diff --git a/metriks/ranking.py b/metriks/ranking.py
index b61ca66..4f6204e 100644
--- a/metriks/ranking.py
+++ b/metriks/ranking.py
@@ -4,10 +4,15 @@
 Metrics to use for ranking models.
 """
 
+from typing import Tuple
+
 import numpy as np
+from numpy.ma import MaskedArray
+from pytypes import typechecked
 
 
-def check_arrays(y_true, y_prob):
+@typechecked
+def check_arrays(y_true: np.ndarray, y_prob: np.ndarray) -> None:
     # Make sure that inputs this conforms to our expectations
     assert isinstance(y_true, np.ndarray), AssertionError(
         'Expect y_true to be a {expected}. Got {actual}'
@@ -36,7 +41,8 @@ def check_arrays(y_true, y_prob):
     )
 
 
-def check_k(n_items, k):
+@typechecked
+def check_k(n_items: int, k: int) -> None:
     # Make sure that inputs conform to our expectations
     assert isinstance(k, int), AssertionError(
         'Expect k to be a {expected}. Got {actual}'
@@ -49,7 +55,8 @@ def check_k(n_items, k):
     )
 
 
-def recall_at_k(y_true, y_prob, k):
+@typechecked
+def recall_at_k(y_true: np.ndarray, y_prob: np.ndarray, k: int) -> float:
     """
     Calculates recall at k for binary classification ranking problems. Recall
     at k measures the proportion of total relevant items that are found in the
@@ -68,7 +75,7 @@ def recall_at_k(y_true, y_prob, k):
             sorted order by y_prob
 
     Returns:
-        recall (~np.ndarray): The recall at k
+        recall (float): The recall at k
 
     Example:
     >>> y_true = np.array([
@@ -118,7 +125,8 @@ def recall_at_k(y_true, y_prob, k):
     return recall
 
 
-def precision_at_k(y_true, y_prob, k):
+@typechecked
+def precision_at_k(y_true: np.ndarray, y_prob: np.ndarray, k: int) -> float:
     """
     Calculates precision at k for binary classification ranking problems.
     Precision at k measures the proportion of items in the top k (in ranked
@@ -134,7 +142,7 @@ def precision_at_k(y_true, y_prob, k):
             sorted order by y_prob
 
     Returns:
-        precision_k (~np.ndarray): The precision at k
+        precision_k (float): The precision at k
 
     Example:
     >>> y_true = np.array([
@@ -178,7 +186,8 @@ def precision_at_k(y_true, y_prob, k):
     return precision
 
 
-def mean_reciprocal_rank(y_true, y_prob):
+@typechecked
+def mean_reciprocal_rank(y_true: np.ndarray, y_prob: np.ndarray) -> MaskedArray:
     """
     Gets a positional score about how well you did at rank 1, rank 2,
     etc. The resulting vector is of size (n_items,) but element 0 corresponds to
@@ -217,7 +226,8 @@ def mean_reciprocal_rank(y_true, y_prob):
     return ma.mean(axis=0)
 
 
-def label_mean_reciprocal_rank(y_true, y_prob):
+@typechecked
+def label_mean_reciprocal_rank(y_true: np.ndarray, y_prob: np.ndarray) -> MaskedArray:
     """
     Determines the average rank each label was placed across samples. Only labels that are
     relevant in the true data set are considered in the calculation.
@@ -241,7 +251,8 @@ def label_mean_reciprocal_rank(y_true, y_prob):
     return ma.mean(axis=0)
 
 
-def ndcg(y_true, y_prob, k=0):
+@typechecked
+def ndcg(y_true: np.ndarray, y_prob: np.ndarray, k=0) -> np.float64:
     """
     A score for measuring the quality of a set of ranked results. The resulting score is between 0 and 1.0 -
     results that are relevant and appear earlier in the result set are given a heavier weight, so the
@@ -322,7 +333,8 @@ def ndcg(y_true, y_prob, k=0):
     return np.mean(sample_ndcg, dtype=np.float64)
 
 
-def generate_y_pred_at_k(y_prob, k):
+@typechecked
+def generate_y_pred_at_k(y_prob: np.ndarray, k: int) -> np.ndarray:
     """
     Generates a matrix of binary predictions from a matrix of probabilities
     by evaluating the top k items (in ranked order by y_prob) as true.
@@ -362,11 +374,13 @@ def generate_y_pred_at_k(y_prob, k):
     index_array = np.argsort(y_prob, axis=1)
     col_idx = np.arange(y_prob.shape[0]).reshape(-1, 1)
     y_pred = np.zeros(np.shape(y_prob))
-    y_pred[col_idx, index_array[:, n_items-k:n_items]] = 1
+    y_pred[col_idx, index_array[:, n_items - k:n_items]] = 1
     return y_pred
 
 
-def confusion_matrix_at_k(y_true, y_prob, k):
+@typechecked
+def confusion_matrix_at_k(y_true: np.ndarray, y_prob: np.ndarray, k: int) -> Tuple[
+    np.ndarray, np.ndarray, np.ndarray, np.ndarray]:
     """
     Generates binary predictions from probabilities by evaluating the top k items
     (in ranked order by y_prob) as true. Uses these binary predictions along with
@@ -433,4 +447,3 @@ def confusion_matrix_at_k(y_true, y_prob, k):
     fn = np.count_nonzero((y_pred - 1) * y_true, axis=0)
 
     return tn, fp, fn, tp
-
diff --git a/setup.py b/setup.py
index 9c70672..61cd427 100644
--- a/setup.py
+++ b/setup.py
@@ -26,6 +26,7 @@
     # Insert dependencies list here
     install_requires=[
         'numpy',
+        'pytypes'
     ],
     setup_requires=["setuptools-scm"],
     extras_require={
diff --git a/tests/test_label_mean_reciprocal_rank.py b/tests/test_label_mean_reciprocal_rank.py
index 374e28d..f8a6b86 100644
--- a/tests/test_label_mean_reciprocal_rank.py
+++ b/tests/test_label_mean_reciprocal_rank.py
@@ -16,6 +16,7 @@ def test_label_mrr_with_errors():
     expected = np.ma.array([1.0, (7/4) / 3, 2/3, 1/3])
 
     np.testing.assert_allclose(result, expected)
+
     print(result)
 
 
diff --git a/tests/test_ndcg.py b/tests/test_ndcg.py
index 7f0144c..dafa738 100644
--- a/tests/test_ndcg.py
+++ b/tests/test_ndcg.py
@@ -207,6 +207,7 @@ def test_ndcg_0_over_0_error():
 
     np.testing.assert_allclose([actual], [expected])
 
+
 if __name__ == '__main__':
     test_ndcg_perfect()
     test_ndcg_errors()