> ## Documentation Index
> Fetch the complete documentation index at: https://docs.dema.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Attribution
> Understand how Dema attributes conversions to specific channels or campaigns. Learn about our supported models (last-click, linear, ML-based multi-touch, ad platform-reported, and causal factor attribution) and how to interpret the results.
## Overview
Dema's **attribution** engine ensures that each conversion (order) is assigned to the right marketing source. By mapping UTM parameters, user sessions, and integrated cost data, Dema helps you see **which channels** and **which campaigns** drive profitable conversions.
Attribution is crucial for identifying where to allocate your marketing budget, comparing performance across channels, and measuring return on ad spend (ROAS) at a more granular level.
***
## Models we support
Dema offers **five** primary attribution models. You can select one that aligns with your business logic or compare models side by side for a complete view of channel impact.
### 1. Last-click
* **How it works**
The **last marketing channel** a user interacted with before placing an order receives 100% credit.
* **Pros & cons**
* **Pros**: straightforward and aligns with many standard analytics platforms.
* **Cons**: may undervalue earlier touchpoints (awareness campaigns).
### 2. Linear
* **How it works**
All channels in the user's path to purchase share conversion credit **equally**.
* **Pros & cons**
* **Pros**: recognizes every touchpoint's contribution.
* **Cons**: can over-credit minor interactions that barely influenced the final purchase.
### 3. ML-based multi-touch (MTA)
* **How it works**
Our algorithmic model evaluates each user's **full journey** and learns how much each touchpoint contributes to the conversion. Rather than assigning fixed percentages (e.g., 30% first-click), it adapts dynamically using historical data.
* **Pros & cons**
* **Pros**: captures complex journeys with more precision, often uncovering undervalued channels.
* **Cons**: requires more data and can be less intuitive at a glance.
### 4. Ad platform-reported
* **How it works**
Uses each ad platform's **native** settings (Meta, Google Ads, TikTok, etc.). Their conversion windows (e.g., 7-day click, 1-day view) define attribution.
* **Pros & cons**
* **Pros**: matches what you see in ad platform dashboards.
* **Cons**: can differ across platforms, making cross-channel comparisons less consistent.
### 5. Dema causal factor attribution (CFA)
* **How it works**
Adjusts MTA or ad platform-reported values using **causal multipliers** derived from [incrementality experiments](/guides/incrementality-testing/how-it-works) and platform-wide benchmarks. Rather than relying solely on observed correlations, CFA calibrates attribution to reflect the true incremental impact of each channel. See the full [Causal factor attribution](/guides/causal-factor-attribution/how-it-works) guide for details.
* **Pros & cons**
* **Pros**: reflects true incremental impact grounded in experimental evidence; works even without your own experiments thanks to platform-wide benchmarks.
* **Cons**: accuracy improves with more experiments; requires configuration of calibration rules and multipliers.
***
## How Dema's ML-based multi-touch attribution works
Dema's **ML-based Multi-Touch Attribution (MTA)** goes beyond simple models like last-click or linear. It uses machine learning to evaluate the entire customer journey and assigns conversion credit based on the actual contribution of each touchpoint. This approach gives you a more accurate understanding of which channels and campaigns drive results.
#### Key features of ML-based MTA
1. **Analyzes the full customer journey**
The model considers every interaction a customer has with your brand—clicks, views, and other engagements—and evaluates their impact on the conversion.
2. **Learns what drives conversions**
Instead of assigning fixed percentages to specific touchpoints, Dema's algorithm uses historical data to learn how much influence each interaction had. It adapts as customer behavior changes over time.
3. **Context matters**
The model accounts for important factors like:
* The order of interactions (e.g., first vs. last touch).
* The time between interactions.
* The type of channel (e.g., paid ads, organic search, email).
4. **Dynamic credit allocation**
Conversion credit is assigned based on the importance of each touchpoint. For example:
* A paid social ad might generate interest and receive 30% credit.
* A remarketing email might nurture the customer further and get 20%.
* The final paid search click that led to the purchase might earn 50%.
***
## Choosing the right model
Each model offers a different perspective on **which channel** or campaign deserves conversion credit. Depending on your funnel length and goals, you may prefer a single model or use **several** at once.
If customers buy soon after clicking an ad, last-click may suffice. For longer journeys, ML-based MTA often reveals hidden touchpoint value.
Running multiple models shows how credit shifts. Channels that seem weak under last-click may appear stronger in linear or ML-based MTA.
Some teams or agencies rely on ad platform-reported conversions. Decide which model is your single source of truth for ROI.
***
## How it works in Dema
1. **UTM & session tracking**
* Dema uses `utm_source`, `utm_medium`, `utm_campaign`, etc., at session start.
* These persist until the user either checks out or the session expires.
2. **Order association**
* When a user converts, Dema identifies the relevant session(s) and divides credit based on your chosen model.
3. **Real-time updates**
* Dema processes data in near real-time, so you can see updates within minutes.
* Overnight jobs reconcile missed events or offline data to ensure accuracy.
4. **Adjusting models**
* You can switch models in the reporting interface.
* Historical data is recalculated, letting you compare how each model allocates credit over time.
**Conversion window**\
By default, Dema uses a 30-day click window for multi-touch attribution.
***
## Going beyond attribution with causal calibration
Attribution models tell you **which channels are involved** in conversions. But they can't tell you whether those conversions would have happened without the marketing spend. Dema's [Causal factor attribution](/guides/causal-factor-attribution/how-it-works) takes your MTA values and adjusts them using evidence from incrementality experiments, or platform-wide benchmarked incremental factors, to reflect the **true causal impact** of each channel.
Learn more about how to calibrate your attribution with causal evidence in the [Causal factor attribution](/guides/causal-factor-attribution/how-it-works) section.
***
## Key takeaways
1. **No single model** works for everyone - match your choice to funnel complexity.
2. **ML-based MTA** can reveal hidden value in mid or early funnel, but it's more complex.
3. **Ad platform data** may differ from Dema's logic; use both for a broader picture.
4. **Causal factor attribution** can adjust any attribution model's values using real experimental evidence to better reflect true incremental impact.