MarcoPolo World School

World School is MarcoPolo Learning’s flagship product. It’s a cute, educational app full of high-budget, PhD-backed content for children aged 3–7. Unfortunately, some parts of the experience weren’t living up to the polish of the shows and games.

During the time I was there, the company was going through a period of growth. Investments were coming in, we were getting media attention, and our engineers and growth specialists had just implemented comprehensive tracking analytics in an effort to identify problems with the overall user experience.

In reality, it was a long process. The flow for a new user started with an un-skippable onboarding, proceeded through a profile creation, hit a paywall, age gates and account creation and activation all before getting to use the app (at which point, an in-app tutorial begins).

Restructuring the flow for a smoother experience

Ideally, I’d wanted to test two versions of the original sign up flow, unaltered other than the screen order. I hypothesised that by moving the payment screen nearer the start of the process, users would have more energy to consider committing to a trial, and avoid the frustration of hitting a block after investing their time. It makes us look more honest upfront, and should also filter out lower quality users at the start, helping identify more meaningful drop-off issues on the following screens.

As we had next to no user testing budget, I reached out to the developers about A/B testing, but the product was built in a way that didn’t allow this with different screen orders. Instead, we decided to incorporate my suggested changes into the redesigned flow, and monitor differences between that and the original flow. Admittedly, it wasn’t a highly controlled test, but we were moving fast and could revert if the stats didn’t prove my theory.

The pricing screen itself had also turned out to be a problem. Users were complaining about confusing options, unexpected charges and information overload.

After discussing with the CEO, we decided to reduce the options to just a monthly and an annual plan. It’s a structure many users are already familiar with, and should reduce decision fatigue.

I made the design much cleaner, highlighting our awards and credentials, showing the monthly costs at a similar height for easier comparison, and drawing more attention to the trial CTA. We also made use of the exit dialogue, which previously had unclear microcopy and provided no reason to stay.

At this point, I believed that reducing confusion was the key to winning users.

Designing for privacy and transparency

Originally, the first steps of the profile creation immediately asked for a child’s full name and date of birth—to some users it felt more like a legal form than innocuous personalisation.

When it came to requesting details, I broke the step into two stages to avoid overwhelm, and offered using a nickname to be less invasive. Additional context reassures the user that there’s a legitimate reason for collecting this data.

In a similar vein, we now only ask for a child’s birth month and year, rather than the full date. We needed this for roadmapped features that change the content depending on a child’s age, and it helped our education experts assess the suitability of the materials for different ages. I looked at more user-friendly options, such as selecting only an age in years, but there’s a big difference between the development of a child who has just turned 3, and one who is about to turn 4.

Adjusting the emotional journey through tone

It was clear from user reviews and complaints that the sign-up felt like a chore. I addressed this across the flow with more enthused copy and fun new visuals, playing on the personalisation aspect of the experience. I also added a progress indicator to show exactly how many steps to expect from the start.

I rearranged the stages so that avatar choice came first, allowing parents to “meet” the Polos characters to see which personality might appeal best to their child, and changing the visuals to feel more like choices in a game UI. Our illustrators and 3D artists had also created some polished new renders and icons for our recent brand refresh, which along with the bright new colour palette, contributed to a higher quality, more dynamic feel. It also allowed the selected Polo to “guide” users through the rest of the flow.

I proceeded to use the child’s name/nickname throughout the flow, to elicit an invested feeling. I also addressed several visual and interaction consistency and quality problems that had contributed to a less polished (and less trustworthy) feel, as well as some accessibility issues. For example, the same sound effect had been used for selecting options and progressing to the next screen, icons were not using the same visual weights, and the disabled CTAs behaved differently on different screens.

The progress screen was actually not a loader at all—it was a timed animation to appease labour perception bias, and appear to be better personalising the app. This is a benevolent deception, designed to increase our product’s perceived value. It did, however, feel dated and sluggish. I addressed this by removing the groggy progress bar, including the child’s name, and creating a more dynamic animation with our updated icons.

We also decided to remove the email confirmation field. It was important that we collected correct addresses, but having to type it twice on a mobile device is a lot of work, adding negative friction at the end of an already tiring process. Instead, we added a simple yes/no check after submitting the form, with verification being prompted once reaching the admin area of the app later.

Assessing the impact

The new flow was ready to be put to the test; as it would be our new baseline for future iterations, we wouldn’t be able to make granular comparisons at this stage.

Instead, we were hoping for some broad improvements to emerge—namely, for an increase in our install-to-trial conversion, and a reduction in drop-off at the profile creation screens.

One month after implementing the redesigned sign-up flow, we were seeing marked improvements in the data.

It seemed that users weren’t so put off by the name and birthday screens any more, and we were even seeing more people passing the age gate (we’d changed it from solving a maths problem to entering a birth year to be more inclusive).

Because users had already committed to a trial and subscription, they appeared to be more invested in the app, and less reluctant to enter information.

We were, however, still seeing a lot of drop off at the payment screen. Our growth analyst had predicted an increase (moving it forwards means a higher number of less-invested users would reach it), but this was more than we’d expected.

Iterating on the payment screen

Since implementing the new sign-up flow, I’d been working on some data-driven advertising with our marketing team. It was interesting to see what kind of tone and visuals influenced out target market the most, and it was surprisingly different from our in-app experience. For example, parents of young children responded best to lifestyle photography, whereas our “selling” screen was very clean and informative. Of course the product itself shouldn’t look like an advertisement, but the goal was the same—to get users on the trial. Perhaps this stage needed to be more emotive?

Out next release was coming up, so I suggested we make some edits to the design. I created a banner featuring lifestyle photography with 2D illustrations from the app content, and made some subtle changes with each iteration.