TRUE,
+    forge(iris[, 3, drop = FALSE], x$blueprint)
   )
 })
 
@@ -390,14 +390,8 @@ test_that("new data classes are caught", {
   x2 <- mold(recipes::step_dummy(rec, Species), iris, blueprint = sparse_bp)
 
   # Silent recovery
-  expect_error(
-    xx1 <- forge(iris2, x1$blueprint),
-    NA
-  )
-  expect_error(
-    xx2 <- forge(iris2, x2$blueprint),
-    NA
-  )
+  expect_no_error(xx1 <- forge(iris2, x1$blueprint))
+  expect_no_error(xx2 <- forge(iris2, x2$blueprint))
 
   expect_s3_class(
     xx1$predictors$Species,
@@ -411,14 +405,8 @@ test_that("new data classes are caught", {
   x3 <- mold(recipes::recipe(Species ~ Sepal.Length, iris), iris)
   x4 <- mold(recipes::recipe(Species ~ Sepal.Length, iris), iris, blueprint = sparse_bp)
 
-  expect_error(
-    xx3 <- forge(iris2, x3$blueprint, outcomes = TRUE),
-    NA
-  )
-  expect_error(
-    xx4 <- forge(iris2, x4$blueprint, outcomes = TRUE),
-    NA
-  )
+  expect_no_error(xx3 <- forge(iris2, x3$blueprint, outcomes = TRUE))
+  expect_no_error(xx4 <- forge(iris2, x4$blueprint, outcomes = TRUE))
 
   expect_s3_class(
     xx3$outcomes$Species,
@@ -441,26 +429,18 @@ test_that("new data classes can interchange integer/numeric", {
   x1 <- mold(recipes::recipe(Species ~ Sepal.Length, iris), iris)
   x2 <- mold(recipes::recipe(Species ~ Sepal.Length, iris), iris, blueprint = sparse_bp)
 
-  expect_error(
-    forge(iris2, x1$blueprint),
-    NA
-  )
-  expect_error(
-    forge(iris2, x2$blueprint),
-    NA
-  )
+  expect_no_error(forge(iris2, x1$blueprint))
+  expect_no_error(forge(iris2, x2$blueprint))
 
   rec <- recipes::recipe(Sepal.Length ~ Species, iris)
   x3 <- mold(rec, iris)
   x4 <- mold(recipes::step_dummy(rec, Species), iris, blueprint = sparse_bp)
 
-  expect_error(
-    forge(iris2, x3$blueprint, outcomes = TRUE),
-    NA
+  expect_no_error(
+    forge(iris2, x3$blueprint, outcomes = TRUE)
   )
-  expect_error(
-    forge(iris2, x4$blueprint, outcomes = TRUE),
-    NA
+  expect_no_error(
+    forge(iris2, x4$blueprint, outcomes = TRUE)
   )
 })
 
diff --git a/tests/testthat/test-forge-xy.R b/tests/testthat/test-forge-xy.R
index 248fb429..22253e08 100644
--- a/tests/testthat/test-forge-xy.R
+++ b/tests/testthat/test-forge-xy.R
@@ -100,24 +100,16 @@ test_that("asking for the outcome is special cased for vector `y` values", {
   expect_equal(xx1$outcomes, xx3$outcomes)
   expect_equal(xx1$outcomes, xx3$outcomes)
 
-  # standard message
-  expect_error(
-    forge(iris, x1$blueprint, outcomes = TRUE),
-    "The following required columns"
+  # This expects:
+  # - standard message: "The following required columns"
+  # - more detail: "`new_data` must include a column with the automatically generated name, '.outcome'"
+  expect_snapshot(
+    error = TRUE,
+    forge(iris, x1$blueprint, outcomes = TRUE)
   )
-  expect_error(
-    forge(iris, x2$blueprint, outcomes = TRUE),
-    "The following required columns"
-  )
-
-  # but also more detail
-  expect_error(
-    forge(iris, x1$blueprint, outcomes = TRUE),
-    "`new_data` must include a column with the automatically generated name, '.outcome'"
-  )
-  expect_error(
-    forge(iris, x3$blueprint, outcomes = TRUE),
-    "`new_data` must include a column with the automatically generated name, '.outcome'"
+  expect_snapshot(
+    error = TRUE,
+    forge(iris, x2$blueprint, outcomes = TRUE)
   )
 })
 
