TEMPORAL BUDGET OPTIMIZATION IN ONLINE ADVERTISING
A method for temporal budget optimization in online advertising, the method comprising using at least one hardware processor for: receiving a user selection of a time period in the future; forecasting, based on historical data associated with an online ad entity, a future return on investment (ROI) function of the online ad entity; receiving a user selection of a point on the ROI function, thereby setting a budget for the time period; and during the time period: (a) tracking a spending of the budget, to determine a remaining budget, (b) periodically updating the future ROI function based on newly-accumulated historical data associated with the online ad entity, and (c) periodically adjusting, in an online advertising platform, a spending pace of the remaining budget, wherein the adjusting is based on the updated future ROI function.
1 . A method for temporal budget optimization in online advertising, the method comprising using at least one hardware processor for:
receiving a user selection of a time period in the future;
forecasting, based on historical data associated with an online ad entity, a future return on investment (ROI) function of the online ad entity;
receiving a user selection of a point on the ROI function, thereby setting a budget for the time period; and
during the time period:
(a) tracking a spending of the budget, to determine a remaining budget,
(b) periodically updating the future ROI function based on newly-accumulated historical data associated with the online ad entity, and
(c) periodically adjusting, in an online advertising platform, a spending pace of the remaining budget, wherein the adjusting is based on the updated future ROI function.
2 . The method according to claim 1 , wherein the forecasting of the future ROI function of the online ad entity comprises:
fetching the historical data associated with the online ad entity, wherein the historical data comprises a historical cost time-series and a historical revenue time-series;
correlating the historical revenue time-series to the historical cost time-series, to produce correlated historical data; and
applying a nonlinear curve fitting algorithm to the correlated historical data, to produce a nonlinear function approximately descriptive of the correlated historical data, wherein, in the nonlinear function, revenue is a function of cost, and wherein the nonlinear function is the future ROI function of the online ad entity.
3 . The method according to claim 2 , wherein the periodically updating of the future ROI function comprises re-executing the fetching, the correlating and the applying.
4 . The method according to claim 2 , wherein the applying of the nonlinear curve fitting algorithm is with an instruction to produce the nonlinear function with a functional form selected from the group consisting of: a polynomial form, a logarithmic form, an exponential form, a trigonometric form and a hyperbolic form.
5 . The method according to claim 2 , further comprising using the at least one hardware processor for computing error bounds of the nonlinear function, based on residuals of the application of the nonlinear curve fitting algorithm to the correlated historical data.
6 . The method according to claim 5 , wherein the nonlinear function has a first functional form, and wherein the method further comprises using the at least one hardware processor for:
applying the nonlinear curve fitting algorithm to the correlated historical data, wherein the applying is with an instruction to produce a different nonlinear function having a second functional form and being approximately descriptive of the correlated historical data, wherein the first functional form is different from the second functional form;
computing error bounds of the different nonlinear function based on residuals of the application of the nonlinear curve fitting algorithm to the correlated historical data; and
comparing the error bounds of the nonlinear function and the error bounds of the different nonlinear function, and indicating which one of the nonlinear function and the different nonlinear function has the smallest error bounds.
7 . The method according to claim 1 , wherein the time period is selected from the group consisting of: up to a week, up to multiple weeks, up to a month and up to multiple months.
8 . The method according to claim 1 , further comprising using the at least one hardware processor for receiving a schedule of one or more future business events expected to occur during the time period, wherein the adjusting is further based on the schedule.
9 . The method according to claim 8 , wherein the receiving of the schedule comprises receiving a business prediction as to each of the one or more future business events.
10 . The method according to claim 1 , wherein the adjusting of the spending pace of the budget comprises adjusting bids associated with the online ad entity.
11 . The method according to claim 1 , wherein the online ad entity is selected from the group consisting of: an individual ad, a group of ads, a campaign and a set of campaigns.
12 . A computer program product for temporal budget optimization in online advertising, the computer program product comprising a non-transitory computer-readable storage medium having program code embodied therewith, the program code executable by at least one hardware processor for:
receiving a user selection of a time period in the future;
forecasting, based on historical data associated with an online ad entity, a future return on investment (ROI) function of the online ad entity;
receiving a user selection of a point on the ROI function, thereby setting a budget for the time period; and
during the time period:
(a) tracking a spending of the budget, to determine a remaining budget,
(b) periodically updating the future ROI function based on newly-accumulated historical data associated with the online ad entity, and
(c) periodically adjusting, in an online advertising platform, a spending pace of the remaining budget, wherein the adjusting is based on the updated future ROI function.
13 . The computer program product according to claim 12 , wherein the forecasting of the future ROI function of the online ad entity comprises:
fetching the historical data associated with the online ad entity, wherein the historical data comprises a historical cost time-series and a historical revenue time-series;
correlating the historical revenue time-series to the historical cost time-series, to produce correlated historical data; and
applying a nonlinear curve fitting algorithm to the correlated historical data, to produce a nonlinear function approximately descriptive of the correlated historical data, wherein, in the nonlinear function, revenue is a function of cost, and wherein the nonlinear function is the future ROI function of the online ad entity.
14 . The computer program product according to claim 13 , wherein the periodically updating of the future ROI function comprises re-executing the fetching, the correlating and the applying.
15 . The computer program product according to claim 13 , wherein the applying of the nonlinear curve fitting algorithm is with an instruction to produce the nonlinear function with a functional form selected from the group consisting of: a polynomial form, a logarithmic form, an exponential form, a trigonometric form and a hyperbolic form.
16 . The computer program product according to claim 13 , wherein the program code is further executable by the at least one hardware processor for computing error bounds of the nonlinear function, based on residuals of the application of the nonlinear curve fitting algorithm to the correlated historical data.
17 . The computer program product according to claim 16 , wherein the nonlinear function has a first functional form, and wherein the program code is further executable by the at least one hardware processor for:
applying the nonlinear curve fitting algorithm to the correlated historical data, wherein the applying is with an instruction to produce a different nonlinear function having a second functional form and being approximately descriptive of the correlated historical data, wherein the first functional form is different from the second functional form;
computing error bounds of the different nonlinear function based on residuals of the application of the nonlinear curve fitting algorithm to the correlated historical data; and
comparing the error bounds of the nonlinear function and the error bounds of the different nonlinear function, and indicating which one of the nonlinear function and the different nonlinear function has the smallest error bounds.
18 . The computer program product according to claim 12 , wherein the time period is selected from the group consisting of: up to a week, up to multiple weeks, up to a month and up to multiple months.
19 . The computer program product according to claim 12 , wherein the program code is further executable by the at least one hardware processor for receiving a schedule of one or more future business events expected to occur during the time period, wherein the adjusting is further based on the schedule.
20 . The computer program product according to claim 19 , wherein the receiving of the schedule comprises receiving a business prediction as to each of the one or more future business events.
21 . The computer program product according to claim 12 , wherein the adjusting of the spending pace of the budget comprises adjusting bids associated with the online ad entity.
22 . The computer program product according to claim 12 , wherein the online ad entity is selected from the group consisting of: an individual ad, a group of ads, a campaign and a set of campaigns.
23 . A method comprising using at least one hardware processor for:
monitoring a spending of an advertising budget of an online ad entity over time; and
automatically adjusting a pace of the spending based on a periodic computation of a future ROI function of the online ad entity.