2. - Air France uses SEM as a part
of their marketing strategy
- They employ different
strategies across different
search engines
Current
State - Air France has an optimal
search engine marketing
strategy
- They have minimized their
spend and maximize their
revenue
Desired
Future
State
- What are the metrics that we need to
be looking at?
- How does strategy differences impact
each of these metrics?
Problem Definition
3. Executive Summary
1. Returns on advertising differ by publisher as much as 3X
2. Similar campaigns can perform differently by a factor of as high as 10X
3. 904 identified low-performing * zero revenue generating keywords resulted in
spends of $75,765
i. The top 10 of these contribute to 30% of the spending
Key
Findings
1. Increase investment in Kayak, Yahoo and MSN
2. Opportunities exist for extending usage of certain keywords across publishers and
related campaigns
3. Revisit strategy for the low-performing keywords
Next
Steps
*low-performing- low click through rate
5. Overture US/Global
Recommendations
• Re-consider creative
• Examine offerings most
commonly found via Overture
An increase in Bookings per impression from 0.002% to 0.004%
would increase Return on Advertising from about 357% to over 830%.
$0.0
$0.5
$1.0
$1.5
$2.0
$2.5
-0.02% 0.00% 0.02% 0.04% 0.06% 0.08% 0.10% 0.12%
Cost/Click
Probability of Booking
Overture
(Global/US)
6. Google US/Global
Recommendation
• Adjust bidding strategies to
lower costs
• Check for under-performing
keywords
Decreasing the average cost per click for Google by just 10 cents
could reduce costs by over $13,000.
$0.0
$0.5
$1.0
$1.5
$2.0
$2.5
-0.02% 0.00% 0.02% 0.04% 0.06% 0.08% 0.10% 0.12%
Cost/Click
Probability of Booking
Google
(Global/US)
7. Yahoo and MSN US/Global
Recommendations
• Increase investment
• Look for successful keywords to
incorporate with Google and
Overture
Overall return on investment for these three publishers is
currently over 1,000%. Increasing investment is certainly worthwhile!
$0.0
$0.5
$1.0
$1.5
$2.0
$2.5
-0.02% 0.00% 0.02% 0.04% 0.06% 0.08% 0.10% 0.12%
Cost/Click
Probability of Booking
MSN -US
Yahoo-
US
MSN -
Global
8. Publisher-Level Insights:
The Best Segments Recommendations
Attributes Yahoo Google & Overture
Average Position [1.2,1.3) >=1.3 <1.2 [1.2,1.3)
Bid Amount - - <8.1 -
Average
Conversion %
9.2 7.4 7.3 9.2
Attributes Yahoo Google & Overture
Average Position <1.3 [1.3,2) <1.3
Average Click-
through %
8.6 8.4 18
Yahoo:
Target position 1 or 2
Google & Overture:
Target position 1
11. Kayak…going forward
• Volume: Does Kayak have the
reach to support these returns
at higher levels of investment?
• Data: Kayak’s data is not well-
integrated, so analyzing success
and failure may be more
difficult
Concerns
• Gradually increase investment:
Make steady increases without
over-committing
• Continually monitor returns: Watch
for under-performing campaigns
Recommendations
12. KPI analyses at a campaign level
Net Revenue | CTR | ROA | CPB
Net Revenue | *CTR | *ROA | *CPB1
3
2
4
Net Revenue | CTR | ROA | CPB
Net Revenue | | CTR | ROA | CPB
Avg. Position
% Exact
Avg. Search Engine Bid
% Branded Keywords
1
2
3
4
*CTR- Click Through Rate
*ROA – Return on Advertising
*CPB- Cost per Booking
13. Evaluating Keywords
How can we evaluate existing keywords?
• Primary Goals
• Clicks
• Conversions
• Basis for Evaluation
• Clicks Per Impression
• Total Conversions
18. Evaluating Keywords
High Click-Through
Rate
Low Click-Through
Rate
Unsuccessful:
Let’s take a closer look…
Keywords with 150 or more
impressions (2,372 records)
1+ Conversions 0 Conversions
?
