Join us 6/23 to hear about "The Insights-Driven Organization" with Forrester

How We Used Behavior-Based Onboarding to Improve PLG Conversion

This story is also published on Medium.

In this ongoing series we document our Product-Led Growth (PLG) wins and losses and share our learnings with the PLG community.

We’ve always believed in the power of our product, but in late 2020 we made a company decision to pursue Product-Led Growth (PLG) in earnest. As we deeply believe in making data-driven decisions—in the power of hypothesis, testing, and iteration—we’ll be sharing our various experiments with the world.

We started by breaking our PLG efforts up into three related KPIs:

  • KPI 1: Increase signup conversion from our marketing site
  • KPI 2: Increase activation rate among people who have signed up
  • KPI 3: Increase conversion rate from trial to the paid version of our product.

In a more traditional SaaS model, we’d assign the first KPI to marketing, the second KPI to product, and the third KPI to sales. But we’re doing PLG here! In PLG, organizational divisions like these tend to fall apart. Our goal in putting this initiative together was in part to break down the invisible walls between product, engineering, sales, marketing, and customer success.

We’re loosely tackling these initiatives in order, though there’s certainly plenty of overlap.

In this post we’ll focus on some key learnings from the second KPI: Increase activation rate among people who have already signed up for the trial.

What’s the goal?

Most simply, our primary goal is to get users to value as quickly as possible.

This goal is motivated by the imperatives of PLG, as PLG generally aims to get users into your product and show them what it can do immediately, so that users are motivated to buy without having to suffer a protracted sales cycle. In some ways PLG flips the standard SaaS model on its head: instead of using your sales team to pitch a vision of what your product can do, in PLG you put your product into people’s hands right away and let it do its magic.

If it works, PLG gives customers a short, self-service sales cycle, and lets companies save resources. PLG also creates the “flywheel” effect: once it’s working, it’ll acquire customers on its own, without (ideally) heavy intervention from your team.

There is a catch, however. Your product needs to deliver. If customers aren’t getting to that “aha” moment quickly enough, they will certainly not stick around.

This second KPI is all about making sure our product delivers.

Mapping our self-serve experience

To go about this systematically, we started by mapping out the main steps to onboarding and activation in our product.

If you’re just starting out with PLG, you’ll need to home in on the one or two user actions that best capture getting value in your product — the things that provide the strongest signal that a user is using your product to do something valuable for them.

To do this, we adopted a framework borrowed from Reforge, who encourages identifying three separate action steps:

  1. The Setup Moment
  2. The "AHA" Moment
  3. The Habit Moment

In Heap, we define those three moments as follows:

  1. The Setup Moment = customer installs the Heap snippet in their product or site
  2. The "AHA" Moment = customer defines an event in Heap
  3. The Habit Moment = customer runs a query using an event they’ve defined

In addition, as we explain in our Guide to Retention, over time we’ve found that in the Heap product, “sharing a report” is the user action that best correlates with long-term retention.

Putting these pieces of information together, we mapped out the main steps from “sign up” to “share a report.” The following is what we aligned upon. This would be the journey we’d try to improve, step by glorious step.

PLG onboarding stepsThe main steps in our onboarding funnel

Setting baselines

Now that we had our funnel laid out, we had to figure out the conversion rate at each step, to get baseline numbers.

Luckily, we have a tool that is amazingly good at this. It’s called Heap!

Because Heap automatically collects all behavioral data from your site, getting data on each of these events was pretty easy. We just pulled up a dashboard and started measuring.

What we found was no surprise. The further a user goes into the funnel, the more likely they’ll be to share a report, which makes them more likely to stay with the product for a longer period.

Behavior-based emails vs time-based emails

Now that we had our funnel set up and our baseline measurements, it was time to hypothesize and test! Our first experiment: set up behavior-based email campaigns to see if we could improve conversion at each step.

Why are behavior-based emails a better strategy?

Well, we used to use a time-based email cadence. On day 1 after signing up, you’d get a welcome email. On day 2 you’d get an email about installation. On day 4 we’d send something about defining events. And so on.

It’s easy to see the inefficiencies with this approach. Mostly it assumes a standard cadence for customers working through the product. In reality this rarely matches people’s experiences. Maybe customers have too much on their plate to install Heap the day after signing up. Or maybe they’re moving fast, and by day 2 they’ve already defined an event and created a report. In that case we’ll just be clogging up their inbox.

With behavior-based emails, you segment your users into behavior-based cohorts — groups of users defined by actions they’ve taken or not taken in the product. Then you set up a special email series for each group.

Luckily, we have tools that can help. The first is, yes, Heap. The second is Marketo.

Want to learn more about cohort analysis? Read our guide here.

Leveraging the Marketo connector

We recently launched our Heap/Marketo integration, which has proven extraordinarily useful for exactly this style of initiative. The idea is this: using Heap we identify key behavior-based cohorts. Then in Marketo we set up a series of emails unique to each group.

For example: we created a cohort for “users who have signed up for Heap but haven’t installed the product.” They get a how-to-guide, and then an email encouraging them to contact support. We set up a cohort for “users who have installed Heap but not yet defined an event.” They get an email about setting product KPIs, and then a guide to defining events. And so on.

Finally, using the Heap/Marketo integration, you automate the process, so that when any user triggers certain thresholds (for example, they’ve installed the product but have gone a week without defining an event), they’re automatically placed in the appropriate Marketo campaign.

It’s this automation part that’s at the heart of any good PLG strategy, since it lets users receive the help they need without you needing to get involved. The more the onboarding experience can run itself, the more efficient you can be.

Building in Marketo

Getting the logic together in Marketo took a few days to get right.

First, we took all of the segments in Heap and synced them to Marketo by simply toggling the “sync to marketo” switch on each segment. This then allowed us to make triggers based on people going in and out of these lists. For instance, we had a segment in Heap that had all people who signed up but didn’t install. As long as they remained in that list, they continued to get emails regarding how and why to install. But as soon as Heap reported back to Marketo that they did install, they would be automatically taken out of that list and placed into the next stream of onboarding emails.

We also did some brainstorming around what the right email sequence for each group should be.

The results

We launched this new process in December 2020. After running for a month, we took a look at the results. They were encouraging:

Conversion liftsIncreases in conversion across the funnel

Overall, from signup to share a report, those who opened the email converted 10% more!

This is the goal of behavior-based onboarding: each group receives the message that’s appropriate for them, and that’s most likely to push them into the next stage in the funnel. And, given that we’ve now automated the process, we can work on improving it while we also focus on other things to optimize.

More tests to run

As we move forward, there are many more tests we’re planning on running, across our PLG process. (See the three initiatives above!) For this particular experience, we’re planning to play with email content, messaging, and sequence, to see if any of those increase conversion at a better rate.

Stay tuned for more!