Advanced Paid Search Tactics

1

• Advanced Paid Search Tactics

Dennis Yu 817-913-0780 Twitter.com/dennisyu BlitzL
ocal.com
2
Paid Search Topics

• A Live Tour of an Optimized Campaign
• 7 Ways to Flush Money Down the Toilet

3
Breakdown of Metrics
CPA
CPA Cost per Acquisition CPC Cost per
Click CR Conversion Rate CTR Click-Through
Rate CPM Cost per 1000 Impressions
CPC
CR
CTR CPM
Influenced by
Influenced by
bids keywords ad targeting
offer landing page
4
Landing Page Comparison

• BEFORE

5
Landing Page Comparison

• AFTER

6
(No Transcript)
7
Campaign Optimization

8
Campaign Optimization

9
Campaign Optimization

10
CTR Trends

• Observed average CTR by position

Averaged over 7,000 data points from real
campaigns across Google, Yahoo, and MSN
11
CTR vs. Position
12
CPA vs. Volume

• First few leads are easier (cheaper) to find
• Set target CPA by value of lead
• Target CPA determines possible profitable volume

24
13
Profit Margin vs. Volume

• Given
• Increased volume increased CPA
• Constant lead value
• Profit margin lead value CPA
• Yields
• Increased volume decreased profit margin
• Conclusion
• Profit decreases for each additional lead, breaks
  even at target CPA.

14
Profit Margin vs. Volume
profit
( Profit Margin x Volume )
15
eCPM

• eCPM Effective Cost Per Thousand Impressions
• eCPM CTR x EPC
• payout (lifetime
  value), conversion rate
• ad copy, negative keywords, targeting
• Can increase by
• Bidding up EPC preferably not
• Improving CTR
• Quality score is determined by eCPM!

16
Bid Optimization
Landing Pages
Keywords
Targeting
Ad Copy
17

• 7 Ways To Flush Money Down the Toilet

18
7 Ways to Flush Money

• Tip 1 Indiscriminate brand bidding

19
7 Ways to Flush Money

• Tip 2 Start off with a zillion keywords

20
7 Ways to Flush Money

• Tip 3 Choose the most expensive software
  possible

21
7 Ways to Flush Money

• Tip 4 No conversion tracking

22
7 Ways to Flush Money

• Tip 5 Using default campaign settings

23
7 Ways to Flush Money

• Tip 6 Staying with keyword targeting

24
7 Ways to Flush Money

• Tip 7 Rely completely on the agency

25
Final Thoughts
CPA
CPA Cost per Acquisition CPC Cost per
Click CR Conversion Rate CTR Click-Through
Rate CPM Cost per 1000 Impressions
CPC
CR
Influenced by
CTR CPM
Influenced by
offer landing page
bids keywords ad targeting
26

• Thank You!

Contact us Dennis Yu Twitter.com/dennisyu 817-913
-0780 To Learn More Go to Blitzlocal.com/smx
27
Initial test promising
equifaxmortgages.com
traffic
equifax.com
leadpoint

13,421 cost
6,496 clicks
563 raw leads
334 paid leads
11,236 revenue

• WHAT WE DID
• - Basic architecture built to drive traffic from
  search engines through to conversion.
• - We ran traffic through the funnel to test the
  value of the Equifax brand and to measure volume
  and efficiency ratios with limited optimization.
• WHAT WE LEARNED
• - Able to demonstrate that we have a process to
  quickly test and learn, going from negative to
  positive margin as we break into new categories
  of traffic.
• - Simple optimization at each stage will drive
  significant scale and margin improvements.
• - We believe we can acquire 10x the traffic at a
  10 margin in the next phase.

Data through August 10, 2007
28
Optimizations for next phase and estimated impact
equifaxmortgages.com
traffic
equifax.com
leadpoint

13,421 cost
6,496 clicks
563 raw leads
334 paid leads
11,236 revenue
2.06 CPC
8.67 CR
59.3 lead quality
33.64 revenue per lead
- All metrics measurable in real time can make
adjustments on the fly. - Data-driven
optimization system. - Goal is to increase volume
at each stage and eliminate blockages in
funnel. Lets walk through each stage of the
funnel to see where our simple implementation
has left money on the table.
29
Optimizations for next phase and estimated impact
equifaxmortgages.com
traffic
equifax.com
leadpoint

