On my website, there are fully-priced items and discounted items (discounted items can be fully-priced also, so this is a time dependent analysis). I need to source discounted items from the existing fully-priced items, but I am unsure which items to source and provide as a discounted item in order to increase sales of discounted items. I also offer multiple items that are the same thing (e.g. 5 types of 'Bose Headphones') so differentiating which should go on sale is important.
My first attempt was to look at metrics such as:
- What items sold the most as discounted vs normal (I am assuming there is an incremental value in having the item discounted).
- What items had the most views (I am assuming a correlation between high views and things people want).
- What items had the work 'discount, deal or cheap' with the search (I guessed this indicated they wanted it cheaper).
- Who is likely to visit over the next week (based on previous visits) and try to customise the discounts for them.
and conduct a multiple regression analysis. However, this doesn't really tell me what products are "better" to source as a deal in order to maximise revenue.
Another approach is to use a linear programming method with the objective function as max(revenue) but then I am still unsure how this helps select the products themselves.
Once this is solved, the next piece would be to understand the propensity of purchase as a discounted product vs not, but this is more straightforward when you have 2 cohorts, which is what I am trying to figure out here.
Does anyone have any ideas on how else this problem can be approached?