Are you letting a few degrees bias your resource estimate? You already know that even a slight misalignment of your search ellipsoid can cause a block to be classified either above or below the cut-off grade, exceeding the confidence band. The search ellipsoid is like our invisible Midas hand that determines which samples influence each block's grade. Its properties fundamentally shape our resource models, yet their optimization often doesn't receive the attention it deserves. To determine its parameters, we primarily use variogram analysis and our geological understanding of the deposit. Let's break down the main critical search ellipsoid properties: ➔ Ellipsoid Radii: These define how far we search for samples in each direction. Ideally, they should align with variogram ranges that capture the spatial continuity of mineralization. ➔ Axis Orientation: The major axis should follow the direction of maximum continuity (typically along strike), with semi-major and minor axes oriented along dip and across the structure. ➔ Sector Division: Dividing the ellipsoid into sectors (octants or quadrants) ensures spatial representativity of samples. ➔ Min/Max Points Per Sector: These parameters balance between having enough data for statistical strength while preventing clustering bias. Using a globally-oriented search ellipsoid when estimating a folded and complex structure almost guarantees error. Instead, consider Dynamic anisotropy to locally adjust your ellipsoid orientation block by block, ensuring it follows the mineralization's true path. As we can adjust the dynamism of our search ellipsoid block by block, I think we should also able to adjust the it's dimension lengths according to the geological zones of our deposit. For instance; in some cases 25 m radii can give the same trust with 50 m radii according to the different zones in our deposit. Common errors I've encountered in projects ➔ Search ellipsoids that ignore geological controls ➔ Over-reliance on default software parameters ➔ Not testing multiple ellipsoid configurations ➔ Using the same ellipsoid dimensions across different domains Dear Colleagues, What search ellipsoid challenges have you faced? Have you tried dynamic Anisotropy ? What parameter selection approaches work best in your experience ? Perhaps, instead of the traditional block modelling approach where we deterministically assign a grade value at the centre of each block, a different methodology could be developed instead of block modelling ? ML algorithms have already started to be applied to block modelling. Even if we continue to use the block modelling concept, could we consider replacing the conventional search ellipsoid which serves as our resource estimation tool with dynamically shaped volumetric shapes that adapt the geometry of data ? Tunc Ozbek, MSc Mining Engineer. https://www.linkedin.com/in/tunc-ozbek-123603128/