[GTM glossary](/glossary.txt) Predictive scoring Contextual Intelligence & AI # Predictive scoring In short Predictive scoring is a method of ranking leads or accounts using a model that learns from historical outcomes. Each record gets a score that estimates how closely it resembles past customers who bought. It differs from rule-based scoring, where a person assigns points by hand. Predictive scoring finds the weights from data. It is a narrower idea than predictive analytics, which also covers forecasting and churn. ## How predictive scoring works The model is trained on records marked won or lost, using traits such as industry, size, tools, and engagement. It then applies the learned pattern to new records and returns a score or tier, such as high, medium, or low. --- ## Predictive vs. rule-based scoring * **Rule-based:** Points set by people. Transparent and quick to start, but reflects assumptions. * **Predictive:** Weights learned from data. Can find non-obvious patterns, but needs enough history and is harder to explain. --- ## How to roll it out ### Check you have enough history You need a meaningful number of past wins and losses with clean fields. If you have only a few dozen deals, start with rules. ### Pick one outcome Score for one result, such as opportunity created, rather than mixing goals. ### Validate before launch Compare scores with real outcomes on recent data. Say the top 20 percent of scored leads should convert clearly better than the bottom half; that is an example target, not a benchmark. ### Use scores to act Route high scorers faster, adjust cadence by tier, and keep a rep override for context the model cannot see. --- ## Common mistakes * **Scoring fit only:** A high-fit account with no timing signal may not buy yet. Combine fit with intent or engagement. * **Opaque scores:** Show the top reasons so reps trust and use the score. * **Training on biased history:** If reps never called certain segments, the model will underrate them. * **Set and forget:** Retrain and compare against outcomes as the market shifts. --- ## Predictive scoring FAQs **What is predictive scoring?** Predictive scoring uses a model trained on past won and lost deals to give each lead or account a score estimating how likely it is to convert. Teams use the score to decide who to contact first and how much effort to invest. **How is predictive scoring different from traditional lead scoring?** Traditional lead scoring uses points that people assign by hand. Predictive scoring learns the weights from historical outcomes. The predictive version can uncover hidden patterns, while the manual version is easier to explain and needs less data. **How much data do you need for predictive scoring?** You need enough closed-won and closed-lost records with consistent fields for a model to learn reliable patterns. With only a few dozen deals, a simple rule-based score is usually more dependable until you collect more history. **Can predictive scoring replace sales judgment?** No. A score is a starting point that ranks likelihood, not a verdict. Reps still know context the model cannot see, such as a champion change or a competing project, so keep room to override and review scores against outcomes. Share term --- ## More related phrases * [Machine learning](/glossary/machine-learning.txt) * [Machine learning model](/glossary/machine-learning-model.txt) * [ICP scoring](/glossary/icp-scoring.txt) ## Similar phrases * [Propensity model](/glossary/propensity-model.txt) A propensity model is a statistical or machine learning model that estimates how likely an account or contact is to take an action, such as buying, churning, or upgrading. * [Predictive lead scoring](/glossary/predictive-lead-scoring.txt) Predictive lead scoring uses a model trained on past deals to rank the leads in your database by how likely each one is to become a customer. * [AI lead scoring](/glossary/ai-lead-scoring.txt) AI lead scoring uses machine learning to rank leads by how likely they are to buy, learning from past wins and losses instead of relying on manually assigned points. * [Fit score](/glossary/fit-score.txt) A fit score is a number that rates how closely a company or contact matches your target customer, using static traits such as industry, size, and tech stack. * [Lead scoring models](/glossary/lead-scoring-models.txt) Lead scoring models are the frameworks used to calculate lead scores, including manual point systems, predictive models, and fit and intent grids.