The key to succeeding with PPC is getting the right ad with the right keyword.  Have multiple ads and test to find the right combinations. Optimization is not an accident and if you have to test and measure.

We’ve done 1000’s of hours of testing and when you test you have an 86% chance of improving performance with a 30-50% average increase and possible gains of up to 400%.  Why Test?

3-5 seconds - only have a couple seconds to attract attension. 

• Bolded Keyword
• Authorized Seller
• Lowest Price
• Think different (can be hard but there are ways to do it)
• have a differernt ad for high performing keywords
• match the right keywords with the right ads - 181% increase in conversions.
• Test widely with different copy and test position

Testing the different conbinations and measuring everything is the best way to actually determine what the best ad, best keyword, best position - in other words you need a system to do this properly.

Interuptions can work well too (need pizza for books?)

Test keywords, credibility words, keywrod segmentation …..TEST, keep testing all the time).

Measure - How to Measure:

• DIY
• tracking urls
• measure CTR and conversion
• in reality conversions/CTR normally go together
• Poor CTR= "disabled"
• Have small AdGroups with the lowest manageable number of keywords per adgroup.  This seems to be the best way.
• Seasoning and Quality Scoring - more leeway for advertiser after a successful account.
• Start with the basics before you try more esoteric keywords.

When your successful what is the next step

• reviving
• sometimes you can avoid minimum bid by testing creating a number of new ads

QUESTIONS:

What parameter do you use for min number of impressions and CTR?

It varies - it’s based on time; sometimes it will take longer for conversion data and there is always latency in conversions.  The right number of impressions and clicks is completely depends on baseline conversions.  Testing in the Long Tail can take a long time (stemming - keywords are close to each other and agregated or you can use landing pages and stemming to associate keywords with actions).

Strategy - thoughts on setting up a 1 year testing program?

Putting together a testing program has some subjectivity to it - if you think the landing pages really suck indicates you should probably test that first.  Testing scent in the ad copy suggesting something in the landing page that does/does not exist - you need to test that.  The best is to go after the lower hanging fruit that you believe you can go for first.

If you take over an exisiting campaign that the customer runs themselves and only has one or two ads - we can jump into some immediate ideas and gains on what is going to work for that clients.  In a sense, a client that has a poorly performing campaign will show the most immediate improvement from a testing program (based on what i am hearing at this session).

Start with the areas that have the highest impression levels - because you’ll able to define the winners quicker.  Low impression pages will take much longer to figure out what works best in ad copy or impression.

We immediately disable "auto optimize" and split up the keyword  list to the smallest possible AdGroups.

Regressive Testing - how do you handle clients who want increase in the program but don’t like to take risks?

Test the ad copy - how many ads are you running - there will be some drop off during testing.  Part of it is there is no pain with no gain.  The nice thing with the ad copy is you can change it without changing the budget.

You can set up some shadow accounts for clients to test and run them in parallel (we ususally get paid for it).

Think about what happens when you conduct a test (you’ll get positive results or negative results).  If you get negative results you can learn something.  Kevin Lee thinks that each clients should allot 15% of your budget for testing they will lose out (this is the same as saying 15% of your budget wil perform variably - but if you don’t test you die).

Kevin Lee - if client is not willing to spend 15% of their budget on testing they will get defeated eventually - and this is true of organic and paid search - it’s true of everything.

Campaign structure needs to be right.

  Question about Referrer data

Kevin replied about the match type being aggregated - depends on the tracking platform how well it’s done.

Joe does not do much testing on Yahoo and they’ve told people they are going to have a new platform since last summer and they’re still delayed - most testing is done on AdWords.  Also the demographics are different for Yahoo Paid Search and Google Paid Search (Google’s crowd is younger and male while Yahoo’s is older and more female).

Done any testing with AdCenter?

We’re pleased with the new changes to the API - and it’s pretty strong - maybe they put even more into it (Kevin Lee - DidIT). It’s possible to do some really neat stuff with the MSN API - that will go away as more advertisers come onto the system).

MSN traffic conversion rates are higher than Google’s (right now).

Is it possible to export an AdWords account and load it into MSN and have it perform as well?

It’s great for testing though MSN gives you a little more room for your ad. You have to start somewhere.  What’s interesting is the current MSN model makes it more difficult to tune your creative to different demographics.  Down the line it may be possible via the API to do this.

Question about the Funnel with broad generic phrases and then getting down to more specific - how to measure the value?

In order to do it right you have to be able to track the actual source - it’s not easy to say follow someone who heard about "boston hotels" on the radio and when directly to the "Commonweath Hotel" directly in organic.