How to Pay Your Holiday Shopping Bill: Unusual Things Never to Do in Adwords (#2)

Here’s another way to pay that oh-so-dreaded credit card bill associated with your Holiday shopping this year…

In the new AdWords user-interface, you can segment your data by device type.  Assuming you’re tracking conversions, you’ll be able to quickly see whether (a) you’re achieving mobile distribution to devices with full internet browsers; (b) whether these prospects are worthwhile to you. (You get this data by going to any active campaign, choosing Ad-Groups –> Filters and Views –> Segment By –> Device)

We often find high performing campaigns on Google and Search don’t do nearly as well on the mobile network.  And when that’s the case, you can just turn it off.

So… “Unusual Things Never to Do in Adwords Tip #2A” is:

  • Never let the mobile network run without evaluating its click through rate, cost per click, and cost per conversion

And while we’re talking about segmentation, let’s throw in Tip #2B:

  • Never let an adgroup run for more than 2500 clicks without checking it’s performance by day of the week.

More often than not there are some higher-than-average responsiveness days, as well as some which fall below the mean (sometimes far below).   Once you know this, you can go into the day-parting settings and adjust your bids accordingly.  Can you smell the money?

Here’s a little secret sauce for getting this right.

Determining a “winning” vs. “losing” day of the week is just like split testing two ads.  You consider each day to be an ad, and then evaluate the number of clicks and conversions (or if you have enough data you use the ultimate metric which is “profit per impression” –> see SuperSplitTester.com) against the MEAN for all days of the week which is listed at the very top of the segmentation list for each adgroup in the new interface.

You definitely want to be SURE of your findings before you hit the dayparting buttons, otherwise you’ll waste money instead.  So let’s walk through an example:

(You get this data by going to any active campaign, choosing Ad-Groups –> Filters and Views –> Segment By –> Day of the Week)

  • In the data above, it looks like Wednesdays performance is significantly above the mean for the group in both Click Through Rate (6.16% vs. 4.50%) and Conversion (29.63% vs. 21.6%).  If that’s true, we should be hitting Wednesday harder.  But if we’re wrong and this is just due to chance, we’d be wasting money.
  • When I  head over to SuperSplitTester.com (note: I used $10 as a “value per sale/opt in” for simplicity – insert your own) and enter these numbers, including the associated impressions for each of these “ads”, I find, indeed, a 90% confidence level that impressions shown on Weds are worth a LOT more than impressions shown on any average day of the week.  While I’d prefer to see a 99% confidence level for day of week analyses (they tend to be less stable and include a lot more random noise than normal split testing), the data I’m seeing is strong enough that I’ll go ahead and bump up my bids significantly on Wednesdays, and watch the results carefully.  Once it reaches 99%, I’ll take additional steps to grab every Weds impression I can!
  • When I run Thursdays data against the average for all days, I initially see it looks like it’s worth a LOT less than the average day, and in fact I’m LOSING money those days.  However, SuperSplitTester.com also tells me I don’t have nearly enough data to be sure even of the difference click through rates, much less the difference in conversion or profit per impressions… so I’ll need to let it run longer.  But I WILL need to keep checking it because there’s a good chance I’m losing money
  • Note: the astute statistician will notice I’ve purposefully introduced an error into my tests above which makes it less likely you’ll find day of week differences until they’re reasonably large.  Extra points for anyone who points it out!

Now, there certainly ARE more sophisticated ways to run segmentation to get an extra bump (especially when you involve your Analytics account), but the average person–even the average advanced marketer– just will never do them so I’m not going to list them here, and if you’ll do the above you’ll be ahead of at least 90% of your competition, I’m sure.

Short story?  Take 20 minutes to run and analyze these simple reports, available right in the basic user interface, and you just might find Santa left you a surprise!

Hope it helps,

Dr. G 🙂

PS – As the hyper-responsive club has developed, it’s evolved to include more and more about my adwords methods since these are most frequently requested by my best customers.    Why not come watch me build a business step by step, and apply the cheat sheets and over my shoulder videos to your own dream?  (The price is going up again on January 1st or when I reach 1,000 active members, whichever comes first)  hyper-responsive club

PPS – I’ll be SPEAKING in Hawaii at Perry Marshall’s Advanced Adwords Seminar.   (I rarely speak at seminars anymore but I couldn’t say no to this one!).  Grab your early bird spot and come meet me there, won’t you? Rob Sieracki, the Director of PPC at Rocket Clicks will be attending and also speaking in his own slot … truth be told, he knows way more than I do at this juncture!

PPPS – Thanks to Jerrold Burke, kick ass ppc analyst at Rocket Clicks for inspiring me to write this post.

Comments

comments