DORA supports running multiple experiments on the same asset, varying input features, deposit types, or Learning Points to test different geological hypotheses. However, targeting decisions require a single source of truth to act on.

The Compare & Merge tool lets you evaluate those experiments side by side and combine their outputs into a single Merged VPS. You can then generate and evaluate exploration targets directly from that result.

In this article:

Before you can compare and merge experiments, all experiments must meet the following conditions:

The Compare & Merge tool is accessed from the DORA index page.

To add or remove experiments from the dashboard, use the + Add button or click the x beside any experiment tag at the top of the dashboard:

The Compare tab contains five sub-tabs. Each provides a different view of how your experiments relate to each other.

This tab displays the Prediction Map for each selected experiment side by side. Use the VPS Threshold slider beneath each map to filter the display.

Use this view to spot spatial differences between experiments — for example, whether a change in input features shifted high-VPS areas or produced a different pattern across the Area of Interest (AOI).

This line chart shows the spread of VPS values across all grid cells for each experiment. Each experiment is represented by its own line.

Use this view to see where most of your VPS values are clustering. A sharp, narrow peak means values are concentrated in a tight range. A flatter, wider curve means values are more spread out across the AOI. Comparing the shapes of the lines helps you assess whether one experiment is producing stronger differentiation between high and low prospectivity areas than another.

A summary table of key statistics for each experiment's VPS values. For a full explanation of each statistic, see Summary Statistics below.

This tab displays the Performance Breakdown matrix for each selected experiment side by side.

Below the metrics, the Performance Breakdown matrix shows the four prediction outcomes for each experiment:

Ideally, True Positives and True Negatives should be high, while False Positives and False Negatives should be as low as possible.

Use this tab to compare model reliability across experiments before deciding which to merge. An experiment with a significantly lower F1 or higher False Positive rate may contribute a less reliable signal to the Merged VPS.

For a full explanation of these metrics and how to interpret them, see Prediction Accuracy and Performance Breakdown.

This tab displays a bar chart and summary table of VPS score statistics for each experiment.

The Statistics Comparison bar chart plots each statistic as grouped bars, with one colour per experiment, so you can compare experiments across the full range of their score distributions at a glance.

The table provides the following statistics for each experiment:

This tab shows how similar each pair of experiments is to each other, cell by cell across the shared grid. Scores are colour-coded, and range from -1 to +1:

Each experiment is perfectly correlated with itself, so the diagonal always shows 1.00. The meaningful values are the remaining cells, each showing the correlation between a pair of experiments.

A lower correlation suggests the experiments are responding to different geological inputs and may produce a more informative Merged VPS when merged. A very high correlation suggests the experiments are similar, and merging them may add limited value.

This tab shows how input features contributed to each experiment's predictions, allowing you to compare feature influence across experiments side by side.

Use the Top N Features slider to control how many of the top-ranked input features are displayed in the charts below.

A horizontal bar chart showing the SHAP Contribution % for each input feature, with one bar per experiment. This lets you compare directly how much each feature drove predictions across your selected experiments.

A colour-coded table showing the same SHAP Contribution % data in grid format. Warmer colours indicate a higher contribution. A value of 0.0 means that feature had no influence in that experiment, or was not included as an input feature.

Together, these two views help you identify which input features are driving your results and whether that pattern is consistent across experiments, or whether different experiments are responding to different geological signals.

For a full explanation of these metrics and how to interpret them, see Feature Importance.

The Merge tab is where you configure and run the experiment merge. Two methods are available:

The recommended method. Bayesian Gaussian is a statistical method that learns from the Learning Points in each of your selected experiments.

It looks at the VPS values at known positive and negative locations across all selected experiments and uses that information to determine how much weight each experiment's model should carry in the final Merged VPS.

Models that more clearly separate prospective from non-prospective locations carry more influence in the merged result.

To merge using the Bayesian Gaussian method:

This method calculates a weighted average of the VPS values from each selected experiment across the shared grid. Equal weights produce a true simple average. You can adjust the sliders to emphasise one experiment over another.

To merge using the Simple Average method:

After the merge completes, the Merge tab displays the Training Results.

Click See results in the banner to review the full Results tab.

The Results tab displays your Merged VPS and allows you to extract exploration targets.

The following statistics are shown at the top of the page:

The Merged Spatial Map displays your Merged VPS. Use the VPS Threshold slider under it to filter the display.

The score distribution chart shows the spread of Merged VPS scores, with the individual input experiment distributions overlaid as lines for reference.

Use the Target Extraction section to generate exploration targets from the Merged VPS using spatial clustering.

For a full explanation of Target Groups in DORA, see Create Target Groups.

After you click Generate Target Groups, the following summary statistics appear:

Each target is plotted as a bubble on the Risk / Reward Plot:

Targets in the upper left of the plot have high prospectivity and low uncertainty — these are your highest priority exploration targets.

Click Map to view the same targets overlaid on the Merged VPS Prediction Map.

Click any target to highlight its Mean VPS, grid cell count, and variance in the Target Statistics table.

To save your Merged VPS to the asset:

To export your Merged VPS for use in GIS or other tools:

Reach out to your dedicated DORA contact or email support@vrify.com for more information.