Prerequisites
To change weights, you need pruning settings permissions.
Steps
Read the score
Open Patents > Intelligence > Pruning. The Keep Score column shows one number from 0 to 100 per family.
Read the color. Green (70 and above) suggests keeping, orange (40 to 69) suggests review, and red (below 40) suggests pruning.
Hover the number to open the breakdown. Each factor shows its score out of 10 and the points it contributes, heaviest factor first.
Look for muted rows. A factor with no data counts a neutral 5 and is shown dimmed. Families not yet scored read Not scored yet for the active pruning cycle.
Click the Keep Score header to sort. The first click sorts ascending, so the weakest families come first.
Click Filter and use Keep Score to keep only a percentile band, such as Bottom 10%.
The same score appears on the Make a Decision page and on the patent detail page's Pruning card.
Configure the factors
Open Settings > Workflows & Automation > Pruning and scroll to Keep Score.
Adjust the Weight of each built-in factor. The defaults are Unique Claim Scope 30, Detectability 20, Deep Research 30, Years Left 10, and Technology Balance 10.
Watch the Total row. Weights must add up to 100 before Save is enabled.
Click + Add custom factor to add your own. Enter a Title and a Description, then set a Weight. ArcPrime scores custom factors from their description.
Click Save. If you added or edited a custom factor, scores recompute and may take a few hours to update.
Click Reset to defaults at any time to return to the built-in weights.
Note: The Deep Research factor uses the newest completed deep research on the case. Without one, a Quick Research or screening project match scores in a narrower band, and a case with no research counts neutral.
Related Articles
Pruning Candidates and Scores — Every column on the candidates list
How to Configure Pruning Settings — Cadence, lead time, and high-touch settings
Running Effective Pruning Cycles — Practices for a repeatable review
Deep Research Overview — The evidence behind the Deep Research factor

