Testing is a great thing, but it’s also important to know when and how to pick a winner and reap the benefits of your hard work.

Choose strategic moments to remove your worse performing ads and apply the lessons that you’ve learned.  There are times, such as the holiday shopping season when user search interest reaches its peak, when it’s in your best interest to run with your best performing creative.

Testing sacrifices some short term efficiency to ultimately improve your account in the long run.  For that reason, there is an opportunity cost to leaving your tests running longer than they need to.  Take a look at this example:
Ad A and Ad B have minor differences in performance, but Ad A still generates two more leads at a cheaper rate.  Here’s what the account would look like running Ad A all by itself:
Even with the small differences in the example, there is a tangible benefit to reallocating your impressions to the absolute best creative.  Imagine what could happen if all of those impressions in your account (where the differences in CPC, CTR and conversion rate might be much bigger than this example) are going to the best possible creative.

There are a number of different processes that you can follow (depending on your rotation settings) to get your best ads to show for those valuable holiday impressions.

If your campaign is using auto-optimization, we’ll automatically bias impressions toward the ads that are statistically likely to perform better.  While poorly performing ads are already being shown infrequently, it might also be time to cut out the fat.

Starting with your highest volume ad groups, identify the worst performing ads and delete them.  You’ll be able to spot them quickly based on relatively poor performance and low impressions compared to the top performers in their ad group.

If your ads are set to rotate evenly (or indefinitely), start with high volume ad groups where ads have had enough impressions with which you can confidently make a decision.

When manually optimizing, you’ll have to decide which metric is most important to you.  Consider creating your own metrics, such as conversions/thousand impressions or revenue/thousand impressions.

As your tests are concluded, don’t restart them during your heavy season.  Let your top ads shine throughout your peak season, but keep track of ad groups that are no longer testing (via labels) so that you can easily restart your tests once things settle down.

Two things to remember for either of these processes:
  1. Make sure to look at your mobile-preferred ads separately from your desktop ads, as these will likely perform differently than their desktop counterparts. 
  2. It's possible that certain keywords within an ad group perform best with ads other than the top performer. This might happen in an ad group comprised of disparately themed keywords (check your new-and-improved opportunities tab for places this may be happening in your account). You can see the performance of each ad against each keyword by segmenting your ad report by keyword. One solution might be to separate such keywords into their own ad group, so they can serve against the ad with which they perform the best.
What if your ad tests don’t have enough data?

This happens all the time, particularly with smaller ad groups and campaigns.  You’ll have three options to end these tests:
Directional: Pick the ad that is in the lead at this moment.  It couldn’t pass through any peer-reviewed scientific journals, but you’re still basing your decision on recent performance.
Referential: Look for trends in messaging within your high volume ad groups.  If it’s the same basic test across both ad groups (i.e. one main benefit against another), you can reasonably infer that the ads will perform similarly.  Ensure that whatever significant tests you refer to are similar enough to make sense.  You wouldn’t want your successful messaging about shoes to decide how you’ll sell vacation plans.  Boots, however, could potentially apply the lessons from shoe ads.
Indeterminate: Continue to wait until your tests get more information, possibly enabling auto-optimization for your ads.  Not acting is a defensible decision, but it won’t improve your performance immediately.  Low volume ad groups are almost always going to continue being low volume, so the sooner you get comfortable making decisions with limited information on those ad groups, the sooner you could start to see performance improve.
Whichever route you decide to take, you should still end up with two ads in each of your ad groups - one for desktop and one for mobile-preferred.

If you’re merely pausing your ads (since you don’t have enough traffic yet), make sure that you’re labelling them intelligently and clearly so that tests can be resumed once the time is right.

Ending tests doesn’t mean you’re against testing.  Testing is a great thing that leads to a healthy account and a well-informed, data-driven advertiser.  Ending a test is a little like pushing your car into the red - you do it when you need to, but if you do it all the time you’re going to eventually stall out.

This holiday season, treat your account right by cleaning out your low performing ad copy.  Then, when things slow down and you have a nice, clean account to work with, you can restart your testing with something new from our Creatives That Clicks whitepaper.

Posted by Matt Lawson, Director, Performance Ads Marketing