Definition
In Salesforce Einstein (formerly used in Community reputation scoring), a decay parameter that determines how quickly the influence of older activities diminishes over time when calculating engagement or reputation scores.
Real-World Example
Consider a scenario where a data analyst at MarketPulse is working with Half-life to uncover trends and patterns hidden in their CRM data. By configuring Half-life, they create visualizations that tell a clear story about business performance. The executive team uses these insights to adjust strategy mid-quarter and the company exceeds its revenue target by 12%.
Why Half-life Matters
In Salesforce Einstein contexts (and previously in Community reputation scoring), Half-life is a decay parameter that determines how quickly the influence of older activities diminishes when calculating engagement or reputation scores. The concept comes from physics, where half-life is the time for a quantity to reduce to half its initial value. In scoring contexts, it controls how recent activity is weighted relative to older activity.
Half-life parameters appear in various scoring mechanisms across Salesforce: engagement scores, lead scores, predictive models, and similar features that need to balance recent versus historical signals. A short half-life (like 7 days) emphasizes very recent activity; a long half-life (like 365 days) treats older activity as still meaningful. Choosing the right half-life depends on the use case: marketing engagement might want short half-lives to react to current behavior, while customer lifetime value might want longer half-lives to capture sustained patterns.
How Organizations Use Half-life
- •Apex Analytics — Tunes the half-life parameter on their engagement scoring model to match the speed of customer behavior changes. Short half-lives caught churn signals faster but were noisier.
- •MarketPulse — Uses different half-lives for different scoring purposes: short for current campaign engagement, long for overall customer health.
- •SilverLine Corp — Tested several half-life values to find the right balance between responsiveness to new behavior and stability of scores.