We’re currently running a DR campaign for a client and we’re tracking “conversions” (online leads) on site.
Without much surprise, I’m seeing much more Post-Impression Conversions than Post-Click Conversions, meaning that people see the banners, don’t click, but come to the site, usually through Google, then convert.
As I was using our Data Management Platform, Artemis, I’ve noticed something weird: During peak hours (7am, noon, 7-8pm) & peak days (Thursday, Friday), the Activity Rate and Cost per Activity was improving greatly for Post-Clicks while seeing the opposite trend for Post-Impressions. The trend is so intense that Post-Clicks are actually more cost efficient during peaks.
I could understand the Post-Click metrics getting better when people have more than 5 minutes to spare, that makes sense, but couldn’t figure out why Post-Impression metrics were suddenly becoming very bad.
The answer came from Google Analytics…
*** Spoiler below ***
I’ve tried hard… but I couldn’t convince myself that people clicking were so different than others and that the individual user itself was the reason why I was seeing such a weird trend. Then it struck me: The Landing Page.
If you click the banner, you get to a Landing Page that basically is the beginning of the Conversion Funnel, while if you saw the banner, then opened another tab and arrived from Google, you have to browse the site a bit to find the Conversion Funnel.
But then… you’ll say… why aren’t Post-Click Conversions always better than Post-Impressions?
This is where Google Analytics saved the day.
The site server isn’t capable of dealing with too many simultaneous users on site. I see average Page Time Load going from less than 5 seconds to over 20 seconds during peaks.
So, if you came from a click and landed directly in the Conversion Funnel, you don’t mind the Page Lag that much, but if you came from Google and have to go through 2-3 pages that take 20 seconds each to load, then arrive in the Funnel and have to wait some more, you say: “(insert censorship here) that (insert more censorship)!!!” and go elsewhere.
I think this is my favorite catch so far, hence the post. It shows the real power of drilling and how you pretty much always should when a number or a trend looks weird.
Now, I wonder how many $$ that very simple server improvement and/or impressions delivery optimization is going to generate.