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

# Markov Attribution Model

> Learn in detail about the Markov model and how it is calculated

# What is the Markov Model?

The **Markov Chains** attribution model (or simply Markov) is an advanced way to understand the real value of each channel, campaign, or ad in conversion journeys. Unlike traditional models (*First Click, Last Click, Linear, Time Decay, etc.*), Markov doesn't distribute credit based solely on fixed points in the journey, but on the importance each interaction has within all observed journeys.

In other words: Markov doesn't just look at *"who brought"* or *"who converted"*, but at the impact each interaction has when present and when absent.

<Info>
  In the Markov model, we don't consider the complete journey, precisely to ensure greater quality and consistency in attribution. Today we work with a **7-day window**, which best balances volume and data accuracy.

  Very long journeys tend to have noise and dilute the relevance of touchpoints closest to conversion, which can negatively impact this model's analysis.
</Info>

## How Markov is Calculated

The Markov model at Nemu follows the **removal effect method**:

1. **We map all journeys**: both those that resulted in conversion and those that didn't.
2. **We simulate removing a channel/campaign/ad**:
   * We remove that point's presence from the journeys.
   * We calculate what the conversion rates would be without that point.
3. **We compare with the original scenario**:
   * If conversion drops, it means that channel played a fundamental role.
   * If there's no impact, it means that channel had little or no role in conversion.
4. **We proportionally assign value**: each channel's share is based on how much it actually contributes to conversion as a whole.

## Practical Example

Imagine we have the following journeys:

* User 1: Google → Meta → Conversion
* User 2: Organic → Meta → Conversion
* User 3: Google → Conversion

If we apply the removal effect to **Meta**:

* Journeys that went through Meta now "lose" that step.
* The result shows **significant drop in total conversions**.

### Conclusion

**Meta** played an essential role in the journey (even without always being the last click).

## Explaining to a Child

Think of putting together a **puzzle**. Each piece helps form the final picture. The Markov model tests what happens if you remove one piece: does the picture still appear or is it incomplete? This way we know which pieces are really important to complete the drawing.

## Advantages of the Markov Model

* Measures real impact of each channel.
* Considers the entire user path, not just beginning or end.
* Helps identify top-of-funnel channels that contribute invisibly in other models.

<Note>
  **Exclusivity:** Very few platforms offer this level of analysis in an accessible way.
</Note>