Chapter 2. Graph Shapes
You don’t need a deep knowledge of statistics to see interesting things in a graph. Human behavior makes a few basic shapes on a graph, and they all mean something.
There are two graph styles that human behavior creates fairly often: Traffic and Structured Behavior.
Note
I have used a bar graph in these examples because it was simpler to understand. Your analytics might use lines or dots or whatever. Don’t panic. They all do basically the same thing. That’s why we’re learning shapes, not graph types.
Traffic Graphs
These graphs show the number of people that did something, over time. Like the number of visitors per day. You can call that “traffic.”
Traffic will always be moving up and down a little because random things in the world happen every day, even when your site doesn’t change at all.
That’s why you can never assume that a small change in traffic was caused by a new feature or a design change.
Now, on with the shapes!
Random/Unexpected/One-time event
People don’t suddenly change behavior without being provoked.
Did you run a weekend campaign? Or, is something technical causing problems on one of your pages? Or, maybe your startup just went public? When you see a spike in a graph (or a sudden drop), try to find out what is causing it, because—as tempting as it is to believe that you have spontaneously become cool—there is always a reason for a spike; sometimes good, sometimes bad.
Predictable Traffic
A mature site (or a boring one) begins to have a clear pattern in visits.
See how there is a pattern repeating over and over, like a wave?
Sites that are popular with “office workers” often get more traffic during week days. If your users are all kids who go to school during the day, weekends might be your big days. That is very common and very normal.
But...
If it’s a healthy pattern it will usually come with a slow growth trend, too. If you see a super-predictable pattern and your numbers are slowly going down, your users might be dying of boredom. Shake things up!
Structured Behavior Graphs
The other big type of graph shows what people are doing. The date or hour when they did it isn’t as relevant. You have a big effect on this type of behavior via your IA.
Note
“Structured behavior” is just a phrase I made up. If you say it at work, you will sound smart, but nobody will know what you’re talking about, so you should explain it.
Exponential/Long Tail
This shows a strong bias toward a certain type of behavior or decision.
The graph shows a shape like a slide. More people will click the first thing than the second thing, and more people will click the second thing than the third thing. Any time there is a visual “order” or a natural sequence—like a menu that users read left to right—the graph will look like this.
The list of “top pages” will usually look like this, too, because you can’t get to page two without going through page one. The detailed view of Time-per-Visit or Pages-per-Visit (see Chapter 6 about time) also usually looks like this, because spending more than 10 seconds on a site is so damn difficult.
Exponential with Unexpected Order
When users are ignoring the structure you gave them, it looks like this.
This one is more interesting. If your data looks like it has all the right pieces but a few are in the wrong order, it means that users have different priorities than you thought. They are clicking the second thing before the first thing sometimes. Those crazy bastards!
Try to change your design/IA so it matches what the data is telling you. Don’t try to change the users, though; they hate that.
Exponential with Power Users
This shows a small group of people who are doing a lot.
This graph looks pretty similar to the first “slide shape” but it has a bump. Some people think the graph is pregnant, but those people are idiots. This is what it looks like when you have a small group of people who are loyal, or very active, or spend a really long time on the site. They are doing much more than the average user, so it makes a bump.
Find out what motivates them and make more of it!
Exponential with Conversion Problem
A huge drop between two bars often indicates a barrier for users.
You want your “slide” to look gentle and smooth. If it has any insane drops or rough spots, those indicate problems. If the home page of your site is really confusing, you might get a graph like this because very few people will get to the second page. A/B tests are a really effective way to find the problem, if it isn’t obvious.