13,421 cost
6,496 clicks
563 raw leads
334 paid leads
11,236 revenue
2.06 CPC
8.67 CR
59.3 lead quality
33.64 revenue per lead
Traffic optimization
Currently only buying traffic from Google, which
is the most expensive of the search engines, but
is also the best place to begin a test, because
of the volume.
Buy traffic from multiple search engines by
replicating existing campaigns across to Yahoo!,
MSN, Ask, AOL, and so forth.
We have only bought search traffic. Larger and
more lucrative channels are email and display.
Intermundo Media has relationships within these
opaque marketplaces to rent opt-in lists and run
banner ads across partner networks.
Current bid strategy is rudimentary does not
take into account differences for geography, the
term they are bidding on, the differentials
between mortgage and refinance payouts, time of
day bidding, negative keyword strategy. Example
bid lower for folks who have bad credit.
Begin multiplying out keyword portfolio and
fine-tuning campaigns so that we can bid the
optimal amount based on a number of variables.
We are just starting to break into high-volume
terms initially at negative margins, but
gradually becoming positive.
Showing the same set of ads to all searchers.
With ad copy testing, we can show ad copy of LA
Mortgage Rates to someone who searches for LA
Mortgage Rates, and show Atlanta refinance to
someone coming in on that term.
Refining ad copy to be more relevant to the user
will increase CTR, which will increase our
quality score, lower our average CPC, and
increase volume.

• ? Systematically vacuum up the nickels that the
  big guys leave on the carpet.

• ? We can get 8-10x the click volume at the same
  price and quality.

30
Optimizations for next phase and estimated impact
equifaxmortgages.com
traffic
equifax.com
leadpoint

13,421 cost
6,496 clicks
563 raw leads
334 paid leads
11,236 revenue
2.06 CPC
8.67 CR
59.3 lead quality
33.64 revenue per lead
Equifax.com site optimization
We have been hit with quality score problems on
the highest volume searches mortgage
calculators, current refinance rates, and so
forth. Because the search engines havent
conferred trust to our new pages, that traffic
has been out of reach. We are primarily on long
tail keywords.
Increase association between mortgage lead
generation site and parent Equifax.com site.
Effect will allow us to bid on high-volume
generate terms. Further, we will begin receiving
natural traffic from organic search, which is
free. Currently, 3 of our traffic is organic,
as we have just been crawled.
Users who search on todays best mortgage rates
and see our ad promoting low mortgage rates
come to our landing page, which doesnt mirror
the users search term. When users lose the
scent of search, they abandon. We have a 53
bounce rate, which could be cut 10-20 with
landing page optimization.
We currently have one landing page that we show
to everyone, regardless of what they are
searching on. Create dynamic landing pages which
insert the text of the users search query. This
is a trivial implementation. We are already
passing through the search term from our ad
campaigns.

• ? We can lower average CPC from 2.06 to 1.70
  while still growing traffic.

31
Optimizations for next phase and estimated impact
equifaxmortgages.com
traffic
equifax.com
leadpoint

13,421 cost
6,496 clicks
563 raw leads
334 paid leads
11,236 revenue
2.06 CPC
8.67 CR
59.3 lead quality
33.64 revenue per lead
EquifaxMortgages.com site optimization
Images and ad copy resonate differently to
different audiences based on geography, interest,
or state of mind.
Landing page testing allows for simple rotation
of landing pages, finding which combination of
creatives, images, and layout will drive the
highest conversion rate.
Search Engine Optimization (SEO) is driving
traffic through natural (unpaid) search results.
Shifting the mix from paid to free will lower
average CPC and increase volume.
Take advantage of Equifaxs good reputation with
search engines it is like a Ferrari with a flat
tire. Add more content to the site there is
none there now. Include a link to a mortgage
calculator. Show articles on how to determine
what you can actually afford, how to shop for the
right mortgage, and so forth.
By sending users from Equifax.com to
EquifaxMortgages.com, we put ourselves at a
disadvantage from the viewpoint of both search
engines and users. To have the lead gen pages
under Equifax.com would allow to buy more traffic
at a lower cost from engines.
Consider moving to mortgages.equifax.com or other
set of subdomains and/or folders. Implement
proper 301 redirects using mod_rewrite, so that
you dont waste link juice.

• ? We can lower our cost per raw lead from 24 to
  under 20 and increase conversion rates from 8.7
  to 10.

32
Optimizations for next phase and estimated impact
equifaxmortgages.com
traffic
equifax.com
leadpoint

13,421 cost
6,496 clicks
563 raw leads
334 paid leads
11,236 revenue
2.06 CPC
8.67 CR
59.3 lead quality
33.64 revenue per lead
Leadpoint optimization
Wasting traffic-- we are buying a mix of traffic
that includes some people that are also
interested in home equity loans, auto refinance,
student loans, and payday loans. We have already
paid for that traffic, so we should show them
those products.
Leadpoint offers these products. Home equity can
be implemented with minimal effort. This will
produce immediate incremental margin.
Lead form optimization 40 of our gross leads
are rejected due to incomplete information, no
longer being interested, not requesting contact,
and so forth.
Streamline forms where possible to increase the
probability that they will complete the process
and also do it accurately. For example,
implement automatic address checking.
The payout we receive from Leadpoint is net of a
20 commission they take. The commission is
tiered based upon volume and negotiated factors.
As we have now reached significant volume (50
leads per day is considered a large account),
we can ask that the commission be reduced a
couple points.

