Gallup tracking poll and AdSense stats

Forum Moderators: martinibuster

Message Too Old, No Replies

Gallup tracking poll and AdSense stats

Graphs show "optical illusion" of spikes and dips

signor_john

7:04 pm on Aug 5, 2008 (gmt 0)

A political-science professor's analysis [huffingtonpost.com] of the Gallup tracking poll has two graphs that show how daily spikes and dips (whether in U.S. presidential candidates' poll numbers or AdSense stats) can be deceptive. The first graph shows the dramatic-looking daily ups and downs, while the second graph shows the much smoother (and more meaningful) statistical trend over a ten-day period.

NOTE: For the purposes of this thread, it doesn't matter which candidate is leading in Gallup's tracking poll. The point is that the "random noise" of daily variations can be more distracting than illuminating. Maybe AdSense should offer 7-, 30-, and 365-day "tracking graph" options to soothe the nerves of jumpy publishers.)

HuskyPup

8:29 pm on Aug 5, 2008 (gmt 0)

Heck s_j, I'm glad my charts don't look like that!

It would certainly keep one's heart fit seeing all those highs and lows:-)

zett

8:15 am on Aug 6, 2008 (gmt 0)

The point is that the "random noise" of daily variations can be more distracting than illuminating

100% agree. That's why I am always charting moving averages against the daily data. This exercise, however, makes the random noise even more ridiculous.

The question that really concerns me - why are we seeing such extreme "random noise" only on the data provided by Adsense?

Traffic and CTR are solid and predictable, and the advertiser base appears to be stable, except for the odd MFA now and then. Yet EPC, eCPM, and revenue are random, really. Why? How can it be that, for example, the average EPC drops to 60% of the previous day, and then jumps back to 100% the next day? I'd say this is strange.

Scurramunga

9:11 am on Aug 6, 2008 (gmt 0)

Whilst you make your point about the need to keep everything in perspective when dealing with short term fluctuations, you will come across as a little partonising to those of us who have made comments based on longer term data trends.

signor_john

2:34 pm on Aug 6, 2008 (gmt 0)

Scurramunga, this thread and the study in question aren't directed at people who understand the value of analyzing long-term data trends.

Are you suggesting that this forum is devoid of members who worry unnecessarily about daily (or even less than daily) variations in their AdSense stats?

And are you also suggesting that I shouldn't have mentioned a study that could be reassuring to those members just because you might misinterpret my post?