Automated optimization of an advertisement placement engine
View Patent ↗A system and method for predicting a performance of a target advertisement placement via a simulation for optimum tuning is disclosed. A simulator receives a set of queries from a production engine, selects a subset of simulation queries from the queries by filtering, modifies query parameters as needed, transmits the simulation queries to the target engine for simulation, collects search results from the target engine, and computes a summary metric, which includes data fields retrieved from the search results and quantities calculated by using a prediction model derived from a production data warehouse. The simulator may further produce a simulation report, an indicator of performance prediction for the target engine, which comprises multiple metrics for the target engine, calculated with varying engine parameters.
1. A system for tuning and optimizing an advertisement placement engine by simulation, comprising:
a memory storing a query log containing queries received from the advertisement placement engine; and
one or more processors in communication with the memory, the one or more processors being configured to:
select a set of queries from the query log, each query having at least one keyword and at least one parameter value;
run a search simulation for each query from the set of queries using the at least one keyword and the at least one parameter value;
determine a bid value and rank for each query based on the search simulation;
compute an expected cost per click for each query based on the bid value, the rank, and a click probability provided in a click through prediction model.
2. The system of claim 1 , wherein the set of queries is selected by applying a filter for a product line on the query log.
3. The system of claim 2 , wherein the queries in the query log is streamed from a production engine into the system.
4. The system of claim 1 , wherein the one or more processors is further configured to modify the at least one parameter values before transmitting the set of queries to the advertisement placement engine.
5. The system of claim 1 , wherein the one or more processors is further configured to perform the steps of:
running the search simulation for each query from the set of queries with a second parameter value;
determining a second bid value; and
compute a second expected cost per click for each query based on the second parameter value.
6. The system of claim 1 , wherein the click probability is determined from click data generated by a production engine.
7. The system of claim 1 , wherein the one or more processors is further configured to calculate an average revenue per impression per query based on the bid value, the rank, and the click probability provided in the click through prediction model.
8. A method of tuning and optimizing an advertisement placement engine, performed by a hardware simulator, the method comprising:
selecting a set of simulation queries from a query log containing queries received from an advertisement placement engine, each query in the set of simulation queries comprising a keyword and a parameter value;
transmitting the set of simulation queries to the advertisement placement engine;
receiving a response to the transmitted set of queries from the advertisement placement engine, the response comprising a bid value and a rank for each query in the set of simulation queries based on the keyword and the parameter value; and
computing an expected cost per click for each query in the set of simulation queries based on the bid value, the rank, and a click probability determined by a click through prediction model.
9. The method of claim 8 , wherein the set of simulation queries is selected by applying a filter for a product line on the query log.
10. The method of claim 9 , wherein the set of queries is streamed from the advertisement engine into the simulator.
11. The method of claim 8 , further comprising modifying the parameter values before transmitting the set of simulation queries to the advertisement placement engine.
12. The method of claim 8 , further comprising the steps of:
transmitting a second set of simulation queries with a second parameter value to the advertisement placement engine;
receiving a second response from the advertisement placement engine; and
computing a second expected cost per click for each query in the second set of simulation queries based on the second parameter value.
13. The method of claim 8 , wherein the prediction model is generated from click data generated by a production engine.
14. The method of claim 8 , wherein the set of summary metric includes an average expected cost per click (CPC) per query and an average revenue per impression per query.
15. A non-transitory machine-readable medium comprising a plurality of machine-readable instructions which, when executed by one or more processors of a simulator, are adapted to cause the simulator to perform a method comprising:
selecting a set of simulation queries, each query in the set of simulation queries including a keyword and a parameter value;
transmitting the set of simulation queries to an advertisement placement engine for simulation;
receiving simulation results for the set of simulation queries from the advertisement placement engine, the simulation results including a bid value and a rank for each query in the set of simulation queries based on the keyword and the parameter value; and
computing an expected cost per click for each query in the set of simulation queries based on the bid value, the rank, and a click probability determined by a click through prediction model.
16. The non-transitory machine-readable medium of claim 15 , wherein the set of simulation queries is selected by filtering a set of queries for a product line.
17. The non-transitory machine-readable medium of claim 16 , wherein the set of queries is streamed from a production engine into the simulator.
18. The non-transitory machine-readable medium of claim 15 , further comprising modifying the parameter values of each query in the set of simulation queries before transmitting the simulation queries to the advertisement placement engine.
19. The non-transitory machine-readable medium of claim 15 , further comprising repeating the steps of:
transmitting the set of simulation queries;
receiving simulation results; and
computing a second expected cost per click for each query based on another parameter value.
20. The non-transitory machine-readable medium of claim 15 , wherein the prediction model is generated from click data generated by a production engine.
21. The non-transitory machine-readable medium of claim 15 , wherein the summary metric includes an average expected cost per click (CPC) per query and an average revenue per impression per query.