One thing I love about newer cars is the information rich displays. I have been travelling a lot and nothing has changed my driving habits as significantly as the information on the dashboard.
Growing up my best friend had a late 70's Cadillac that had a crude display, which reported your miles per gallon at that moment. There was no aggregate or average, just what you are doing right now. We found ways to manipulate it to report up to 70 miles per gallon (get going fast, let off the gas and you could watch the miles per gallon move up).
The car I drive now is more sophisticated in terms of reporting. I can set multiple trip mileages, gallons of fuel used, length of time the car is on, and probably 15 other driving metrics. But no metric has changed the way I drive more than the average miles per gallon. I watch it meticulously; I obsess over it. The way I drive has been drastically altered by this single metric. I speed up slowly; I wait as long as possible to use the brakes; I take routes with fewer stop lights even if it is a slightly longer route (if you don't think this helps save $$, track your mileage for a month on each route, you may be surprised); etc.
Not surprisingly, what we did in the Caddy actually cost us money (not that we cared, gas was cheap in the mid-90's). We could momentarily get high miles per gallon, but only after excessive acceleration. For a metric to work for you it needs context and attention. When a metric gets appropriate context and attention it should change the way you drive. Why "should?" Because, if you give it context and attention and it doesn't change the way you drive, it is probably the wrong metric.
Context on the Web
Getting context from web analytics is simply taking the metric out of isolation. Traffic metrics, like visits, by themselves are almost meaningless. So, you got 1,000 visits today. So, what? Is that good? Is it bad?
- Time Context -- using history to give your metrics context is the most basic way to make some sense of it. We got 1,000 visits today and 1400 visits yesterday. Now we have some way to make a better judgement call about good or bad. A different time context would be: we got 1,000 visits today and 550 visits on the same day last year.
- Average Context -- average is really just a spin on the time context, but it can be useful. Use it with caution, though. We got 1,000 visits today and our average for the last month was 990 visits per day. Average can give you context, but it can be misleading if you have spikes or valleys.
- Action Context -- actions per metric is a great way to get some context around your metric. If your site has a newsletter sign up then Sign-Ups per Visit is certainly a meaningful metric; watching it closely will change how you present information, how you hook your visitors, and maybe the size and color of the Sign-Up button. These metrics generally have diminishing returns, the trick is finding the sweet spot where you get maximum action for minimum input.
- Revenue Context -- if you can attach an income amount to your metric that is useful context (assuming you want to make money). Dollars/visit is pretty simple. Get more money per visit and you will make more money. Again, there can be diminishing returns here too, so don't operate with your eyes closed. For this to be really valuable you need to understand the costs associated too, that's why ROI (return on investment) gives you even better context than just $/visit.
- Competitive Context -- seeing what your competition is doing is more difficult, but highly valuable. We had a 1000 visits, our competitor had 400 visits. Good job!
Selecting the best context depends on what is most important to your business and why you are tracking it. Average miles per gallon was perfect for changing how I drive, but it didn't cause me to drive less. WHY matters.
Attention to Information
There are really two parts to giving your metric the appropriate attention: (1) Cutting the fat and (2) Periodic monitoring.
Once you have your data collection method in place you need to home in on the most valuable metrics. Usually, there are a few, but rarely more than five. There will be 100's of reports available to you, but only a few will give you actionable insights.
Average miles per gallon was the right metric for me. I could have selected a more sophisticated metric like $/mile driven. After all, that is really what mattered to me, the effect it had on my pocket book. Tracking $/mile driven would have caused a lag in the data and been too painful to be of value.
The other thing to consider is that I can't control gas prices, but I can control how I drive. I may drive more efficiently than ever before and a rise in gas prices will skew my data. You will have a similar problem with your web metrics. In an ideal world ROI is the the "end" of all metrics, but it can be complicated, like $/mile driven. For a lot of ecommerce businesses revenue/month, or revenue/visit, are simple, timely, and you have a fair amount of control over the outcome.
Finally, when you get your critical few metrics, watch them like a hawk, get periodic reports. Notice what happens when you change your marketing messaging, notice what happens when you do an email blast, notice what happens when you don't do anything. When there is a change investigate the cause, be relentless about discovering the activities that improve your metric. When you find something that works, do it more. Things are constantly changing, so don't stop watching.
If you are watching a metric and you don't find ways to improve your business consider whether you are giving it appropriate context and attention. If you are and it still doesn't change your behavior, maybe you need to select a different metric.