Hello and thank You for taking time to view our job description.
We are an eCommerce store engaged in pay per click advertising who has determined in order to optimize our advertising we are fundamentally facing a Knapsack problem which will require a genuine computer scientist with excellent algorithm expertise and a solid foundation of mathematics to conquer the problem at hand.
To give an idea of our problem:
We have numerous groups of ads, which we call ad groups, with each of these ads having associated search terms, and a goal cost per conversion we need to stay under. The cost of conversion is calculated by multiplying the bid * all clicks received, divided by the total quantity of conversions generated.
We have a database of every possible search term which will be eligible for any given ad group.
Each search term will have it's own quantity of clicks, a quantity of conversions it will generate, and a minimum bid required to activate the search term.
There will be many search terms per ad group, and the minimum bid required, as well as clicks and quantity of conversions will vary widely within the group of search terms, and we can only set 2 bids for each group of search terms.
Whatever bid you determine will be applied to all search terms in the set.
If your bid is too low for a group of terms, you may not activate well performing terms, which you will then miss possible conversions. If your bid is too high you may be spending much more for a search term which you could had gotten at a lower cost, and this may push your average cost per conversion higher, exceeding the goal cost per conversion of the ad group.
From each group of search terms we want to segment into a set which will have the first bid applied, a set which will have the second bid applied, and a set which we will not advertise at all.
The total number of either of the segments of search terms within the group which we want to bid on + the quantity of terms we are opting to not advertise at all must be < n quantity.
An optimal set of terms will combine the primary set of terms, the sub set of terms, and the negative set of terms to allow for maximum possible quantity of conversions for that ad group, while having average cost per conversion <= that ad group's goal CPA.
We will be able to provide a more in-depth description of the problem and guideline requirements, as well as the data set and some sample examples we have upon interview.
Ideally this project will produce a local ran application in which we can feed a CSV file of data, and the application will provide a resulting CSV file of the results.
If can be written in Java, this will be a plus.
Thank You so much for your time, and we look forward to hearing from you