How adopting a UX mindset changed the Shopify Help Center

Three ways we used UX thinking to improve customer support

This article was co-written with Lina Tovbis.

Shopify’s Help Center is massive, with a wide range of resources crafted to help Shopify’s users face the challenges of operating a business. As Shopify Support scaled to match the company’s product growth over the past few years, we began to recognize an increasing need for UX thinking to build the best support experiences.

When we looked at user feedback, support ticket dives, and search behavior, we could see a disconnect between the language and terms used by Gurus (what we call front-line Support team members), users, the Help Center, and Shopify’s core product. Likewise, data made it clear that the Help Center suffered from a content discoverability problem: the Help Center’s UI was underused, and search did not show the most relevant results as often as it might.

We could see that the Help Center was bursting with useful information, but it could only be great if it was usable, discoverable, and tailored to meet the shifting needs of our growing user base.

Working as a systems designer (responsible for researching, analyzing, and designing high-value support experiences) and as a senior operations analyst (similarly concerned with understanding and improving support services), we are deeply familiar with Shopify Support, and we collaborate regularly with the various teams that influence support development.

To help us succeed, we started incorporating UX best practices. We started by experimenting with things like heat mapping, surveys, user interviews, and A/B testing to find out what research methods work for support. In addition to “learning by doing”, we paired with UXers at Shopify, took relevant UX courses, and attended UX conferences.

One of the first things we did was look closely at support data to understand how users experience Shopify’s Help Center. The more we looked, the more questions we had: how do we know what valuable support means for Help Center users? What does success mean for a support resource? Does success mean the same thing for Shopify’s users as it does for Support’s business obligations?

The more we learned, and the more questions we asked, the more feathers we ruffled in Support — this was how we knew we were on to something. With this in mind, here are three of our projects highlighting how a UX mindset can improve outcomes for people working on support experiences.

Making the most of user feedback

One of our first successes was a redesign of the Help Center’s user feedback form. Each page of the Help Center features a feedback form that lets users tell us how helpful a documentation article is, as well as what successes and struggles they had in their support journey to that point.

The user feedback form has changed a few times over the years: at first, it was a binary yes/no form, but the data associated with this version of the form was incomplete and regularly incorrect. It included little insight on why users rated pages they way they did, with no qualitative input to complement the quantitative score assigned by users. The form’s data compromises made it hard to work confidently with its results.

To add detail to the form’s results, we adopted a five-point scale measuring user sentiment. This form also introduced a qualitative component to gather a wider range of user insights. We hoped that by pairing qualitative feedback with more granular (but still quantitative) sentiments captured by the five-point scale, we might start to report more specifically on documentation helpfulness.

The five-point scale worked well enough: it saw decent user engagement, it gave a sense of user satisfaction with Help content, and it offered a glimpse of common user struggles with specific articles.

To understand the qualitative feedback users were submitting with the form, we applied a manual language trend analysis model to identify recurring trends across anonymized feedback submitted over time. We reviewed thousands of feedback items, noting recurring language across comments to identify common trends and user struggles with specific documentation articles.

For example, whenever multiple users would speak to struggles finding a piece of information that exists but is not clearly associated with a documentation article, we would flag the ways that documentation suffered from content discoverability issues. Likewise, if feedback showed multiple users speaking to an article’s lack of clarity, we would flag the page for not reflecting the documentation team’s style requirements. With time, we identified dozens of recurring, actionable feedback trends.

After completing this manual analysis, we generated reports with recommended actions for the documentation team to resolve user issues with Help Center sections and articles, prioritizing critical trends based on how commonly they were noted from feedback. We would share these reports with technical writers, and we would follow up with secondary reports some time after changes were applied, to understand the impact.

Despite offering some good insights on user struggles, the five-point form was imperfect: in particular, it became clear that the five-point rating scale was unnecessary. Most users rated pages either high or low, with little engagement in the middle.