@@ -125,10 +117,7 @@ test_that("new_data can be a matrix", {
   x <- mold(iris[, "Sepal.Length", drop = FALSE], iris$Species)
   iris_mat <- as.matrix(iris[, "Sepal.Length", drop = FALSE])
 
-  expect_error(
-    xx <- forge(iris_mat, x$blueprint),
-    NA
-  )
+  expect_no_error(xx <- forge(iris_mat, x$blueprint))
 
   sep_len <- iris$Sepal.Length
   pred_tbl <- tibble::tibble(Sepal.Length = sep_len)
@@ -147,18 +136,9 @@ test_that("new_data can only be a data frame / matrix", {
   x2 <- mold(iris[, "Sepal.Length", drop = FALSE], iris$Species, blueprint = sparse_bp)
   x3 <- mold(iris[, "Sepal.Length", drop = FALSE], iris$Species, blueprint = matrix_bp)
 
-  expect_error(
-    forge("hi", x1$blueprint),
-    "The class of `new_data`, 'character'"
-  )
-  expect_error(
-    forge("hi", x2$blueprint),
-    "The class of `new_data`, 'character'"
-  )
-  expect_error(
-    forge("hi", x3$blueprint),
-    "The class of `new_data`, 'character'"
-  )
+  expect_snapshot(error = TRUE, forge("hi", x1$blueprint))
+  expect_snapshot(error = TRUE, forge("hi", x2$blueprint))
+  expect_snapshot(error = TRUE, forge("hi", x3$blueprint))
 })
 
 test_that("missing predictor columns fail appropriately", {
@@ -170,22 +150,22 @@ test_that("missing predictor columns fail appropriately", {
     blueprint = bp
   )
 
-  expect_error(
-    forge(iris[, 1, drop = FALSE], x1$blueprint),
-    "Sepal.Width"
+  expect_snapshot(
+    error = TRUE,
+    forge(iris[, 1, drop = FALSE], x1$blueprint)
   )
-  expect_error(
-    forge(iris[, 1, drop = FALSE], x2$blueprint),
-    "Sepal.Width"
+  expect_snapshot(
+    error = TRUE,
+    forge(iris[, 1, drop = FALSE], x2$blueprint)
   )
 
-  expect_error(
-    forge(iris[, 3, drop = FALSE], x1$blueprint),
-    "'Sepal.Length', 'Sepal.Width'"
+  expect_snapshot(
+    error = TRUE,
+    forge(iris[, 3, drop = FALSE], x1$blueprint)
   )
-  expect_error(
-    forge(iris[, 3, drop = FALSE], x2$blueprint),
-    "'Sepal.Length', 'Sepal.Width'"
+  expect_snapshot(
+    error = TRUE,
+    forge(iris[, 3, drop = FALSE], x2$blueprint)
   )
 })
 
@@ -341,9 +321,8 @@ test_that("new data classes are caught", {
   x <- mold(iris[, "Species", drop = FALSE], iris$Sepal.Length)
 
   # Silent recovery
-  expect_error(
-    x_iris2 <- forge(iris2, x$blueprint),
-    NA
+  expect_no_error(
+    x_iris2 <- forge(iris2, x$blueprint)
   )
 
   expect_s3_class(
@@ -357,9 +336,8 @@ test_that("new data classes are caught", {
   iris3$.outcome <- iris2$Species
   iris3$Species <- NULL
 