✔
20. Keyword Click Charges Average Position Search Engine Bid
Average Cost per
Click
Flight $4,247 1.9 $8.75 $3.83
travel to france $3,296 2.3 $6.25* $5.55
greece flights $2,667 2.9 $6.25 $3.41
europe flights $2,602 1.8 $6.25 $3.01
italy travel $2,566 2.5 $6.25 $4.12
romantic vacation $2,510 2.7 $6.25 $1.67
turkey travel $2,297 2.5 $6.25 $3.50
spain travel $2,027 2.2 $6.25 $3.94
athens hotel $1,545 2.1 $3.28 $2.55
travel to paris $1,459 1.9 $5.36* $1.47
AirFrance Average $2,521 1.93 $5.72 $1.89
Most Expensive Keywords in the X Group
*Estimated data included
21. Keyword Click Charges Average Position Search Engine Bid
Average Cost per
Click
Flight $4,247 1.9 $8.75 $3.83
travel to france $3,296 2.3 $6.25* $5.55
greece flights $2,667 2.9 $6.25 $3.41
europe flights $2,602 1.8 $6.25 $3.01
italy travel $2,566 2.5 $6.25 $4.12
romantic vacation $2,510 2.7 $6.25 $1.67
turkey travel $2,297 2.5 $6.25 $3.50
spain travel $2,027 2.2 $6.25 $3.94
athens hotel $1,545 2.1 $3.28 $2.55
travel to paris $1,459 1.9 $5.36* $1.47
AirFrance Average $2,521 1.93 $5.72 $1.89
Most Expensive Keywords in the X Group
*Estimated data included
23. Tested Campaigns
Tested Publishers
Opportunity
Opportunity
Air France Branded
Air France Global Campaign
French Destinations
Geo Targeted Boston
Geo Targeted DC
Geo Targeted Detroit
Geo Targeted Los Angeles
Geo Targeted New York
Geo Targeted Philadelphia
Google Yearlong 2006
Paris & France Terms
Western Europe Destinations
Air France Brand & French Destinations
Geo Targeted Houston
Geo Targeted Chicago
Geo Targeted San Francisco
Geo Targeted Seattle
Geo Targeted Atlanta
Geo Targeted Cincinnati
Google - Global
Google - US
Yahoo - US
Overture - Global
MSN - Global
MSN - US
Overture - US
Opportunity for key Keywords
24. Conclusion
1. Opportunity to increase investment in Kayak, Yahoo and MSN
2. Revisit strategy for the identified low performing keywords – low click through rate, 0 bookings and high cost
3. Revisit strategy for the identified high performing keywords – high click through rate, low # campaigns
We wanted to make sure that we evaluated keywords which had had a chance to perform. This resulted in a tension between wanting to evaluate with confidence, and wanting to evaluate a significant portion of the keywords. We chose to evaluate those that had met a minimum threshold of impressions. Specifically 150, which is about where we’d expect a typical keyword to have been clicked on a few times.
The dividing like for Click-Through rate is about 3.3% which is based on the average observed rate within our records.
We wanted to make sure that we evaluated keywords which had had a chance to perform. This resulted in a tension between wanting to evaluate with confidence, and wanting to evaluate a significant portion of the keywords. We chose to evaluate those that had met a minimum threshold of impressions. Specifically 150, which is about where we’d expect a typical keyword to have been clicked on a few times.
The dividing like for Click-Through rate is about 3.3% which is based on the average observed rate within our records.
We wanted to make sure that we evaluated keywords which had had a chance to perform. This resulted in a tension between wanting to evaluate with confidence, and wanting to evaluate a significant portion of the keywords. We chose to evaluate those that had met a minimum threshold of impressions. Specifically 150, which is about where we’d expect a typical keyword to have been clicked on a few times.
The dividing like for Click-Through rate is about 3.3% which is based on the average observed rate within our records.
We wanted to make sure that we evaluated keywords which had had a chance to perform. This resulted in a tension between wanting to evaluate with confidence, and wanting to evaluate a significant portion of the keywords. We chose to evaluate those that had met a minimum threshold of impressions. Specifically 150, which is about where we’d expect a typical keyword to have been clicked on a few times.
The dividing like for Click-Through rate is about 3.3% which is based on the average observed rate within our records.
We wanted to make sure that we evaluated keywords which had had a chance to perform. This resulted in a tension between wanting to evaluate with confidence, and wanting to evaluate a significant portion of the keywords. We chose to evaluate those that had met a minimum threshold of impressions. Specifically 150, which is about where we’d expect a typical keyword to have been clicked on a few times.
The dividing like for Click-Through rate is about 3.3% which is based on the average observed rate within our records.
The important takeaway from this slide is not the money spent, but the distribution: just a few of the keywords cost us a lot in click charges, with many of rest being very inexpensive. Remember, this is the group of keywords which have not yet produced a conversion. From here, we’ll look at the top 10 most expensive keywords in this category.
Engine representation: Google = 12, Yahoo = 10, Overture 7, MSN = 2
Overall average Bid: $5.72 (excludes missing data), overall average of avg. cost per click: $1.89
Each of these columns is a measure of the amount spent on the keywords. Click charges is the sum total. With average position, a lower number represents spending more relative to other groups bidding on the same keyword. Search engine bid represents the maximum we are willing to pay for a click on the keyword, where our overall average is $5.72. The average cost per click is a representation of how much clicks on these keywords tend to cost us, where our overall average is $1.89.
These keywords are expensive: over $25,000 was spent on them alone with no revenue generated. For this group of keywords, we recommend either seriously reducing or possibly even eliminating support.
Engine representation: Google = 12, Yahoo = 10, Overture 7, MSN = 2
Overall average Bid: $5.72 (excludes missing data), overall average of avg. cost per click: $1.89
Each of these columns is a measure of the amount spent on the keywords. Click charges is the sum total. With average position, a lower number represents spending more relative to other groups bidding on the same keyword. Search engine bid represents the maximum we are willing to pay for a click on the keyword, where our overall average is $5.72. The average cost per click is a representation of how much clicks on these keywords tend to cost us, where our overall average is $1.89.
These keywords are expensive: over $25,000 was spent on them alone with no revenue generated. For this group of keywords, we recommend either seriously reducing or possibly even eliminating support.