Onboarding flow | BorrowMyDoggy

Educating users on a unique proposition through sign-up and profile creation 

The challenge

BorrowMyDoggy offers a unique service: instead of dog owners paying for traditional dog care, they connect with local dog lovers who don't have their own dog. They then build a mutually beneficial relationship where "borrowers" get to spend time with a dog, and "owners" receive help with their dog care for free. BorrowMyDoggy supports and insures this relationship in exchange for a subscription fee. 

When I joined BorrowMyDoggy, the existing sign-up process was clearly a significant problem. It had numerous bugs and UX issues and failed to educate users about the nature of the service. This led to several key areas of concern:

• Many users completed sign-up thinking it was a standard, paid dog-sitting service

• Many users switched their profile type multiple times during the process due to confusion

• Profiles created were of poor quality, creating a negative impression of the product

• A lot of under 18s signed up and subscribed, only to be told that they could not pass the safety checks and so could not use the service

• It was common for borrowers to ask for payment, which was not allowed, and took a great deal of effort from the customer service team to police

First iteration

I designed first iteration of the sign-up process for the new mobile app and implemented several essential principles:

• Each step had one question.

• It featured an engaging, visual process with characterful illustrations.

• It provided education on the product and key principles.

• It effectively handled existing accounts.

• It incorporated robust error handling and reporting.

• It has accessibility at its core, ensuring an inclusive experience for all

• It featured a modern, map-based location UI.

• Confirmation steps were added to prevent common mistakes.

• The ability to enter conflicting information was removed.

• It enforced clear response to essential binary questions.

• It featured a modular approach, to allow for reconfiguration in response to data.

Moving to web

The new mobile app sign-up process appeared to be a significant improvement over the older web version, and initial feedback was positive. However, technical limitations prevented us from gathering any qualitative data on the app to make a fair comparison between the two platforms. Therefore, I designed the web sign-up process by using similar patterns to the mobile app to ensure a consistent user experience, knowing that we had far greater opportunities to gather data on our web platform.

During the design phase, I closely collaborated with the lead developer, Arthur, to create and fine-tune new form components. Arthur used React to prototype these components, and together, we focused on improving the user experience by paying attention to details such as input validation, feedback mechanisms, and error handling. Once these basic building blocks were complete I proceeded to design and specify the components and layouts that would form the entire process, paying attention to edge cases. I manually tested each of the components with screen readers to ensure that the experience was inclusive for those using assistive technologies.

The sign-up process consisted of two main steps: account creation and profile creation, which users had to complete one after the other. In an ideal scenario, I would have investigated whether delaying profile creation to a later stage in the user journey would yield a more user-friendly experience and increased subscriptions. However, implementing this change would have demanded substantial structural adjustments to our product before we could even evaluate its impact. 

Given our very small team that we had to break our work down into small projects that could deliver incremental value and so I made the decision to address this issue at a later date.

Testing, testing..

After completing the initial version of the new web-based sign-up process, I conducted a series of A/B tests to compare it with the old process. The account creation part of the new process immediately outperformed the old one, showing an 11% increase. However, the profile creation section experienced a significantly higher drop-off rate compared to the old process.

Recognising that the profile description step was the most challenging and that I had previously increased the minimum character requirement in an effort to enhance quality, I decided to address this issue first. With some reluctance, I reduced the minimum character requirement back to 1 character to align it with the old process.

Despite these adjustments, when we retested the profile creation, it continued to underperform when compared to the old version. The new process undoubtedly required more screens to complete due to the added educational content, error correction, and focus. Internally, there were discussions and challenges regarding whether this approach was the right one. However, I still believed that the benefits outweighed the concerns.

To gain more insights, we introduced step drop-off tracking into the profile completion flow. This revealed that the profile description step was still responsible for 60% of the drop-off in the profile completion process. This was a critical finding that we couldn't have assessed in the old flow because it combined the profile description with other form items on a single page. 

Profile writing automation

I realised our members needed more help in crafting their profiles, and the empty text box could be intimidating. So, I came up with a solution: using their previous answers to generate a description as a starting point. At the time public awareness of AI like ChatGPT was limited  and so I meticulously designed a set of rules to create natural-sounding descriptions from the users' previous inputs, both binary and textual, while ensuring grammatical correctness. I also took the opportunity to educate users further on how to use BorrowMyDoggy by modelling the ideal language and attitude to get the most out of the product. 

We then conducted A/B tests with two different approaches compared to the old sign-up process. The first version filled in the profile description text box with the generated description, giving users the option to edit it. The second version presented a "write it for me" call-to-action, allowing users to choose their preferred approach.

The outcome

The initial findings showed that the pre-filled text box improved profile completion rates by 10% compared to the old process, while the version with the call-to-action (CTA) only resulted in a modest 2% improvement over the old process. Although it was tempting to choose the pre-filled version, I was concerned about its potential effects on other members. To make an informed decision, I examined:

• The number of profiles with unedited generated descriptions

• The length of descriptions

I also took into account feedback from our customer service team, who reported concerns about "cookie-cutter" descriptions, which were resolved when members understood their origin.

By analysing the data I discovered that pre-filling the profile description discouraged users from creating longer, more unique descriptions. The proportion of unchanged descriptions was also significantly higher in the pre-filled variant. In conclusion, I decided that, although the profile completion rates were not as high, the CTA version offered a better overall member experience. I also made adjustments to the generated description to clarify its origin and alleviate concerns about any malicious intent.

This decision allowed us to proceed with replacing the old sign-up process, effectively resolving numerous technical issues and roadblocks. We addressed all significant bugs and accessibility issues with this completely new code and ensured that conflicting data could not be added by changing the way that it was requested. The new modular structure made it easier to make changes, and we revisited it multiple times to test and further improve. 

Not only did we accomplish the key business objective of boosting sign-ups, but we also received positive feedback from the customer service team regarding:

1. Reduced under-age sign-ups

2. Fewer requests for payment in borrower profiles

3. Increased user understanding of the appropriate profile type