Analytics are important for one simple reason: they help you understand how people use your product, so you can improve it. This leads to more sales.
In the real world you can learn all kinds of things just by observing: “The kids usually run straight to the pastries;” “People often bump their shopping carts into that display;” and “Tons of people are confused why the granola isn’t with the other cereals.” But online you can’t see anything, unless you’ve got the tools to watch and learn.
Analytics tools offer a view into how your customers experience your product. But collecting your data and finding the right suite of tools to learn from is a huge challenge. Here’s the roadmap we use when advising our customers how to get their analytics in order:
Start with key metrics. It’s easy to get overwhelmed by the amount of data you could collect and analyze. Don’t go nuts. Start with three, and only three, metrics that best indicate the health of your business. These key metrics will be the core representation of your business, and you’ll build all kinds of things to improve these key numbers.
There are two types of key metrics -- let’s go into some examples.
The first are “totals.” Facebook’s total metric might be the number of active users within the past 30 days. Dropbox might measure the total number of pro users.
Unfortunately, it can be hard to avoid BS metrics with totals. Totals give you sense of the scale of your business, but it’s hard to know whether things are getting better or worse.
The second (and usually better) type of metric is a conversion rate. These are excellent metrics to track because they show conclusively whether you are improving your business or not.
An ecommerce company might live and die by their sales conversion rate -- the percentage of people who visit their site and complete an order. Entrepreneur.com probably cares most about their page views per user, since page views are how publishers price ads.
Measure your metrics. Now that you’ve chosen your key metrics, you need to collect the right data to measure them.
Conversion rates measure the percentage of users who performed a certain action (such as viewing the homepage) and then later performed a second action (such as completing an order).
The best way to measure conversion rates is to use a technology called event tracking. Events are things people do on your website (such as “viewed checkout page”). To set up this kind of tracking you’ll need a developer to add code to your website or app, so that each time a user performs one of these actions, your analytics tool of choice will be notified.
Like most other software-as-a-service companies, we at Segment are chiefly concerned with how signups become paying users. To measure our key metrics, we track the following events in a customer’s lifecycle:
- Signed up
- Created project
- Sent project data → they have activated our product
- Enabled integration → they used Segment to integrate and send analytics data to a third-party tool, the whole point of our product
- Started subscription
- Upgraded subscription
- Downgraded subscription
We use these events to calculate our key metrics:
- Activation rate: Percentage of users who “signed up” and then “sent project data”
- Conversion rate: Percentage of users who “signed up” and then “started subscription”
- Churn rate: Percentage of users who “started subscription” and then “downgraded subscription”
Run experiments. Now that you know what’s important to your business, and how to measure it, you can start running experiments to improve your numbers.
If your activation rate is important, you could try to simplify your user onboarding or run an email campaign to up that metric. If signup rate is low, you could try adding a customer testimonial on your signup page, and see if that helps. If your churn rate is high, you can run a survey to ask people why they stopped using your services, and then use that information to create a new experiment.
Once you know what’s important from your key metrics, you can think of a lot of ways to improve them. Event tracking lets you determine, with confidence, whether these experiments worked or not.