• ? We can decrease our rejected lead rate from
  40.7 to 35.

33
Optimizations for next phase and estimated impact
equifaxmortgages.com
traffic
equifax.com
leadpoint

13,421 cost
6,496 clicks
563 raw leads
334 paid leads
11,236 revenue
2.06 CPC
8.67 CR
59.3 lead quality
33.64 revenue per lead
payout optimization
After the user has completed the mortgage lead
form, they can be sold a series of other
products credit reports, credit protection, auto
refinance, and so forth. The data can also be
added to Equifaxs list to be resold.
We can add an incremental 5-10 margin by
promoting other offers. If we can monetize leads
better than average, then we can afford to
profitably bid higher for traffic, while still
maintaining margins.
While refinance leads are generally worth 55-60
and new home leads are worth 3-5, there are
significant variations even within these areas.
Home value, down payment, credit worthiness, and
other factors can make even a refinance lead
worth between 5 and 140.
Tweak bids based upon value of the lead. We are
currently optimizing to a CPA, which assumes all
leads are worth the same. In other words, we are
bidding to a cost per raw lead goal of 24.
Thus, there are significant efficiencies to be
gained through automation of such bid strategies.

• ? We can increase our revenue per lead from
  33.64 to north of 40.00.

34
Which Ad Is Better?
5.875 30 Year Fixed Rate 0 Points. Refi or
Purchase Let Equifax Find You a Great
Deal! www.Equifax.com/Mortgage-Rates
Impressions Clicks CTR Cost CPC Cost/Conversion
2,579 61 2.37 105.36 1.73 15.05
5.875 30 Year Fixed Rate 0 Points. Refi or
Purchase Let Equifax Find You a Great
Deal! www.Equifax.com/Mortgages
5,235 141 2.69 249.67 1.77 24.97
Why Is Ad 1 Better?
35
Split testing methodology
5.875 30 Year Fixed Rate 0 Points. Refi or
Purchase Let Equifax Find You a Great
Deal! www.Equifax.com/Mortgage-Rates
Impressions Clicks CTR Cost CPC Cost/Conversion
2,579 61 2.37 105.36 1.73 15.05
5.875 30 Year Fixed Rate 0 Points. Refi or
Purchase Let Equifax Find You a Great
Deal! www.Equifax.com/Mortgages
5,235 141 2.69 249.67 1.77 24.97
These two ad variations came from the
mortgage_rates ad group in our master campaign.
While the CTRs are similar between these two,
the conversion rate (and therefore, cost per
conversion) are dramatically different.
Why? All keywords in this ad group have
mortgage rates in them best mortgage rates,
current mortgage rates, and so forth. When we
convey the users intent back to them, by
mentioning mortgage rates in the ad copy, we
give them confidence that they will get what
theyre looking for. The net effect is a 40
improvement in conversion rates for this ad
group. In other ad groups, having refinance in
the display url outperforms mortgage-rates so
what works in one ad group will not necessarily
carry to other groups. A typical secondary
benefit, though not demonstrated here, is an
increase in CTR. When a users search term is
included anywhere in the ad, it will get bolded.
In this example, if we were to send users to a
landing page that mentions mortgage rates, we
would see a further improvement in conversion
rates. Currently, no matter what term the user
comes in on, they get the exact same landing
page. The key is to give users exactly what
they are looking for. Thus, we need a multitude
of ad groups, each with their own tightly themed
keywords, ad copy, and landing pages.
36
When can we safely determine a winner?
5.875 30 Year Fixed Rate 0 Points. Refi or
Purchase Let Equifax Find You a Great
Deal! www.Equifax.com/Mortgages
Impressions Clicks CTR Cost CPC Cost/Conversion
221 8 3.62 9.18 1.15 0.0
Fixed Rate Mortgages Find Great Loan Rates from
Equifax! Compare Over 300 Lenders in
Minutes www.Equifax.com/Mortgages
51 0 0.00 0 0 0.0

