Six rootin, tootin myths about web analytics
A large chunk of my job at Fairfax NZ is about numbers – wonderful, glorious numbers – mostly looking at web traffic at Stuff.co.nz and its competitors. It’s work that lots of people might think is pretty dull, but those lots of people would be both wrong and unawesome.
Web analytics is so cool I can hardly keep my trousers straight writing about it. It’s like the Wild Wild West of statistics, and I’m like Will Smith, except in a movie that doesn’t suck. Plus I’m a better dancer.
So in the spirit of gunslingers everywhere, I’m going to shoot down some myths about the world of web analytics.
Myth: You need to be good at maths to understand traffic data.
Fact: Not really. While having a feel for numbers certainly helps, if you’re looking for broad trends it’s usually good enough to eyeball the data, especially if it’s in a nice chart. That’ll give you a good sense of what’s going on, and for day-to-day life that’s usually good enough.
Myth: My site got lots of hits. I rule!
Fact: First, a quick sum-up of the major web metrics: (metric = way to measure something)
Hits – Any request to a server. If your page has five images on it and that’s it, there’ll be five requests to the servers for images, so you’ll get five hits. A lot of people say “hits” when they mean “page impressions”. Ignore “hits”, they don’t mean anything unless you’re talking about server loads.
Page View or Page Impression (PVs/PIs) – Someone reading one web page once on your website. For example, every time someone reads an article on Stuff.co.nz, a little bit of code runs that pings an analytics server, registering one page view. Lots of page views is therefore good news generally, though what counts as a “page” is open to interpretation, and varies between different audience measurement firms. It’s starting to fall out of favour, but is still widely used because it’s related to ad inventory (you serve ads every time a reader loads a page).
Unique Browsers/Visitors (UBs/UVs) – Basically translates into “people visiting my site”. It’s actually more like “unique computers visiting my site until they clear out their cookies” – so if you check a site at work and at home, you count as two UBs, even though you’re the same person. Generally though, it’s a pretty good indication of the size of your audience.
UBs are measured over period (hour, day, week, month usually), which you can change depending on what you’re trying to find out. So if someone visits your site twice in a week, you see two daily UBs (one for each day) but only one weekly UB.
Visits or Sessions – Every time a UB hits the site, it counts as a visit (broadly speaking at least). So a person who visited Stuff.co.nz twice yesterday and once today, seeing 10 pages in total records three visits, 1 daily UB for each day, 1 weekly UB for both days, and 10 PIs.
Engagement – Basically time spent on your site (or sometimes on a page). This is growing in popularity because if people spend a lot of time on your site, they’re more connected and loyal to your brand, and are more likely to interact with ads. This compares to readers coming from search engines or social media, who often just pop in for one article then leave.
There are other metrics too, like PI/UB or average daily UBs, but usually people stick to what’s above.
So what should you look at? Depends on what you’re trying to do. Want to serve lots of ads? Try increasing your PIs. Want to serve those ads to lots of people? Then it’s UBs. Want to make sure your audience is actually seeing the ads? Then probably engagement.
Hits don’t really enter into it.
Generally you want to aim to increase whatever metric advertisers are currently focused on. No metric is perfect though, and every metric can be gamed (raised without really giving the advertiser what they’re after).
Myth: Wow, I didn’t know anything when I started reading this post, but now that you’ve told me that, I’m going to jump in and figure out exactly how people use my site.
Fact: You’ll screw it up. The first few times at least. The truth is, you don’t need a great head for numbers to get a top-level sense of what’s going on. But the deeper you go, the more precise you try to be, the trickier things get. If you want detailed site analytics, you need to find someone with a good feel for numbers and basic understanding of stats and maths. It’s very, very easy to trick yourself.
Jumping in the deep end is the best way to learn and you should totally do it, just check what you’re learning is actually right by asking someone who knows.
Myth: Ok then, I’ll just hire some numbers guy to tell me what’s going on.
Fact: You’re better off hiring someone who knows your site to work out the numbers, than hiring someone who knows numbers to work out your site. Knowing about work practices, site navigation, technical quirks is just as important as regression curves, rolling averages, and how to do a percent increase (I die a little inside every time someone asks me how to do that last one).
Myth: As long as your traffic is up on last week, all is well.
Fact: Well, it’s better than going down, but traffic can fluctuate for all kinds of reasons, and what you want to know is whether the overall trend is good or bad. You’re better off comparing one week to the average of other weeks like it (eg for a news website – no public holidays, no earthquakes in Christchurch, you get the picture). That’ll show you where your traffic is really going.
Getting a traffic spike is great and all, but if those new visitors don’t keep coming back, you’ll end up back where you started and the site will never really grow.
Myth: As long as your traffic is going up over the long term, all is well.
Fact: Traffic on the web is growing in general. So you can gain PIs and UBs compared to a year ago, but still be losing market share if you’re not gaining them faster than the market is naturally growing. If your growth rate is better than your competitors’, you’re doing fine though.
More myth-busting tomorrow, cowpokes.