diff --git a/src/finn/custom_op/fpgadataflow/streamingmaxpool_batch.py b/src/finn/custom_op/fpgadataflow/streamingmaxpool_batch.py
index d9ffea4d9cd8895fdf55a497e8c7d0e49808ac95..0131ca40bd9f32b04a7fd6e0b8444884897580a8 100755
--- a/src/finn/custom_op/fpgadataflow/streamingmaxpool_batch.py
+++ b/src/finn/custom_op/fpgadataflow/streamingmaxpool_batch.py
@@ -145,6 +145,12 @@ class StreamingMaxPool_Batch(HLSCustomOp):
             return int(exp_cycles)
         else:
             # TODO: adjust inaccurate formula
+            warnings.warn(
+                """Estimated latency for layer {} can be (2-3x) lower
+                 than actual latency!""".format(
+                    self.onnx_node.name
+                )
+            )
             return int(ifm_dim[1] * (ifm_dim[1] + (ifm_dim[1] / k[1])))
 
     def get_instream_width(self):