Moving Average of Adsense Income

How big a window to get smooth results?

[ronburk]

I'm just starting out in AdSense and have only low traffic. I could look at the numbers and get excited about how one day my income is up, and upset about why it's down later, but of course that's silly with such small numbers. So, I instead calculated the moving average of my income. Once I increased the window of the moving average to about 30 days, I got a very smooth curve (slowly rising income, which makes sense since I am regularly adding more income-generating content).

My question is, is the size of window required to get a pretty smooth line different for each website, reliant on the level of traffic, or...? Anybody else care to post how big a window required to get a smoothed moving average of their AdSense income? Of course, this presumes a relatively stable website (e.g., if you just made a wrenching structural change last week, there probably shouldn't be a smooth line in moving average income at that point.)

Note: I think this is an abstract enough topic that it doesn't violate AdSense TOS.

[ken_b]

It isn't just big changes to a website that can dramatically affect earnings. Sometimes rather small changes even just rewording part of the text on a page, or adding a link to a new page, can apparently cause quite a difference.

Adding new pages can also make a big difference. Sometimes a single new page can drastically affect income.

As far as the window goes, I think 30 days is probably a minimum time frame to use. 90 days may be a lot better.

[wgonz]

Hi Ron! Hi Ken!

I’ve been working with 7 days in order to filter out the weekend effects.

Some advertiser budget may be based on weekly or monthly bases, so the minimum of 30 days make sense.

Now reading your posts, I’m thinking to employ a multiple of 7, may be 28 days or 35 days window.

wgonz

[loanuniverse]

I use 7 days and the curve is fairly smooth. Unfortunately, it is not going in the right direction. Using 7 days is the minimum to cancel out the weekend effect in my case {where traffic drops by 50%}

[hotftuna]

I use a daily total of all my earnings then do a 7 day and 30 day average of the total.

[tamen]

On a somewhat related note; What do you use to make those statistics?
I would like to know my moving averages as well. Moving averages, sounds almost dirty ;)

[darkmage]

Hi Ronburk.

The short answer is you probably can't get enough data for future predictions. When projecting the future, an average can really only work when the input behaves in a regular, consistent manner. Even then, you may have an average but it won't tell you anything specific about future trends.

Adsense is extraordinarily complex and your site is just a tiny piece of a very big dynamic puzzle. Just when you think you get enough, all the parameters will change.

[linear]

I was looking for a weekly cycle as described by a few, but my sector seems not to be disposed toward that. I looked at 7- and 30-day plotted on the same chart as my daily numbers. That was back before I got convinced that trying to overanalyze the stats wasn't a productive use of time. (Having said that, it's a breeze in excel to do it.)

[loanuniverse]

On a somewhat related note; What do you use to make those statistics?
I would like to know my moving averages as well. Moving averages, sounds almost dirty ;)

I use Microsoft Excel.

Steps

1. Download the report in CSV format
2. Open with Excel
3. Get EPC by adding the earnings of seven days by clicks of seven days. Once you get the formula right, you can copy it down and fill the whole column.
4. Graph it.

[bernis]

there is a special software ... google for "csv adstats"

[suidas]

A seven-day average works for me, My site is composed of various subsites with quite different keywords, so the average tends to be pretty stable over a seven-day period.

When I do it by single topic I get very different results—wild swings up and down. I had chalked this up to game-theory instabilities in the AdWords keyword markets, particularly in low-competition markets. But maybe this is over-thinking it.

Anyway, this pattern of topic swings and general stability has concinved me that all posts about how Adsense was great or terrible on this day or week came from people with single-keyword sites. I don't think someone up in Google corporate is looking at a screen going "Oh my God, Adsense is up 300% today! WHAT'S GOING ON GUYS?"

[trader]

there is a special software ... google for "csv adstats"

Bernis or anyone, Where do get that software?

[ve3cnu]

google for "csv adstats"!

[trader]

Now I get it. I was thinking he meant there was an actual script or software for sale with the name 'Google CSV Adstats.'

[ve3cnu]

glad you got it.
sometimes the no url policy here is difficult.

[trader]

I was going to get it but I am leary of installing a program named .NET Framework 1.1? What is it? Does it take up much resources? Does it run OK? Is it a risk? I always hate downloading files to my PC, especially ones which serve an unknown purpose and never heard of it, and only needed to run 1 program.

Also, regarding CSV Adstats, I see it is on a France website: nix.fr - Is the program in French? Also wondering why it say's this: "Warning, This beta version will stop to work the 12/07/2004" P.S. The fact is stops working soon and the bad English does not give me much confidence!

What happens after that date, oddly it does not say anything about that from what I can see?

[ve3cnu]

The .NET Framework is Microsoft's answer to java.

It is no more (un)safe than the O/S itself.

[amznVibe]

Sometimes daily stats are not enough information.
It helps to watch the numbers before noon and after noon, etc.
(though adsense stats lag makes it impossible to be precise)

I actually log my stats 4 times a day at the same times to make sure nothing funky is going on (like some way off target or raunchy ads being displayed). All depends on your site(s) volume if that much attention is worthwhile.

[trader]

Thanks but does anyone know about CSV Adstats, on a France website: nix.fr and why the program apparently stops running next week? See my prior post.