scope3data / methodology

Measurement methodology for advertising emissions
https://methodology.scope3.com
Apache License 2.0
29 stars 10 forks source link

Traffic shapping accounting for properties emissions #210

Open maelmrgt opened 1 month ago

maelmrgt commented 1 month ago

Hello everyone,

I would love to contribute to your repository to implement a new feature allowing to account for traffic shaping policy during carbon emissions computation, on a publisher side. I would appreciate your thoughts on the implementations proposed below.

Here is my use case. At Greenbids, we want to be able, for our partner publishers, to account for whether traffic shaping is being deployed on their properties. This distinction is important because impressions on properties with traffic shaping should result in lower carbon emissions.

This leads me to consider two options:

First solution assumes that corporate emissions are tied to the ad selection process, which I am not entirely sure about. However, in solution 2, determining or setting a value for ad_selection_gco2_per_impr (which should depend on the type of property) can be challenging.

Thanks for your time and this repository. I appreciate how well it articulates and models the advertising ecosystem. Maël

bokelley commented 1 month ago

Hi Maël - agree that this is important. If you look at https://methodology.scope3.com/calculations#ad-platform you will see that there is a field distribution_rate_by_bidder_by_country which is where traffic shaping is incorporated. In the Scope3 implementation, we have an API where publishers or their traffic shaping vendor can upload daily data about how bid requests are being shaped: https://docs.scope3.com/docs/atp-traffic-shaping

maelmrgt commented 1 month ago

Hi Brian,

Thank you for your prompt response and detailed explanation. From what I understand, this functionality is incorporated into the Scope3 implementation, but it is not yet available in the open-source repository.

To elaborate on my issue: as a publisher, I have two identical properties (same number of visits per month, same pages per visit, same number of SSPs connected ..). The only difference is that traffic shaping is deployed on one property, resulting in fewer ad calls sent (regardless of the bidder), making it a greener property.

When modeling these two properties using the open-source repository, I would, for example, add a facts parameter like traffic_shaping_coefficient: 0.8 for one of the properties. However, I am unsure how to account for this difference when modeling them using model_publisher_emissions.

Please let me know if you need further details or if my explanation is unclear. Maël

bokelley commented 1 month ago

Maël - feel free to submit a PR to update the open source code for this!