-  expect_error(
-    xx_iris3 <- forge(iris3, xx$blueprint, outcomes = TRUE),
-    NA
+  expect_no_error(
+    xx_iris3 <- forge(iris3, xx$blueprint, outcomes = TRUE)
   )
 
   expect_s3_class(
@@ -383,14 +361,8 @@ test_that("new data classes can interchange integer/numeric", {
     blueprint = bp
   )
 
-  expect_error(
-    forge(iris2, x1$blueprint),
-    NA
-  )
-  expect_error(
-    forge(iris2, x2$blueprint),
-    NA
-  )
+  expect_no_error(forge(iris2, x1$blueprint))
+  expect_no_error(forge(iris2, x2$blueprint))
 })
 
 test_that("intercept is not included as a predictor", {
@@ -412,14 +384,8 @@ test_that("intercept is not included as a predictor", {
     "(Intercept)" %in% colnames(x2$blueprint$ptypes$predictors)
   )
 
-  expect_error(
-    xx1 <- forge(iris, x1$blueprint),
-    NA
-  )
-  expect_error(
-    xx2 <- forge(iris, x2$blueprint),
-    NA
-  )
+  expect_no_error(xx1 <- forge(iris, x1$blueprint))
+  expect_no_error(xx2 <- forge(iris, x2$blueprint))
 
   expect_equal(
     colnames(xx1$predictors),
diff --git a/tests/testthat/test-levels.R b/tests/testthat/test-levels.R
index 2a556df6..a2004499 100644
--- a/tests/testthat/test-levels.R
+++ b/tests/testthat/test-levels.R
@@ -66,9 +66,9 @@ test_that("Can extract levels from an outcome", {
     NULL
   )
 
-  expect_error(
-    get_outcome_levels("a"),
-    "`y` is of unknown type"
+  expect_snapshot(
+    error = TRUE,
+    get_outcome_levels("a")
   )
 
   expect_equal(
diff --git a/tests/testthat/test-model-offset.R b/tests/testthat/test-model-offset.R
index 205a69c2..40b46b94 100644
--- a/tests/testthat/test-model-offset.R
+++ b/tests/testthat/test-model-offset.R
@@ -38,9 +38,9 @@ test_that("(offset) is not recognized as an offset", {
 })
 
 test_that("Only numeric columns can be offsets", {
-  expect_error(
-    mold(~ Sepal.Width + offset(Species), iris),
-    "Column, 'offset"
+  expect_snapshot(
+    error = TRUE,
+    mold(~ Sepal.Width + offset(Species), iris)
   )
 })
 
@@ -117,9 +117,9 @@ test_that("offset columns are stored as predictors", {
   iris2 <- iris
   iris2$Sepal.Length <- NULL
 
-  expect_error(
-    forge(iris2, x$blueprint),
-    "Sepal.Length"
+  expect_snapshot(
+    error = TRUE,
+    forge(iris2, x$blueprint)
   )
 })
 
diff --git a/tests/testthat/test-mold-formula.R b/tests/testthat/test-mold-formula.R
index 8540a3e4..d4342504 100644
--- a/tests/testthat/test-mold-formula.R
+++ b/tests/testthat/test-mold-formula.R
@@ -110,9 +110,8 @@ test_that("can mold and not expand dummies", {
 })
 
 test_that("errors are thrown if `indicator = 'none'` and factor interactions exist", {
-  expect_error(
-    mold(~fac_1, example_train, blueprint = default_formula_blueprint(indicators = "none")),
-    NA
+  expect_no_error(
+    mold(~fac_1, example_train, blueprint = default_formula_blueprint(indicators = "none"))
   )
 
   expect_snapshot(error = TRUE, {
@@ -311,54 +310,54 @@ test_that("can mold formulas with special terms", {
 
 test_that("formulas with non-existent columns are caught", {
   bp <- default_formula_blueprint(composition = "dgCMatrix")
-  expect_error(
-    mold(fac_1 ~ y + z, example_train),
-    "predictors were not found in `data`: 'y', 'z'"
+  expect_snapshot(
+    error = TRUE,
+    mold(fac_1 ~ y + z, example_train)
   )
-  expect_error(
-    mold(fac_1 ~ y + z, example_train, blueprint = bp),
-    "predictors were not found in `data`: 'y', 'z'"
+  expect_snapshot(
+    error = TRUE,
+    mold(fac_1 ~ y + z, example_train, blueprint = bp)
   )
 
