diff --git a/.github/workflows/pytest.yaml b/.github/workflows/pytest.yaml
index ec5addb6..49133f01 100644
--- a/.github/workflows/pytest.yaml
+++ b/.github/workflows/pytest.yaml
@@ -8,9 +8,23 @@ on:
     - cron: "0 12 * * 1"
   pull_request_target:
     types: [opened, synchronize, reopened, ready_for_review]
+
 jobs:
-  call-run-python-tests:
+  call-run-python-tests-unit:
+    uses: openclimatefix/.github/.github/workflows/python-test.yml@issue/pip-all
+    with:
+      # pytest-cov looks at this folder
+      pytest_cov_dir: "quartz_solar_forecast"
+      os_list: '["ubuntu-latest"]'
+      python-version: "['3.11']"
+      extra_commands: echo "HF_TOKEN=${{ vars.HF_TOKEN }}" > .env
+      pytest_numcpus: '1'
+      test_dir: tests/unit
+
+  call-run-python-tests-all:
+    # only run on push, not external PR
     uses: openclimatefix/.github/.github/workflows/python-test.yml@issue/pip-all
+    if: github.event_name == 'push'
     with:
       # pytest-cov looks at this folder
       pytest_cov_dir: "quartz_solar_forecast"
@@ -18,3 +32,4 @@ jobs:
       python-version: "['3.11']"
       extra_commands: echo "HF_TOKEN=${{ vars.HF_TOKEN }}" > .env
       pytest_numcpus: '1'
+      test_dir: tests
\ No newline at end of file
diff --git a/pytest.ini b/pytest.ini
new file mode 100644
index 00000000..7c95c298
--- /dev/null
+++ b/pytest.ini
@@ -0,0 +1,3 @@
+[pytest]
+markers =
+    integration: marks tests as integration tests
\ No newline at end of file
diff --git a/quartz_solar_forecast/eval/metrics.py b/quartz_solar_forecast/eval/metrics.py
index fedd2220..ea79ce2b 100644
--- a/quartz_solar_forecast/eval/metrics.py
+++ b/quartz_solar_forecast/eval/metrics.py
@@ -35,7 +35,7 @@ def metrics(results_df: pd.DataFrame, pv_metadata: pd.DataFrame, include_night:
         .mean(),
         4,
     )
-    print(f"MAE: {mae} kw, normalized {mae_normalized} %")
+    print(f"MAE: {mae} kw, normalized {100*mae_normalized} %")
 
     # calculate metrics over the different horizons hours
     # find all unique horizon_hours
@@ -74,7 +74,7 @@ def metrics(results_df: pd.DataFrame, pv_metadata: pd.DataFrame, include_night:
         )
 
         print(
-            f"MAE for horizon {horizon_group}: {mae} +- {1.96*sem}. mae_normalized: {100*mae_normalized} %"
+            f"MAE for horizon {horizon_group}: {mae} +- {1.96*sem:.3g}. mae_normalized: {100*mae_normalized:.3g} %"
         )
 
         # TODO add more metrics using ocf_ml_metrics
diff --git a/tests/eval/test_evaluation.py b/tests/integration/eval/test_evaluation.py
similarity index 83%
rename from tests/eval/test_evaluation.py
rename to tests/integration/eval/test_evaluation.py
index e6548b2b..9a9cbb58 100644
--- a/tests/eval/test_evaluation.py
+++ b/tests/integration/eval/test_evaluation.py
@@ -1,8 +1,10 @@
 from quartz_solar_forecast.evaluation import run_eval
 import tempfile
 import pandas as pd
+import pytest
 
 
+@pytest.mark.integration
 def test_run_eval():
 
     # create a fake dataframe
@@ -15,7 +17,7 @@ def test_run_eval():
                 "pv_id",
                 "timestamp",
             ],
-            data=[[8215, "2021-01-26 01:15:00"], [8215, "2021-01-30 16:30:00"]],
+            data=[[7593, "2021-08-21 12:00:00"], [7593, "2021-10-04 20:00:00"]],
         )
 
         testset_filename = tmpdirname + "/test_dataset.csv"
diff --git a/tests/eval/test_nwp.py b/tests/integration/eval/test_nwp.py
similarity index 92%
rename from tests/eval/test_nwp.py
rename to tests/integration/eval/test_nwp.py
index e9c3659b..0b18f18f 100644
--- a/tests/eval/test_nwp.py
+++ b/tests/integration/eval/test_nwp.py
@@ -1,8 +1,10 @@
 from quartz_solar_forecast.eval.nwp import get_nwp
 import pandas as pd
+import pytest
 
 
 # can take ~ 1 minute to run
+@pytest.mark.integration
 def test_get_nwp():
     # make test dataset file
     test_set_df = pd.DataFrame(
diff --git a/tests/eval/test_pv.py b/tests/integration/eval/test_pv.py
similarity index 91%
rename from tests/eval/test_pv.py
rename to tests/integration/eval/test_pv.py
index 51739547..56698c6e 100644
--- a/tests/eval/test_pv.py
+++ b/tests/integration/eval/test_pv.py
@@ -1,7 +1,8 @@
 from quartz_solar_forecast.eval.pv import get_pv_truth, get_pv_metadata
 import pandas as pd
+import pytest
 
