question on Budget Allocator Max_response_allocator

[tgtod002]

Max Response Allocator

Question on the current logic for Max_response in budget allocator.

It looks like the current logic when computing max_response_expected is to compare histSpend with Expected Spend. What this does is to compare a previous share allocation at a different budget (histSpend) vs. a new budget (ExpectedSpend).

Shouldn’t the comparision be ExpectedSpend(with old share allocation) to ExpectedSpend (with optimized Share allocation)?

For example. If we have a historical budget of $1m and historically, it was split Media 1(70%) and Media2(30%). If we want to optimize on a new budget (let's say $2M). Should we be comparing the response of new budget with Media 1($2M*.70) and Media2($2M .3) vs. an optimized share allocation of Media 1 ($2MOptmizedSharemedia1) and Media2($2M*optimizedsharemedia2). Then compare/show the expected responses for both.

I think currently initSpendUnit = histSpendUnit which compares historical spend vs. new Expected spend.

Hope that makes sense.

Robyn version 3.7

[gufengzhou]

Hi, we actually compare the optimised spend share with the initial spend share, as well as the optimised response (or allocation) share with initial. I'm not 100% sure about your question.

You can also check the script in R/R/allocator.R

[gufengzhou]

Please reopen if this issue reoccurs.