AdWords 911: The 4 Step Emergency Fix for Dying AdWords Accounts

Gear your KPI to the business's goals:

• Lead focus?  Use cost per conversion, number of leads, and conversion per impression.  Or get more advanced and make the client tell you how much the leads are worth, and assign that value in the conversion code.  This value may be wrong, but has psychological power.  You can argue about the right value later on.  You're more likely to get the right one when you start with a guess than if you never use a value.
   
• Retail? Use ROAS.  To do this, you have to make sure the shopping cart is passing revenue totals into the conversion code dynamically.  In AdWords' new conversion tracking area, when creating a conversion event, you have to specify a dummy value if you want those lines to show in the conversion code- replace that dummy value with your dynamic revenue total variable.  You'll also look at CTR, CR, conversions, etc., but ROAS includes all of that and average transaction amount (order size) and the cost per conversion, so it's a mega-metric.

2.    Check Campaign Structure

If you don't have campaigns organized correctly (for geotargeting or specific budgets for sets of keywords), you'll have administrative problems.  Make sure you find out these needs from the client.  We use a PPC questionnaire when we start with a client to uncover these needs and direct our campaign structures.  Also make sure you've set ads to rotate, not optimize, if you plan to follow the advice in #4 below.

3.    Check Adgroup Structure

Do the adgroups each have a topic and keyword similarity?  If not, ads will end up too general to get good results.  Segment keywords by topics and similar words.  Then write more ads.

4.    Are They Split-Testing Ads? 

Every AdGroup needs at least two ads, and no more than four ads except in rare situations.  I usually test just two ads per AdGroup at one time.  Testing ads is the NUMBER ONE way to win at AdWords, second in importance only to using the right metrics.  If you think the existing ad is doing well, test a variation of it.  If it's not producing, test a new ad against it that represents a completely new concept- new offer, new message, new emphasis.

5.    What's Working and What's Not?

Only at this point does it make sense to look at reports (has your client ever run one?), the account snapshot, etc.  If you had to make major changes in the first four steps above, wait a few weeks (unless it's a high spending, high volume account) to get the stats you need to be sure what's going on, but this is where you start asking:

• Which ads are working, and which aren't?
• Which keywords are performing?
• Which keywords are bid too low to be visible?
• Which keywords need to be in a different adgroup?
• Do we need new adgroups?
• What keywords and ads should be paused?

I think step #5 is a completely different topic- how to optimize a well-organized AdWords campaign, and deserves a separate blog post... so more on that later.