Making use of moving averages - at tutor2u HQ

I’ve been trying to get my head round where our website will be in 1-2 years time - a kind of a sales forecast, but using volume not value data. So I thought I’d try some moving averages (BUSS3 - sales forecasting!)...

What prompted me to try the exercise was some good news from our Google Analytics service (a tiny piece of code on every T2U webpage records every imaginable piece of activity on the site. We passed a major milestone for us last week - 1 million monthly unique users on the www.tutor2u.net site for the first month since we started using Google Analytics (back in August 2006).

I was aware that the daily website traffic (“demand”) goes up and down in a similar pattern each week (busy Monday - Thursday; much quieter on Friday, Saturday and Sunday). There are also well-established seasonal peaks and troughs in our website traffic. Its very busy in the run-up to exams and during term-time; it goes much quieter during the school holidays (though not as much during the crucial Easter revision period).

But what about the trend? What does the data show, and can we extrapolate a similar trend over the next year or two?

I started by looking at one measure of website activity - the number of people who get the daily Google Feedburner RSS feed from our Economics Blog delivered to their inbox. Economics was our first subject, and we’ve offered an RSS feed for just over three years now (feeds for other subjects were added later).

The data from Google Feedburner showed wide variation in the number of people who click on a content link each day. These variations caused by the timing and seasonal issues above, but also by the volume and type of new content included in the daily feed.

However, the 30-day Moving Average of people visiting the Economics Blog via Feedburner shows a much clearer trend line:

Can I reasonably extrapolate further growth in this data for 2010 and 2011? Possibly. There’s nothing in the historical data to suggest a likely tail-off in RSS subscriptions. And I’m confident that we can continue to add lots of useful content to the Economics Blog which will encourage subscribers to click on their RSS feeds or daily emails. But as the number of competitors grows (for example, the EBEA has started an Economics Blog, as has Philip Allan Updates), then we have to compete harder for our user’s attention. On the flip-side, more and more teachers are making use of RSS feeds to help them filter the available information on the Web, so maybe we can continue to tap into that trend.

When I look at the data set of daily website visitors, a different moving average is required. I need something which ideally smoothes out those weekly and seasonal variations. So this time I opted for a 365-day moving average. I.e. the average number of daily unique website users, but calculated over the last 365 days. The chart below illustrates that particular trend line:

The trendline shows a less convincing case for extrapolation. It looks like we flatlined for a year or so back in mid-to-late 2007 and early 2008. It is only since we launched our new subject blogs that we’ve seen a rapid growth in daily unique website users. It seems that we hit a new daily high almost every week recently. But someone looking at investing in the website would want to see a sharper gradient to that 365-day moving average I suspect