-  expect_error(
-    mold(y + z ~ fac_1, example_train),
-    "outcomes were not found in `data`: 'y', 'z'"
+  expect_snapshot(
+    error = TRUE,
+    mold(y + z ~ fac_1, example_train)
   )
-  expect_error(
-    mold(y + z ~ fac_1, example_train, blueprint = bp),
-    "outcomes were not found in `data`: 'y', 'z'"
+  expect_snapshot(
+    error = TRUE,
+    mold(y + z ~ fac_1, example_train, blueprint = bp)
   )
 })
 
 test_that("global environment variables cannot be used", {
-  expect_error(
+  expect_snapshot(
+    error = TRUE,
     {
       y <- 1
       mold(fac_1 ~ y, example_train)
-    },
-    "predictors were not found in `data`: 'y'"
+    }
   )
 })
 
 test_that("cannot manually remove intercept in the formula itself", {
   bp <- default_formula_blueprint(composition = "dgCMatrix")
-  expect_error(
-    mold(fac_1 ~ y + 0, example_train),
-    "`formula` must not contain"
+  expect_snapshot(
+    error = TRUE,
+    mold(fac_1 ~ y + 0, example_train)
   )
-  expect_error(
-    mold(fac_1 ~ y + 0, example_train, blueprint = bp),
-    "`formula` must not contain"
+  expect_snapshot(
+    error = TRUE,
+    mold(fac_1 ~ y + 0, example_train, blueprint = bp)
   )
 
-  expect_error(
-    mold(fac_1 ~ 0 + y, example_train),
-    "`formula` must not contain"
+  expect_snapshot(
+    error = TRUE,
+    mold(fac_1 ~ 0 + y, example_train)
   )
 
-  expect_error(
-    mold(fac_1 ~ y - 1, example_train),
-    "`formula` must not contain"
+  expect_snapshot(
+    error = TRUE,
+    mold(fac_1 ~ y - 1, example_train)
   )
 })
 
@@ -390,13 +389,11 @@ test_that("`NULL` can be used to represent empty RHS formulas", {
     mold(~0, example_train, blueprint = bp)
   })
 
-  expect_error(
-    x1 <- mold(~NULL, example_train),
-    NA
+  expect_no_error(
+    x1 <- mold(~NULL, example_train)
   )
-  expect_error(
-    x2 <- mold(~NULL, example_train, blueprint = bp),
-    NA
+  expect_no_error(
+    x2 <- mold(~NULL, example_train, blueprint = bp)
   )
 
   expect_equal(nrow(x1$predictors), 12)
@@ -404,9 +401,8 @@ test_that("`NULL` can be used to represent empty RHS formulas", {
   expect_equal(nrow(x2$predictors), 12)
   expect_equal(nrow(x2$outcomes), 12)
 
-  expect_error(
-    y <- mold(~NULL, example_train, blueprint = default_formula_blueprint(intercept = TRUE)),
-    NA
+  expect_no_error(
+    y <- mold(~NULL, example_train, blueprint = default_formula_blueprint(intercept = TRUE))
   )
 
   expect_equal(colnames(y$predictors), "(Intercept)")
@@ -431,14 +427,8 @@ test_that("intercepts can still be added when not using indicators (i.e. model.m
 
