---
title: "What type of events can be used in a behavioral model?"
slug: "what-type-of-events-can-be-used-in-a-behavioral-model"
updated: 2025-11-05T15:47:09Z
published: 2025-11-05T15:47:09Z
---

> ## 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.

# What type of events can be used in a behavioral model?

A Behavioral segmentation — or **Likelihood to Buy model** — is the one that ingests behavioral data (in-app usage, website activity, marketing campaigns, etc.) of users/visitors to measure the engagement of the lead (i.e. has the lead tried to engage with you?). Because the input of the model uses behavioral data, the scores are recomputed several times throughout the day to reflect as much as possible the level of intent of the lead at a certain time.

## How is the Likelihood to Buy score built?

Essentially the score is built on **"Active events"** which are triggered when the lead has taken an action (e.g. connected to their account, added users to the account, opened email, visited a web page, registered to a webinar, filled out a form ....) and we ignore any **"Passive event"** (e.g. received an email or invitation and didn't take action, system events fired such as "enrichment"). In addition, the score is also **degraded by "Negative events"** that suggest the person is not interested (e.g. requests to unsubscribe, bounced emails).

## How will the Likelihood to Buy score evolve?

The score will be **updated in batch every 4-12 hours** with any new activity.

The score will **decrease** over time to differentiate the people who recently showed activity or requested a demo versus the ones who did it 6 months ago. We usually apply **a time decay of 90 days**, meaning **the score will return to 0 if there was no event logged for this person in the previous 3 months.**

## What Behavioral data points can be used?

The idea is to catch events which are most likely leading to a conversion and reflecting **intent**. It means identifying what events will distinguish a "hot" lead from a lead just surfing the web or vaguely curious of your product.

***The list of activities below is given as example and is not exhaustive. In addition, not all events below are necessary, as it depends on what you would like the Likelihood to Buy score to reflect the most and on data availability.***

- **App Usage**

This is highly specific to your product and what main events you are tracking. Examples:
  - New user added / deleted
  - Account deactivated
  - Core product action
  - Created project
- **Web Activity**

  

Examples:
  - Demo request / contact us form submission
  - Sign up for free account
  - Chatbot conversation
  - Watched demo
  - Webpages viewed : home page, pricing page, features page, blog, forum, case studies, partners, blog pages
- **Marketing Activity**

  

Examples:
  - Webinar registration, attendance
  - Conference / event registration, attendance
  - eBook, case study or Whitepaper downloads
  - Signup for newsletter
- **Email Activity**

  

Examples:
  - Marketing emails opened, clicked, replied
  - Email bounced (hard or soft)
  - Unsubscribe request
- **Sales Activity**

  

Examples
  - Call scheduled / attended / cancelled

If you are sending all of your data, we will need to understand:

- how you link those events to a lead (email address)
- if this link and tracking is reliable
- which events are post- or pre-conversion
- which events are non-user activity (technical process triggered like enrichment_provider, data_submitted, etc.)

## We have worked in the past with the following integrations:

- Segment
- Marketo
- Salesforce campaigns responses
- Salesforce tasks
- Eloqua
- Hubspot
- Amplitude
- Stripe
- Kissmetrics
- ... and some more

## How does behavioral data look like?

| Attribute |  | Format | Example | Description |
| --- | --- | --- | --- | --- |
| ```plaintext event_key ``` | *required* | String | "abc123" | A unique key identifying the event. If you do not have one, we suggest creating a combination of event_text + contact_key + event_timestamp |
| ```plaintext event_text ``` | *required* | String | “signup”, “login”, “invited a friend” | The action taken by the user. |
| ```plaintext event_timestamp ``` | *required* | Unix time | “1436172703” | The time at which the event happened |
| ```plaintext contact_key ``` | *required* | String | "abc123" or "paul@madkudu.com" | The unique identifier of the user who performed the action. This needs to be the same as the contact_key field in the identify file. |
| ```plaintext event_* ``` | *optional* | String or Numeric |  | properties describing the event (e.g. event_url for the url of visited page, event_form_title for the title of form submitted...) |

**Example**:

From **Segment's** track object we get:

- the event id, the event name, the contact id, the event timestamp

From **Salesforce Campaigns** we get:

- the campaign type, the campaign member status, the lead/contact id and campaign member created date

If you have any question, you can shoot us an email at [product@madkudu.com.](mailto:product@madkudu.com)

The HG Insights Team

-