The following is a guest post by Peter Reinhardt, co-founder at Segment.io and author at Analytics Academy. He loves startups, math, physics and lots of data!
Growth hacking and data-informed product development are all the rage. But getting started is a bit daunting. This article will show you how to put together a simple, great analytics setup, with some tips to avoid the most common mistakes. It'll put you on the fast-track to measurably improving your product.
The most common question I see at Segment.io is "Which analytics service should I use?" What we've found is that different analytics tools are helpful for different parts of your "funnel". Most companies end up relying on three or four analytics tools to get a complete picture.
When a visitor first comes to your site they'll skim your landing pages and check out your company and product. A small percentage of them will sign up, usually 1-5%. A bunch of them, 50-80% usually, will just leave right away.
The first questions to answer here are
The best tools for this part of the funnel require zero setup. Google Analytics, Chartbeat and GoSquared all give you this information in real-time, for free or dirt cheap. If you're a marketer running ads, or writing blog articles as content marketing, you'll live and die on these numbers.
But if you're a really good marketer, you'll dive a bit deeper and make sure that all this traffic is actually producing value for your company. After all, lots of traffic isn't worth squat unless some of those people pay! So really good markters measure "Conversion to Signup".
Google Analytics is the best tool for measuring your signup conversion rate. The trick is to simply record a custom event when the user signs up:
// normal Google Analytics _gaq.push(["trackEvent", "All", "Signed up"]); // or with Google Analytics enabled via Segment.io analytics.track("Signed up");
This is an enormously powerful set of insights, because it shows you exactly which marketing efforts are working, and which aren't. Optimize accordingly!
Quick tip: if you're recording the "Signed up" event right after a user submits a signup form (makes sense, right?) then make sure your event tracking request gets off the page. This is tricky, because you'll have to
event.preventDefault() the submit event, and then re-trigger the submit event after 300ms or so, when the custom event has had time to be sent to Google's servers.
The simplest way to avoid this is to use the open source library analytics.js. Analytics.js has two utility functions
analytics.trackForm() that handle the event binding and delays for you.
Measuring the top of the funnel will streamline your marketing efforts, because you'll see exactly which referrers and content are working to really convert visitors to signups. Once you've done that with Google Analytics, it's time to dive deeper into your funnl.
After a visitor signs up for your product, you need to activate them into a dedicated (paying!) customer. Most sites have some kind of setup process, like downloading, installing and logging into a desktop app... uploading a profile picture, setting your bio, and finding some friends... A signed up user is no good unless they've really started using and getting value from your product.
By tracking conversion rates in the setup process, you can figure out where to improve the setup experience, and how to help your users succeed faster. The key here is to record an event at every step of your setup process. For Segment.io, that process looks like this:
From step 1 to step 5 we see 40% conversion. But that wasn't always the case! At first we only had 9% conversion. The silliest part was that we only had 60% conversion between step 1 and step 2, because we had an email confirmation. So we moved the email confirmation, and conversion on the first two steps jumped from 60% to 97%. Then we improved the setup guide copy, and added more libraries to choose from. That more than doubled conversion on steps 2-5 from 15% to 40%.
The two best tools for looking at this kind of "mid-funnel" data are Mixpanel and KISSmetrics. They can let you analyze arbitrary funnels based on event data, showing you the conversion rates at each step of the funnel. If you want to get more qualitative feedback about why people aren't converting, grab an Olark or Qualaroo account and pounce on users who pause in your funnel. Ask them if they hit a snag, they'll tell you!
As soon as you start tracking the steps in your own setup funnel, you'll find all kinds of ways to make the conversion rates go up.
The "bottom of the funnel" is retention and referrals. These are both super juicy topics, but too much to cover today. We'll write more about these topics at Analytics Academy. Instead, let's talk about about how to actually improve your top and middle funnels.
At the Growth Hacking Conference last week, the legend himself Stan Chudnovsky talked about how to experiment successfully with conversion rates. (Stan is the guy who first came up with "Connect to your Email to see who else you know", and a bunch of other brilliant hacks we all know now.)
His first recommendation was to understand your customer as well as possible, and think deeply about their psychology. How can you massage their ego? communicate the right thing at the right time? One of his examples was PayPal. Most people have a $0.00 balance in PayPal itself, but that's connected to bank accounts with large balances. In the PayPal mobile interface it shows in huge glaring letters "Balance: $0.00". People don't want to complete a transaction when PayPal tells them they have zero balance! If you're lost for how to improve conversion, get yourself into your users shoes and use your product as they would. It will become obvious.
His second recommendation was about managing your own psychology. If you run an experiment and it fails to produce positive results, it's super easy to say "screw it, that's as good as it'll get." But that's where Stan points out the obvious: 4/5 experiments will fail. Keep motivated, keep trying new things, and expect to fail. Don't let yourself stop experimenting just because the first four attempts failed. Use those failures to learn... what's the fifth experiment that'll give you that 20% boost?
The great secret to growth hacking at big companies is that they can run 50 experiments in one day. They just have that much traffic. You don't have that much traffic, so your experiments will have to run for 5 days to be significant. Don't get down, power on and keep trying new things!
As you get your own analytics systems set up, make sure you're using the right tools for the job. Google Analytics is great for the top of the funnel, and Mixpanel/KISSmetrics are great for mid-funnel conversion analysis. Keep experimenting with new ways to improve conversion at each step. If you learn something amazing, or even something face-palming obvious in retrospect, I would love to hear about it! Email me or follow me on Twitter.
Posted by Peter ReinhardtLinkedIn Twitter Website