 test_that("`data` is validated", {
   bp <- default_formula_blueprint(composition = "dgCMatrix")
-  expect_error(
-    mold(fac_1 ~ num_2, 1),
-    "`data` must be a data.frame or a matrix"
-  )
-  expect_error(
-    mold(fac_1 ~ num_2, 1, blueprint = bp),
-    "`data` must be a data.frame or a matrix"
-  )
+  expect_snapshot(error = TRUE, mold(fac_1 ~ num_2, 1))
+  expect_snapshot(error = TRUE, mold(fac_1 ~ num_2, 1, blueprint = bp))
 })
 
 test_that("full interaction syntax is supported", {
@@ -563,13 +553,13 @@ test_that("`.` notation works as expected", {
 # the `"."` as a variable.
 test_that("`.` notation fails on the LHS", {
   bp <- default_formula_blueprint(composition = "dgCMatrix")
-  expect_error(
-    mold(. ~ fac_1, example_train),
-    "The left hand side of the formula cannot contain `.`"
+  expect_snapshot(
+    error = TRUE,
+    mold(. ~ fac_1, example_train)
   )
-  expect_error(
-    mold(. ~ fac_1, example_train, blueprint = bp),
-    "The left hand side of the formula cannot contain `.`"
+  expect_snapshot(
+    error = TRUE,
+    mold(. ~ fac_1, example_train, blueprint = bp)
   )
 })
 
diff --git a/tests/testthat/test-mold-recipe.R b/tests/testthat/test-mold-recipe.R
index ee5a3777..52e19270 100644
--- a/tests/testthat/test-mold-recipe.R
+++ b/tests/testthat/test-mold-recipe.R
@@ -136,10 +136,10 @@ test_that("can pass `strings_as_factors` through to `prep()`", {
 
 test_that("`data` is validated", {
   skip_if_not_installed("recipes")
-
-  expect_error(
-    mold(recipes::recipe(Species ~ Sepal.Length, data = iris), 1),
-    "`data` must be a data.frame or a matrix"
+  
+  expect_snapshot(
+    error = TRUE,
+    mold(recipes::recipe(Species ~ Sepal.Length, data = iris), 1)
   )
 })
 
diff --git a/tests/testthat/test-spruce.R b/tests/testthat/test-spruce.R
index df474ddc..9a4b6e9d 100644
--- a/tests/testthat/test-spruce.R
+++ b/tests/testthat/test-spruce.R
@@ -4,8 +4,8 @@ test_that("spruce - numeric", {
   expect_s3_class(spruced, "tbl_df")
   expect_equal(colnames(spruced), ".pred")
 
-  expect_error(spruce_numeric("hi"))
-  expect_error(spruce_numeric(matrix(1)))
+  expect_snapshot(error = TRUE, spruce_numeric("hi"))
+  expect_snapshot(error = TRUE, spruce_numeric(matrix(1)))
 })
 
 test_that("spruce - class", {
@@ -14,8 +14,8 @@ test_that("spruce - class", {
   expect_s3_class(spruced, "tbl_df")
   expect_equal(colnames(spruced), ".pred_class")
 
-  expect_error(spruce_class(1))
-  expect_error(spruce_class("hi"))
+  expect_snapshot(error = TRUE, spruce_class(1))
+  expect_snapshot(error = TRUE, spruce_class("hi"))
 })
 
 test_that("spruce - prob", {
@@ -27,29 +27,20 @@ test_that("spruce - prob", {
   expect_s3_class(spruced, "tbl_df")
   expect_equal(colnames(spruced), paste0(".pred_", pred_levels))
 
-  expect_error(spruce_prob(1, prob_matrix))
-  expect_error(spruce_prob(pred_levels, 1))
-  expect_error(spruce_prob("a", matrix("a")))
+  expect_snapshot(error = TRUE, spruce_prob(1, prob_matrix))
+  expect_snapshot(error = TRUE, spruce_prob(pred_levels, 1))
+  expect_snapshot(error = TRUE, spruce_prob("a", matrix("a")))
 
-  expect_error(
-    spruce_prob(c("a", "b"), matrix(1, ncol = 3)),
-    "2"
+  expect_snapshot(
+    error = TRUE,
+    spruce_prob(c("a", "b"), matrix(1, ncol = 3))
   )
 
