Archive for March, 2011

AdSense: Revenue Prediction

Until very recently, the AdSense UI featured “today’s estimated earnings”; which incorrectly suggested that this is the amount you could expect to receive that day. That wasn’t the case; it really just showed how much had been earned so far (‘estimated’ to give Google some moving space). The interface now labels this as “Today so far”, which is far more accurate.

It did give the idea of creating our own revenue prediction model, which would be useful for allowing you to understand whether today’s earnings was on track with expected performance.

We’ve now figured out how to collect revenue data from AdSense and analysed how the intraday earnings trends over a 24-hour period, so had just one missing piece to solve. The time and percent values from the prior analysis were placed into an array and a function created to return an estimate of the full day’s earnings based on this historical data.

The output now reads as:

Next: getting this data on your mobile.

Update: Grab the script and build your own prediction model.

 Tags: AdSense   Published: 25th March '11


The M&A team at Lunatude have been busy. I’ve been speaking with webmasters for months and finally stumbled onto a site fitting for my portfolio. is an online repository of fan fiction for the insanely popular Twilight series.

Users come along and share their stories. Others can edit and review these stories. Pretty simple. The site is in need of some serious TLC; it’s experiencing falling traffic and growing competition. It’ll be a challenge, but it may just pay off.

Here’s the current homepage:

Want to see an example story? Try Dream’s Shadow, apparently it’s quite popular.

Redevelopment work has begun in earnest and an all new design should appear in a couple of weeks time.

I’ve tagged this, and some other posts, with the all new Introducing tag, so you can see a list of the other sites I’ve toyed with over the years.

 Tags: Business, Introducing   Published: 19th March '11

AdSense: Intraday Earnings Trend

Google Analytics can display a break down of a day’s traffic into hourly chunks, but AdSense is strictly limited to just whole days. I’m sure Google has a few fair reasons not to provide this (data reliability, computation time and so on). For educational purposes, we can try to accomplish this data-set ourselves.

Thankfully, the AdSense interface does tell us how much we’ve earned so far during today:

Now that we know how often Adsense updates its data, we can configure our script to efficiently collate the right amount of data that’ll allow us to do some number crunching.

The script is now setup to query the rolling balance every 15-minutes and store the value in an SQL database. With a spreadsheet to hand, we can then look at the relative increase in earnings at each interval. Here’s the chart showing that data for a 7-day period.

The dark green line provides the mean average over the 7-days, smoothed as a moving average. I opted for the PST timezone for the X-axis, since most of my site users originate form the US (and California specifically).

I’d like to show a nice bell curve chart to show when the major growth occurs, but my spreadsheet/statistics skills are a bit limited. It’s pretty clear to see that the rapid acceleration is in the morning (6-10am) with fairly flat growth thereafter.

Next up, let’s do something somewhat useful with this finding.

Update: Grab the script and collect your own data, plus build a prediction model and mobile app!

 Tags: AdSense   Published: 7th March '11

AdSense: Data Update Frequency

I’m sure I’m not the only one with a minor addiction in checking their AdSense statistics. The Overview page shows the running total of that day’s earnings. Many have asked, how often does this figure update? So I thought I’d figure that out.

Pretty simple; I just needed to check the figure every minute of the day and make a note when it had changed. A handy script completely automated the process. This is the result:

Each line marks an update (Numbers on the Mac isn’t very good with histograms). I wasn’t collecting data every minute for the first couple of hours, hence the bunched up markings. Updates during the day weren’t entirely regular, but they did generally occur once every hour at around 40-50 minutes past the hour. Each update seemed to take a minute or two longer than the last, so the exact time is properly always in flux.

This was an important piece of information to know in order to carry out my next experiment. Stay tuned.

Update: Grab the script and collect your own data, plus a whole lot more.

 Tags: AdSense   Published: 1st March '11