Gavin Cheng

SI 618 Individual Project Proposal

Gavin Cheng (gavinch)

1. Project Motivation

Air France, born in 1933, is one of most successful airline company in the world. From years ago, Air France began to try search engine marketing (SEM). However, the result is not very satisfactory. In this project, I want to analyze the SEM data to get some recommendations to improve Air France's online marketing efficiency.

2. Data Source Summary

Air France got its online advertising data from DoubleClick, an Internet ad services company that help companies to do online advertisement. I will buy the dataset and company background information from Kellogg School of Management, Northwestern University. The data is in xlsx or csv format. It has all kinds of variables related to SEM including Publisher Name (text), Keyword (text), Bid Strategy (text), Impression (numeric), Clicks (numeric), Total Click Charges (text), Gross Revenue (text), Click through Rate (numeric) and so on.

3. Exploratory Questions & Research Methods

* What search engines delivered the most value to Air France?

Compute the Return on Advertising Dollar Spent (ROA) by publish. Then get the mean for each publisher using ddply. Visualize the data in bar graph to compare ROA. Another way to do this is to use R to get the correlation between expense on each publisher and ROA. Then get some insights from the correlation result and graph.

* How should keywords be changed to increase overall value gained from investment with a search engine publisher?

Firstly, I can test whether the data that geographically based campaigns is sensitive to certain cities. This certainly should be implemented one step deeper, on a publisher-basis. I can draw bar graph to compare ROA or revenue.

* What are the most important key performance indicators (KPIs)?

Use correlation analysis. Get which KPIs are the most important ones from correlation results and graphs. Try Factor Analysis: use factal() function to produce maximum likelihood factor analysis.

* What will increase tick sales and improve return of advertisement?

Combine the data of probability of booking, cost per click and total cost. Then use Bubble Chart to some insights and recommendation.