diff --git a/cfa_rt_postprocessing/main_functions.py b/cfa_rt_postprocessing/main_functions.py
index 1066119..790253d 100644
--- a/cfa_rt_postprocessing/main_functions.py
+++ b/cfa_rt_postprocessing/main_functions.py
@@ -405,6 +405,16 @@ def merge_and_render_anomaly(
     except Exception as e:
         console.log(f"Failed to upload the flu anomaly report: {e}")
 
+    # === Calculate the categories for the samples =====================================
+    console.status("Calculating the categories from the samples")
+    p_growing = calculate_categories(final_samples)
+
+    # Save it to file as parquet, and as CSV
+    p_growing_pq_file = internal_review / "p_growing.parquet"
+    p_growing_csv_file = internal_review / "p_growing.csv"
+    p_growing.write_parquet(p_growing_pq_file)
+    p_growing.write_csv(p_growing_csv_file)
+
     # === Clean up =====================================================================
     conn.close()
     console.log(f"Cleaning up {root} folder")
@@ -437,6 +447,55 @@ def render_report(
     )
 
 
+def calculate_categories(samples_file: Path) -> pl.DataFrame:
+    """
+    Takes in the path to the samples parquet file, calculates the five categories for
+    each geo_value, disease, and reference_date.
+
+    Returns a DataFrame with the columns:
+    - geo_value
+    - disease
+    - reference_date
+    - p_growing
+    - category
+
+    The samples file is fairly large, so use duckdb to help things go faster and more
+    efficiently.
+    """
+    conn = duckdb.connect()
+    conn.sql(
+        f"""
+    -- First create a 'view' of the samples. Don't use a table, because that would read
+    -- it all into RAM. ATM, a single samples file is about 1.7GB, but attempting to load
+    -- it into RAM crashes my 32GB machine.
+    CREATE OR REPLACE VIEW samples AS
+    SELECT reference_date, geo_value, disease, "value" as Rt
+    FROM '{str(samples_file.absolute())}'
+    WHERE "_variable" = 'Rt';
+
+    -- Calculate the p_growing for each geo_value, disease, and reference_date
+    CREATE OR REPLACE TABLE p_growing AS SELECT
+    geo_value, disease, reference_date,
+    AVG(IF(Rt > 1, 1, 0)) AS p_growing,
+    CASE
+        WHEN (p_growing > 0.9) AND (p_growing <= 1.0) THEN 'Growing'
+        WHEN (p_growing > 0.75) AND (p_growing <= 0.9) THEN 'Likely Growing'
+        WHEN (p_growing > 0.25) AND (p_growing <= 0.75) THEN 'Not Changing'
+        WHEN (p_growing > 0.10) AND (p_growing <= 0.25) THEN 'Likely Declining'
+        WHEN (p_growing >= 0.0) AND (p_growing <= 0.10) THEN 'Declining'
+    END AS five_cat_p_growing
+    FROM samples
+    GROUP BY ALL
+    ORDER BY ALL;
+    """
+    )
+
+    p_growing = conn.sql("SELECT * FROM p_growing").pl()
+    conn.close()
+
+    return p_growing
+
+
 if __name__ == "__main__":
     # Some sample inputs for testing. Need to move something like this to an actual test
     args = {