-
+@pytest.mark.integration
 def test_get_pv_metadata():
     test_set_df = pd.DataFrame(
         [
@@ -16,6 +17,7 @@ def test_get_pv_metadata():
     assert "latitude" in metadata_df.columns
 
 
+@pytest.mark.integration
 def test_get_pv():
     # make test dataset file
     test_set_df = pd.DataFrame(
diff --git a/tests/test_forecast_no_ts.py b/tests/test_forecast_no_ts.py
deleted file mode 100644
index 0450379d..00000000
--- a/tests/test_forecast_no_ts.py
+++ /dev/null
@@ -1,29 +0,0 @@
-import pandas as pd
-from quartz_solar_forecast.forecast import run_forecast
-from quartz_solar_forecast.pydantic_models import PVSite
-
-
-def test_run_forecast_no_ts():
-    # make input data
-    site = PVSite(latitude=51.75, longitude=-1.25, capacity_kwp=1.25)
-
-    current_ts = pd.Timestamp.now().round("15min")
-    current_hr = pd.Timestamp.now().round(freq='h')
-
-    # run gradient boosting model with no ts
-    predications_df = run_forecast(site=site, model="gb", ts=current_ts)
-    # check current ts agrees with dataset
-    assert predications_df.index.min() == current_ts
-
-    print(predications_df)
-    print(f"Current time: {current_ts}")
-    print(f"Max: {predications_df['power_kw'].max()}")
-
-    # run xgb model with no ts
-    predications_df = run_forecast(site=site, model="xgb", ts=current_ts)
-    # check current ts agrees with dataset
-    assert predications_df.index.min() == current_hr
-
-    print(predications_df)
-    print(f"Current time: {current_ts}")
-    print(f"Max: {predications_df['power_kw'].max()}")
\ No newline at end of file
diff --git a/tests/api_tests/test_api.py b/tests/unit/api_tests/test_api.py
similarity index 100%
rename from tests/api_tests/test_api.py
rename to tests/unit/api_tests/test_api.py
diff --git a/tests/data/test_make_test_set.py b/tests/unit/data/test_make_test_set.py
similarity index 100%
rename from tests/data/test_make_test_set.py
rename to tests/unit/data/test_make_test_set.py
diff --git a/tests/data/test_process_pv_data.py b/tests/unit/data/test_process_pv_data.py
similarity index 100%
rename from tests/data/test_process_pv_data.py
rename to tests/unit/data/test_process_pv_data.py
diff --git a/tests/eval/test_eval_forecast.py b/tests/unit/eval/test_eval_forecast.py
similarity index 100%
rename from tests/eval/test_eval_forecast.py
rename to tests/unit/eval/test_eval_forecast.py
diff --git a/tests/eval/test_metrics.py b/tests/unit/eval/test_metrics.py
similarity index 100%
rename from tests/eval/test_metrics.py
rename to tests/unit/eval/test_metrics.py
diff --git a/tests/inverters/test_process_enphase_data.py b/tests/unit/inverters/test_process_enphase_data.py
similarity index 100%
rename from tests/inverters/test_process_enphase_data.py
rename to tests/unit/inverters/test_process_enphase_data.py
diff --git a/tests/inverters/test_victron.py b/tests/unit/inverters/test_victron.py
similarity index 100%
rename from tests/inverters/test_victron.py
rename to tests/unit/inverters/test_victron.py
diff --git a/tests/test_forecast.py b/tests/unit/test_forecast.py
similarity index 100%
rename from tests/test_forecast.py
rename to tests/unit/test_forecast.py
diff --git a/tests/unit/test_forecast_no_ts.py b/tests/unit/test_forecast_no_ts.py
new file mode 100644
index 00000000..0df021b6
--- /dev/null
+++ b/tests/unit/test_forecast_no_ts.py
@@ -0,0 +1,28 @@
+import pandas as pd
+from quartz_solar_forecast.forecast import run_forecast
+from quartz_solar_forecast.pydantic_models import PVSite
+
+
+def test_run_forecast_no_ts():
+    # make input data
+    site = PVSite(latitude=51.75, longitude=-1.25, capacity_kwp=1.25)
+
+    current_ts = pd.Timestamp.now()
+
+    # run gradient boosting model with no ts
+    predictions_df = run_forecast(site=site, model="gb")
+    # check current ts agrees with dataset
+    assert predictions_df.index.min() >= current_ts - pd.Timedelta(hours=1)
+
+    print(predictions_df)
+    print(f"Current time: {current_ts}")
+    print(f"Max: {predictions_df['power_kw'].max()}")
+
+    # run xgb model with no ts
+    predictions_df = run_forecast(site=site, model="xgb")
+    # check current ts agrees with dataset
+    assert predictions_df.index.min() >= current_ts - pd.Timedelta(hours=1)
+
+    print(predictions_df)
+    print(f"Current time: {current_ts}")
+    print(f"Max: {predictions_df['power_kw'].max()}")
\ No newline at end of file
diff --git a/tests/test_generate_forecast.py b/tests/unit/test_generate_forecast.py
similarity index 100%
rename from tests/test_generate_forecast.py
rename to tests/unit/test_generate_forecast.py
diff --git a/tests/utils/test_file_path.py b/tests/unit/utils/test_file_path.py
similarity index 100%
rename from tests/utils/test_file_path.py
rename to tests/unit/utils/test_file_path.py