Measuring For Success Part II
Written by Dave Collins, SoftwarePromotions Ltd.
Last month we looked at the baffling phenomena of developers who don’t do metrics. We went through my personal Top Eight Reasons for getting to grips with metrics, the tools I think you need for doing so, and why you shouldn’t rely on your “free” website stats.
Now that we’ve got the basics out of the way, let’s take a look at some of the issues that you need to be aware of when trying to separate the gold from the sand in your server logs.
I assume that everyone reading this will understand what a search engine robot is, but do you also know how active they can be when visiting, scanning and indexing your website?
Assuming that your website has a lot of solid content and is regularly updated, then the search engine spiders will be paying a lot more attention to your website than you may realise.
Looking at the logs for our main website (www.softwarepromotions.com), I can see that all of the main pages (around thirty or so) have been hit by Google’s spiders within the last three days.
And depending on your choice of analysis software, the search engine spiders may show up as regular hits. Be aware.
(2) Browser cache.
Most web browsers by default use some sort of system of caching the web pages they have viewed. The problem is that if a person comes to your website and their browser displays the cached version instead, they won’t see any new content, and they may not even show up in your logs.
On top of this, many ISPs use their own caches to prevent having to download and display massive numbers of the identical pages to many users.
There are various solutions out there, but the most important thing is to be aware of the impact this may have on your logs.
(3) Shared IP addresses.
This particular issue is more or less self-explanatory. Large numbers of people may share the same IP addresses, and depending on the log analysis application you’re using, this may distort the data.
Ten people all sharing the same IP address may show up as one person. And if your log analysis software is clumsy, they may also display one person visiting your site twice as two different visitors.
(4) Blocked referrals.
Irrespective of which software you use, the most common referrer will almost certainly be identified as no referrer/unknown, or will simply be left out of the report altogether. So chances are that you won’t know where most of your visitors came from.
There are five main possible reasons for this:
i) The visitor came to your website from their bookmarks or favorites.
ii) The visitor clicked on an email link. Perhaps your signature, someone else’s email or the many thousands of spam you’ve been sending, Shame on you.
iii) A new window was generated by the website displaying your link.
iv) The URL was entered manually.
v) The visitor is using a browser setting, add-on or plugin to protect their privacy. Privacy sucks.
Whatever the reason, there’s little you can do about it. But forewarned is forearmed.
(5) Download managers.
These have the potential to make a horrible mess of your data. Most work by simultaneously hitting the same file several times. So even though one single person is downloading the file, the fact that their download manager is set to create five active download sessions means that the logs will display five hits. Not good.
The more astute readers of this article will by now have noticed two things.
Firstly that I’m pointing out problems without offering solutions. Why? Partly because not all problems have solutions, partly because I’m trying to stop this article from being too long, but mainly because my aim is to simply bring these issues to your attention.
If at some point you wonder why the website you’re advertising on isn’t showing up as a referrer, then you’ll thank me. Maybe.
The second thing noticed by the more discerning reader is that I have made repeated mention of the way that your log analysis software functions.
The fact is that no two log analysis applications behave in the same way. And the astonishing fact is that if you run (for example) a month’s data through five different log analyzers, you probably won’t find a single item that any two will agree on. For that matter you may be pushing it to find two applications that even come close to each other.
There are two main reasons for the lack of agreement. One is that some of the applications are simply badly written and clumsy, and seem to lean more towards speed than accuracy.
The second reason is that each application has its own way of dealing with the grey areas as identified above.
Whether it’s down to different people using the same PCs and/or internet connection, browsers that block referral information or people walking away from their machines leaving the browser open, there are a massive number of facts that are open to interpretation.
The more accurate applications often try to interpret anomalies and recognise patterns, but this sometimes means that log analysis is less exact a science than many users might expect.
So once you know what the issues are and how much (or little) to trust your log analysis app of choice, the next obvious question is what to look for.
Many of the better applications try to only pick out the pertinent facts, yet there’s also a lot to be said for having access to all the information, so as to be able to filter out what you’re looking for.
The importance and relevance of the data available really depends on what you’re looking for. Here are some of the standard items that I usually look for.
The visitors to your website come in three flavours; and it’s not the good, the bad and the ugly. Total visitors, unique visitors and return visitors.
If your log analysis software only lets you see the total number, then you should probably be looking at patterns and trends. Almost all websites follow some sort of weekly pattern, and many will also demonstrate other regular trends, in terms of days or times of the year with more/less traffic and downloads.
It’s also important not to let your own software’s habits confuse you here. For example, I’ve seen applications that hit a page on the developer’s website when they start up. If you’re doing the same sort of thing, you should (a) be aware of this and (b) hit a page, image or file that isn’t linked in from your website.
Long term analysis should also let you see which are your website’s busy and quiet times of the year. Do these patterns conform to your sales trends too? If not then why not? Opportunities abound.
Knowing which of your pages are popular entry points is vital, as visitors arriving here will receive their first impressions of your website, product/s and company.
Are your most common entry pages set up to do so? Can they be improved? Are there clear links to the rest of your site? And should so many people be arriving at these pages?
Often overlooked, the pages that most of your visitors leave from is also of extreme importance. Many websites with a healthy level of incoming links and search engine prominence will have a fairly high number of visitors who simply aren’t interested in what you’re selling.
Many companies will therefore see the main index page as the most popular entry page, and may see a surprisingly high number of people leave from the page without going further.
This is more or less to be expected, but when you look through which of your other pages have high exit rates, you may be quite surprised. Pages that you may have thought were very effective may actually prove to be black holes – swallowing up visitors who disappear without a trace.
Another often overlooked vital statistic.
A basic example.
Let’s say you’re running a Google AdWords campaign that sends you 500 visitors a day for $0.02 each. $10 for 500 visitors might seem like quite good value.
But what if you looked through your logs and saw that of those 500 visitors, the average time spent was 0.5 seconds?
Something would be very wrong, obviously.
How long visitors should spend on your pages depends on who they are, where they come from, how much information is on the page and how you’re presenting it. But if you take all these factors into account you should be able to identify which pages are working well, and which could do with some improvement.
The links clicked from each page tell you a lot, but don’t forget to look at the page and see where these links are physically located.
Chances are that links towards the top of the page will get more clicks than those at the bottom.
I’ve seen web pages for software with only one download link, right at the very bottom of a long page. And this is a good idea? (Hint: No.)
Where the links are placed, how they stand out and the text and/or images used can massively affect how many clicks they receive. Experiment and reap the benefits.
Most log analysis applications have some sort of path function, that lets you see the most common routes that visitors take when travelling through your website. Most are very inaccurate but will still probably open your eyes to behavioural patterns that you could never have predicted. Watch, recognise, learn and respond.
Some of the pages on your website are more important than others. Obviously you want visitors to buy and try your software, but you probably have other pages that you feel do a good job of convincing them along the way.
How popular are these pages? And more importantly, how much more can you do to increase the percentage of visitors who go to these pages? Another hint: the answer is a lot.
The bottom line is that log analysis isn’t just about number crunching. It’s about understanding why your visitors do what they do, realising what can be done to improve the figures and patterns, and getting a better return on the traffic that you’re already receiving.
Thorough log analysis is no more about number crunching than dentistry is about brute force with a drill. The software performs all the numerical analysis for you. Now you need to apply what you know, understand what you see and join the dots. It’s somewhere between in-depth detective work and a jigsaw puzzle with many fiendishly small pieces.
But the effort will pay off. Set aside time for regular analysis of your web logs and you can only gain. Visitors to your website are hard to come by, so the more of them you can convert to downloads and sales the better.
Be seen, be sold.