Ad Spend Optimizer
In advertising, it's of interest just how much to spend on an ad before the investment is no longer worthwhile. In this demo I attempt to show how D3 can connect to both a regression model, as well as to a numerical optimization (don't worry, users don't have to be aware that it's actually doing so), to help determine this optimal amount.
Below you'll see three curved lines; these are the result of running a regression on the relationship between a hypothetical company's advertising spending (the 'X' variable), and the revenue from this company's customers who go on to make a purchase (the 'Y'), within three customer segments. The lines appear curved because of how these returns on this spending diminish: the greater the number of impressions already purchased for a segment, the lesser the impact in purchasing a new one.
Try these things with this demo:
- Change the "budget # impressions" -- to show the expected resulting revenue or profit
- Click "optimize budget" -- to discover the profit-maximizing investment total across segments
(To determine this amount mathematically, I'm having D3 connect to an independent optimization routine, which sets the segments' marginal ROIs to zero, the first-order condition)
- Change to "profit view" -- to see how the curves peak at their points of maximal profit
- Change the "cost per impression" -- and then click "optimize budget" again to see how these maximal profit points change in response
- Drag a red dot -- to intervene manually into a specific segment's budget; the visualization re-optimizes the other points using what budget remains
- Click "optimize budget allocation" -- to reset the segment-by-segment budgets to optimal, subject to the total overall budget constraint
(Mathematically here, I'm having D3 solve a system of equations to determine the investment points at which the marginal ROIs become equal)