-  expect_error(
-    spruce_prob(c("a", "b"), matrix(1, ncol = 3)),
-    "3"
+  expect_snapshot(
+    error = TRUE,
+    spruce_prob(c("a"), matrix(1, ncol = 2))
   )
 
-  expect_error(
-    spruce_prob(c("a"), matrix(1, ncol = 2)),
-    "1"
-  )
-
-  expect_error(
-    spruce_prob(c("a"), matrix(1, ncol = 2)),
-    "2"
-  )
 })
 
 test_that("`spruce_numeric_multiple()` generates the correct output", {
diff --git a/tests/testthat/test-standardize.R b/tests/testthat/test-standardize.R
index b3ce7fd6..bb6aff38 100644
--- a/tests/testthat/test-standardize.R
+++ b/tests/testthat/test-standardize.R
@@ -71,7 +71,7 @@ test_that("standardize - data.frame", {
 
   bad2 <- data.frame(x = "a", stringsAsFactors = FALSE)
 
-  expect_error(standardize(bad2), "These columns have unknown types: 'x'.")
+  expect_snapshot(error = TRUE, standardize(bad2))
 
   good <- bad
   colnames(good) <- c("a", "b")
@@ -88,6 +88,6 @@ test_that("standardize - data.frame", {
 })
 
 test_that("standardize - unknown", {
-  expect_error(standardize("hi"), "`y` is of unknown type 'character'")
-  expect_error(standardize(Sys.time()), "`y` is of unknown type 'POSIXct'")
+  expect_snapshot(error = TRUE, standardize("hi"))
+  expect_snapshot(error = TRUE, standardize(Sys.time()))
 })
diff --git a/tests/testthat/test-tune.R b/tests/testthat/test-tune.R
index 169a0657..19d0be2a 100644
--- a/tests/testthat/test-tune.R
+++ b/tests/testthat/test-tune.R
@@ -10,7 +10,7 @@ test_that("tune `id` value", {
 })
 
 test_that("`id` is validated", {
-  expect_error(tune(1), "The `id` should be a single character string.")
-  expect_error(tune(c("x", "y")), "The `id` should be a single character string.")
-  expect_error(tune(NA_character_), "The `id` can't be missing.")
+  expect_snapshot(error = TRUE, tune(1))
+  expect_snapshot(error = TRUE, tune(c("x", "y")))
+  expect_snapshot(error = TRUE, tune(NA_character_))
 })
diff --git a/tests/testthat/test-validation.R b/tests/testthat/test-validation.R
index 464575aa..69530a1d 100644
--- a/tests/testthat/test-validation.R
+++ b/tests/testthat/test-validation.R
@@ -5,9 +5,9 @@ test_that("validate_outcomes_are_univariate()", {
 
   expect_silent(validate_outcomes_are_univariate(1))
 
-  expect_error(
-    validate_outcomes_are_univariate(iris),
-    "The outcome must be univariate, but 5 columns were found."
+  expect_snapshot(
+    error = TRUE,
+    validate_outcomes_are_univariate(iris)
   )
 })
 
@@ -16,53 +16,53 @@ test_that("validate_outcomes_are_numeric()", {
     validate_outcomes_are_numeric(mtcars)
   )
 
-  expect_error(
-    validate_outcomes_are_numeric(iris),
-    "'Species': 'factor'"
+  expect_snapshot(
+    error = TRUE,
+    validate_outcomes_are_numeric(iris)
   )
 
   date <- as.POSIXct(as.POSIXlt(as.Date("2019-01-01")))
   x <- data.frame(x = date, y = factor("hi"))
 
-  expect_error(
-    validate_outcomes_are_numeric(x),
-    "'x': 'POSIXct', 'POSIXt'\n'y': 'factor'"
+  expect_snapshot(
+    error = TRUE,
+    validate_outcomes_are_numeric(x)
   )
 })
 
 test_that("validate_no_formula_duplication()", {
   expect_silent(validate_no_formula_duplication(y ~ x))
 
