Which media dimension should be chosen for calculating MTA contributions?

Problematic

The risk of biased interpretation of performance
In omnichannel marketing strategies, the points of contact with users are numerous, distributed over time, and carried by very different activations:
- “common thread” campaigns broadcast continuously,
- one-off campaigns (launch, promotions, events),
- campaigns with significant media investments ,
- and others with little advertising pressure .
Backhand Index Pointing Right Challenge: Without the right level of aggregation for calculating contributions in the MTA model, performance analysis can become biased.
Examples
- A common thread campaign (Meta Ads always active) will automatically capture more points of contact… and therefore more contribution.
- A one-off campaign (e.g. TikTok for a special operation) may seem to underperform , even though it is simply present in fewer journeys than the campaign that is inherently red thread (on the day of its launch, it has no history).
- A Google campaign with large budgets will be able to capture a lot of contributions thanks to the volumes generated, to the detriment of a less exposed but efficient advertising agency .

Objective

Make the attribution model aligned with management objectives
By choosing the attribution dimension ( channel , medium or campaign ), the company can:
- Correct exposure-related biases (frequency, duration, budget).
- Maintain a strategic (channel) or operational (campaign) vision.
- Avoid drawing erroneous conclusions about the effectiveness of a weakly diffused activation with a high marginal ROI.


Configuring Attribution Dimensions

The Eulerian platform allows the Client to configure the dimension on which marketing contributions are calculated each night.
3 dimensions are available: channel , support or campaign .
This choice influences the reading of the performance of each marketing lever.


Situational analysis with an example of a user journey

For the sake of simplicity, we will consider the MTA algorithm used to calculate the contributions in this example to be a linear algorithm (each contact point receives an equivalent contribution).
Let's take as an example a marketing history that contains 4 touchpoints before a sale.
    Click Meta Ads
    Click Meta Ads
    Click TikTok Ads
    Google Ads Click
Here are the attributes of each touchpoint:
Channel
Publisher
Campaign
Social
Meta
Campaign1
Social
Meta
Campaign1
Social
TikTok
Campaign2
SEA
Google
Campaign3


Choice of attribution dimension

1. By Channel

In this case, contributions are measured according to the channel dimensions.
  • Result :
  • Social Channel: present 3x (Meta x2 + TikTok x1)
  • SEA Channel : present 1x (Google x1)
  • Distribution (e.g. model with linear distribution):
  • Social Channel: 3/4 of the contribution (75%)
  • SEA Canal: 1/4 of the contribution (25%)


2. By Support

Here, contributions are measured by medium :
  • Result : Three rules in the report:
  • Meta Support: Present 2x
  • TikTok Support: present 1x
  • Google Support: present 1x
  • Distribution :
  • Meta Support: 2/4 (50%)
  • TikTok Support: 1/4 (25%)
  • Google Support: 1/4 (25%)


3. By Campaign

Here, contributions are measured per unique campaign .
  • Result : Three distinct campaigns:
  • Campaign 1 (Meta): Presents 2x
  • Campaign 2 (TikTok): presents 1x
  • Campaign 3 (Google): presents 1x
  • Distribution :
  • Campaign 1: 2/4 (50%)
  • Campaign 2: 1/4 (25%)
  • Campaign 3: 1/4 (25%)


Check Mark Button Recommendations


I want to avoid one-off campaigns being overwhelmed by recurring campaigns.
If the one-off campaigns belong to the same publisher, and this publisher is "always-on", then we recommend configuring the MTA model on the publisher dimension.
If the one-off campaigns do not belong to the same publisher, but belong to the same channel, and this channel is "always-on", then we recommend configuring the MTA model on the channel dimension .