---
title: "Behavioral model refresh"
slug: "behavioral-model-refresh"
updated: 2025-07-17T13:50:41Z
published: 2025-07-17T13:50:41Z
---

> ## Documentation Index
> Fetch the complete documentation index at: https://help.madkudu.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Behavioral model refresh

This comprehensive guide provides a 3-week approach to refreshing your behavioral model from initial assessment through deployment. The process balances data-driven insights with business intelligence, ensuring your refreshed model not only performs statistically well but also supports your sales team's daily operations and strategic objectives.

The methodology outlined here emphasizes validation at every step, cross-functional collaboration, and iterative improvement to deliver a model that drives measurable business impact.

## Prerequisites - Week 1

- Confirm which behavioral integrations are connected
- Confirm if the model(s) to update are lead-level, account-level, or both
- Establish success criteria for the refresh project
- Confirm which audience and conversion mappings should be used to measure the performance of the model
- [Assess the performance of the behavioral model against those with the predictions pages](/v1/docs/monitoring-your-model-performance)
- [Spot check the deep dive to draw first conclusions on what to change in the weights](/v1/docs/deep-dive-reports)
  - Identify patterns in false positives and false negatives
- Internally align on what changed since the last fit model refresh (New behavioral integration? New events?)
- Internally align on the hand raiser events the behavioral model should always qualify

## Step by step configuration - Week 2

- [Update audience](/v1/docs/audience-mapping) and [conversion mappings](/v1/docs/conversion-mapping), if necessary
- [Check the unmapped events report](/v1/docs/event-mapping-output-report-1)
- [Map missing events in the event mapping](/v1/docs/event-mapping)
- [Create relevant aggregations](/v1/docs/what-are-aggregations)
- Duplicate the live behavioral model(s)
- [Update dataset dates, audience, and conversion](/v1/docs/likelihood-to-buy-training-and-validation-datasets)
- [Update the weights](/v1/docs/likelihood-to-buy-event-weights)
  - Qualify hand raiser events (demo request, contact us form, viewed pricing page…)
  - Put a weight of 0 for:
    - Sales actions
    - Frequent low engagement actions (open email, viewed blog page…)
  - Leave the statistically suggested weights for the rest of the events

- Adjust the thresholds
- Review the performance of the model and a validation sample

## Feedback and iteration - Week 2&3

- Do several sample reviews with peers as a sanity check
- Share some examples with Sales to get their feedback
- Iterate on model configuration based on feedback

## Go Live - Week 3

- Deploy the new model
  - Consider deploying on a Friday afternoon to avoid disrupting business

**Estimated time: 10 hours**