-  expect_error(
-    validate_no_formula_duplication(y ~ y),
-    "'y'"
+  expect_snapshot(
+    error = TRUE,
+    validate_no_formula_duplication(y ~ y)
   )
 
   expect_silent(validate_no_formula_duplication(y ~ log(y)))
 
-  expect_error(
-    validate_no_formula_duplication(y ~ log(y), original = TRUE),
-    "'y'"
+  expect_snapshot(
+    error = TRUE,
+    validate_no_formula_duplication(y ~ log(y), original = TRUE)
   )
 
-  expect_error(
-    validate_no_formula_duplication(y + x ~ y + x),
-    "'y', 'x'"
+  expect_snapshot(
+    error = TRUE,
+    validate_no_formula_duplication(y + x ~ y + x)
   )
 
   expect_silent(validate_no_formula_duplication(y ~ .))
 
-  expect_error(
-    validate_no_formula_duplication(y ~ . + y),
-    "'y'"
+  expect_snapshot(
+    error = TRUE,
+    validate_no_formula_duplication(y ~ . + y)
   )
 
   # offset() is a weird special case but this is ok
   expect_silent(validate_no_formula_duplication(offset(y) ~ offset(y)))
 
-  expect_error(
-    validate_no_formula_duplication(y ~ offset(y), original = TRUE),
-    "'y'"
+  expect_snapshot(
+    error = TRUE,
+    validate_no_formula_duplication(y ~ offset(y), original = TRUE)
   )
 })
 
@@ -74,9 +74,9 @@ test_that("validate_outcomes_are_factors()", {
   date <- as.POSIXct(as.POSIXlt(as.Date("2019-01-01")))
   x <- data.frame(x = date, y = "hi", stringsAsFactors = FALSE)
 
-  expect_error(
-    validate_outcomes_are_factors(x),
-    "'x': 'POSIXct', 'POSIXt'\n'y': 'character'"
+  expect_snapshot(
+    error = TRUE,
+    validate_outcomes_are_factors(x)
   )
 })
 
@@ -85,9 +85,9 @@ test_that("validate_outcomes_are_binary()", {
     validate_outcomes_are_binary(data.frame(x = factor(c("A", "B"))))
   )
 
-  expect_error(
-    validate_outcomes_are_binary(iris),
-    "'Sepal.Length': 0\n'Sepal.Width': 0\n'Petal.Length': 0\n'Petal.Width': 0\n'Species': 3"
+  expect_snapshot(
+    error = TRUE,
+    validate_outcomes_are_binary(iris)
   )
 })
 
@@ -96,17 +96,17 @@ test_that("validate_predictors_are_numeric()", {
     validate_predictors_are_numeric(mtcars)
   )
 
-  expect_error(
-    validate_predictors_are_numeric(iris),
-    "'Species': 'factor'"
+  expect_snapshot(
+    error = TRUE,
+    validate_predictors_are_numeric(iris)
   )
 
   date <- as.POSIXct(as.POSIXlt(as.Date("2019-01-01")))
   x <- data.frame(x = date, y = factor("hi"))
 
-  expect_error(
-    validate_predictors_are_numeric(x),
-    "'x': 'POSIXct', 'POSIXt'\n'y': 'factor'"
+  expect_snapshot(
+    error = TRUE,
+    validate_predictors_are_numeric(x)
   )
 })
 
@@ -115,8 +115,8 @@ test_that("validate_prediction_size()", {
     validate_prediction_size(mtcars, mtcars)
   )
 
-  expect_error(
-    validate_prediction_size(mtcars[1:5, ], mtcars),
-    "The size of `new_data` \\(32\\) must match the size of `pred` \\(5\\)."
+  expect_snapshot(
+    error = TRUE,
+    validate_prediction_size(mtcars[1:5, ], mtcars)
   )
 })