diff --git a/website/docs/features.mdx b/website/docs/features.mdx index e3d4e59c5..fc7c04009 100644 --- a/website/docs/features.mdx +++ b/website/docs/features.mdx @@ -456,6 +456,17 @@ following deepdive articles: - **"The convergence of marginal ROAS in the budget allocation in Robyn"**: [here](https://medium.com/@gufengzhou/the-convergence-of-marginal-roas-in-the-budget-allocation-in-robyn-5d407aebf021) - **"Hitting ROAS target using Robyn’s budget allocator"**: [here](https://medium.com/@gufengzhou/hitting-roas-target-using-robyns-budget-allocator-274ace3add4f) +--- +## The reach & frequency allocator (prototype) + +Reach and frequency planning is conducted frequently by advertisers and media agencies. To further harvest the power of **[Project Halo](https://wfanet.org/leadership/cross-media-measurement)** and provide more actionability for R&F data, we're experimenting on a new feature "reach and frequency allocator". It answers the question “what’s the optimal combination of reach & frequency” on a channel given a total budget and average CPM. It consumes the saturation information from R&F data and uses nonlinear optimization to find the optimal point with highest response. + +Decription of the graphic below: Given 100k budget and 6$ CPM, as well as the estimated saturation curve for reach and frequency separately using Halo data, the optimum R&F combination is 5.67M reach x 2.94 frequency, with the maximum response (sales) of 317.8k$. + +ModelResults1 chart + +It's currently a prototype and not yet available. An MVP is expected in 2025. + --- ## Model refresh diff --git a/website/static/img/rnf_allocator_prototype.png b/website/static/img/rnf_allocator_prototype.png new file mode 100644 index 000000000..2c15e63da Binary files /dev/null and b/website/static/img/rnf_allocator_prototype.png differ