• Sometimes there isn't enough data from which to
  draw a reasonable conclusion
• Though we have not seen clicks or conversions in
  Ad 2 doesn't mean it should be deleted. With
  only 272 total impressions on both of these ads,
  there isn't enough data to be able to choose one
  ad over the other.
• Accurate split testing requires that we set click
  thresholds to determine when we have enough data
  to make a statistically significant decision.
  Required sample size is based on confidence
  interval, level of error to be detected, and
  estimated conversion rate. As a rule of thumb,
  for this category of conversion, wed like to see
  100 observations before making a decision. To
  have less than that is to potentially judge a
  variation to be a winner, when it was due to
  chance.
• Sometimes, there may appear to be enough data to
  make a decision
• However, we may be comparing across a time period
  where changes may disproportionately have
  affected one ad. For example, lets say we have
  been running one ad for a week then the next
  week, double all bids and create a second ad.
  When we compare performance over the last two
  weeks, ad 2 will appear to have a higher CPC and
  likely a better cost per conversion, as it wasnt
  running during the first week. We have to
  evaluate performance in light of adjustments
  weve made over the comparison period.
• Having to gather enough data across all the
  accounts, campaigns, ad variations, keywords, bid
  strategies, and landing pages requires a
  significant number of observations. The purpose
  of a test budget is to allow us to gather enough
  data to continue to optimize. Sometimes the
  smallest change can double or triple the click
  through rate. Often, what is most effective
  cannot be predicted. For example, having the
  registered symbol after Equifax give us a
  significant boost to CTR, while keeping
  conversion rates constant. With A/B testing, as
  we find winners, we will make new variations
  based on what weve learned, choosing ads in a
  winner stay on fashion.

37
Keyword Strategy Develop the master campaign
refinancing
Ditech loans
mortgages
Step 1 From public and proprietary sources for
keyword data, we dump raw terms into our process
for sorting.
Home loans
Home equity loans
etc., etc.
Step 2 Sort into themed groups. For example,
all terms with mortgage go down one path, while
those that contain home loan go down another
branch. This treeing goes down multiple
levels, until we have a few hundred groupings.
Home loan
mortgage
For each group (e.g. Mortgage refinance) we sort
into branded terms vs. non branded terms. Now we
can geo-multiply and use dynamic keyword
insertion (DKI) to make the ads more effective.
other
Mortgage refinance
From there strip the keywords of geo modifiers,
and then re-multiply by geo and other words.
Branded terms (can't DKI)?
non-branded
This process is semi-automated, but will always
require some level of human judgement. When we
find new raw keywords, we can pour them into the
top of the funnel and re-start the process.
Ad group 1
Ad group 3
Ad group 2
38
Keyword Strategy Campaign replication
It is through the permutations performed in the
last slide that we are able to build a keyword
portfolio of 700,000 terms, 1,000 adgroups, two
dozen campaigns, and 14 accounts.
The master account is our primary testing
vehicle, and will, understandably, drive a
negative ROI, as we test new concepts. The
master campaign is nationally targeted, to give
us the broadest volume. We replicate the master
account into slave accounts that serve by region
(all searchers who live in the San Francisco DMA)
or by type (master campaign with bad credit
negative terms removed). Periodically, the slave
accounts are refreshed to benefit from updates to
the master--- the latest keyword groupings, ad
copy, negative keywords, bid strategies, and so
forth.
Master Account
Master account replicated into slave accounts,
which are locally rather than nationally
targeted. bad credit slave is also a
duplicate, but doesnt have bad credit negative
keywords.
Los Angeles
San Fran.
NYC
Bad credit
39
Bid Strategy How much should you bid on a
keyword, time of day, match type, etc?
Ideally, bid according to observed value
(maximize ROI).
variables
- home value - credit score - geography - type
new home, refinance, home equity, etc. - others
We have an automated process that raises or
lowers the bid depending on these variables.
For example
Home loan in San Fransisco (high value real
estate) Bid Up 35 Keyword has bad credit
Bid Down 50 Refinancing
Bid Up 150
End Result Bid Up 135
This allows us to tune our bid down to the
keyword level.
40
Bid Strategy
Currently, we optimize bidding to a CPA of 24.
Traffic
Lead Form
Acquisition

Since we are able to tie the conversion all the
way back to the source of traffic, we can see
what keyword, ad, campaign, customer location,
and other variables drove value. One thing that
we could do, but have not turned on yet, is
bidding according to lead value, as opposed to
CPA. With CPA-based bidding, we are assuming
that all leads are worth the same. However,
mortgage leads can vary between 2-6 for a new
home lead and 50-60 for a general refinance
lead. Thus, averaging out to 24 will often
cause inappropriate over and under
bidding. Bidding by multiple variable value was
out of scope for our limited test. The objective
was to demonstrate that with rudimentary
optimizations, we could demonstrate a path
towards scale and better margin.