Tray

Web AppDashboardOnboarding

Tray is a workflow building platform that helps you automate your or your company's repetative everyday tasks.

The Team

Product designer, Product manager, Engineering manager and 5 Engineers.

The Problem

Newly onboarded users don’t know where to start from and don't know how to build on Tray which leads to low engagement. We also have a large support team that is overwhealmed with simpler questions that take a lot of their time.

The Goal

Introduce an onboarding experience that would show and teach newly onboarded users how to use and build on Tray by increasing their user score and stickiness by reaching a good conversion rate of the onboarding experience.

Automation applied to an inefficient operation will magnify the inefficiency.
- Bill Gates

What we decided to do is to understand and fix our process and problems before we bake it into the app and automate it.

The Approach

I already knew that the user score of our users and their stickiness is based on how much they know about the platform and their ability to build on Tray by looking at the data coming from Tray Academy. (Tray academy is a place where you can learn the basic concepts of building on Tray. ) We also know that contextual hand held experiences make it easier for users to understand instructions. Interrupting at each step insures that they remember each task better which has been observed in the Zeigarnik Effect.

Meanwhile I started interviewing some of our newly onboarded users and started looking into the types of apps and workflows they usually try to build. I also wanted to understand how our user score is defined and what actions are needed so that a user is successfully onboarded.

That helped me understand some of the pains but also helped me define what a well onboarded user looks like. The user score is defined by a lot of different things like, how many workflows you’ve built, have you used connectors, or specific operations. Have you used the documentation or Tray academy. Depending on your score you would fall into three different groups of people:

Anyone around score 200 and above builds quite complicated workflows and most of them use connectors or services such as Salesforce, Pipedrive, Hubspot, AWS and many more.

Anyone below 150 hasn’t built anything useful but is highly engaged, and they usually try to build something they need, they use the documentation and the Academy.

Users below score 20 are usually just looking around and comparing tools or just want to see the tool and it’s capabilities, but they drop off quite quickly.

These three groups also mapped quite well to the three main personas I've defined earlier, which are mostly based on user’s technical skills.

Based on all that information we came up with the hypothesis that if we show them a tutorial that helps them build using the main concepts and tools in Tray, then their user score should go up and make them more sticky and engaged. We came up with few concepts which I ran usability tests on (can't share the report as archived projects on Maze are locked unless you upgrade your plan 😭) and we rolled out the first experiment to 50% of our trial users.

As with any experiment that didn’t really go down extremely well, but gave us a good start and a better understanding of what worked and what didn’t. Having all the feedback also helped me put together a much detailed journey map based on the technologist persona that we were focusing on.

It's unclear what the goal or purpose of the steps are. A quick background before the step-by-step guide would be useful
- Sarah

Conversion wasn’t great for few reasons that we found out by interviewing and getting an in app feedback from some of our users...

The tutorial was way too long
It was too descriptive and not focusing only on the task itself
Not each task was tailored to the technical level of the user
You coudn’t easily close and come back to it

Based on that I kicked off another workshop with the squad to explore “how might we’s”, map the experience we’re focusing on and come up with some ideas using the crazy 8s framework.

We decided to focus on some quick wins like reducing the length of the tutorial and decoupling the what from the why.

These changes put us back on track and improved significantly the conversion rate of the tutorial. Most of the people that decided to progress with the tutorial also completed it which was a signifficant improvement.

Now that we knew the conversion was better, we focused on showing the right tutorials and make sure we give information about how difficult a tutorial is, so less experienced users can start off with the easier ones or the ones that are more relevant to them.

I ran a quick card sorting exercise and see how we could fit some of the main concepts into more tutorials that could be grouped into themes as well as figure out the order in which we do them.

We ended up with quite a few ideas on how we could expose these tutorials even better, but for the time being we still wanted to experiment and keep our focus. We made the decision to reuse our modal component, so we can test the concept as quickly as possible.

What we've learned

By delivering small experiments quickly we were able to learn a lot of different things. We understood how the user score can be helpful for us, we were able to define what the onboarding for Tray could be and set a good ground for onboarding to be properly integrated into the app. Another interesting finding was to find how much templates were important for people to understand and build on Tray. Onboarding and tutorials looks like is going to be just one piece of the puzzle along with improving the properties panel, so it's more user friendly and more instructive. Tray is a very complicated tool, that requires a very different approach, which is also why experimentation early on is a must.

The impact

What's next?

We have still loads to do and a lot of ideas on how to move forward. We already had customer feedback pointing us to add an additional walkthrough of the main elements users are interacting with. For instance, what’s a connector, or what’s an operation… what do they do… before we even drag them into the tutorials.

Also we want to make the documentation more contextual to where they’re stuck and offer a quick and easy way to get back to where they’ve left off.

We need to remind them about unfinished tutorials and lastly we need to start delivering templates that are relevant to their industry and level.