Similarly, the form’s qualitative prompts were very open ended, leading users to leave a lot of general comments. Many users claimed pages were either “great!” or “the worst.” This data was hard to translate to actionable insight for technical writers, leading to a lot of feedback labeled “general positive” and “general negative.” Although actionable trends did exist in more specific user comments, parsing them was fully manual: labor intensive and time consuming.

Taking these shortcomings into account, we reconsidered the Help Center’s feedback form. We wanted the new form to provide data requiring less intensive trend analysis. We returned it to a Yes/No binary and expanded it with multiple-choice options based on trends parsed from the previous form’s qualitative results. To add additional specificity to the form’s results, we ended the form with a qualitative input for users to describe what they did or did not like about their experience relative to the multiple choice option they chose.

By funnelling users from a simplified binary rating, through multiple choice options mapping to top feedback trends, to a qualitative input informed by the selected multiple choice option, we aimed to produce more specific results.

On release, the new form was a success: in the first months after shipping, engagement with the yes/no options more than doubled the number of ratings recorded by the five-point scale. Trends recorded by the multiple choice options mapped exactly to trend percentages recorded by manual analysis of past results, and the new form successfully produced a simple helpfulness percentage for each page. All the while, analysis of qualitative inputs continued to expose actionable trends based on user reception.

To improve how teams used these feedback results, we worked with the Support Data team to produce a reporting dashboard . This centralized the results and it reduced the analysis needed to draw insights.

Combined, this provided stakeholders with more context than ever about how users feel about specific documentation articles.

Personas and journey maps

With the success of the feedback form, we were asked to explore another project rooted in UX: understanding the customer service support experience using customer service journey mapping and support user personas.

Creating a journey map can help us understand customer experiences with our products and services across different support channels. This way we can find where customers have difficulties with the support experience, or where there are gaps that we can improve.

Customer support user personas remind us that we are not the customer and that not all customers are the same. Having personas is crucial for knowing how to help these customers and they help employees remember who the customer is by sharing details about them.

In the first phase of the project, we were able to:

• Come up with an initial set of merchant personas (6 in total) to create journey maps for. There were some personas that already existed at Shopify, but they were not focused on the support experience.
• Create a map of merchant touch points with support.
• Start identifying areas of high effort or areas where there are gaps in the service experience, including trying to understand which personas we might be under-serving.

This was the template that we settled on for our personas:

Because every merchant is different, we could have come up with thousands of personas and tweaked them to match all of those different merchants, but we wanted to limit ourselves to a few personas so that we can have a general understanding of particular support users. We also wanted to focus on users’ support experiences that may differ based on certain traits. To do this, we came up with some defining characteristics to create our Support personas. Some of the characteristics that we included were:

• Merchant type — are they the store owner, or are they a staff member?
• Support tools that they use — what are their preferred methods for learning or gathering information?
• What makes a good day — how do they currently interact with Shopify? What is working right now? What’s causing a high-effort experience?
• Personality — what are the problem-solving behaviours we should consider? For example, willingness to be guided. For our personas, we referred to these four personality profiles.

For the second phase of this project, we focused on research and data. We broke out into smaller working groups made up of people from Support, UX, Service Design, and Data, and picked areas of focus:

• Continuing to develop existing personas. Mainly, focusing on personas that might be underserved or missing entirely, and understanding which personas might be having the most high-effort experiences with support.
• Working with the Data team to make sure that our personas and business models are validated by data.
• Setting up interviews with Gurus (Shopify’s frontline support staff) to review the existing personas and learn about how Gurus would interact with each persona.

We used data from the annual merchant survey that the UX research team puts together to add more details to our personas and make sure they align with what real merchants say about their experiences. We were happy to see that our personas lined up quite accurately with the demographics and business models in the annual merchant survey.

We interviewed 12 tenured Gurus and had them review and tweak the personas to reflect their experience helping merchants. We talked about which personas they encountered most and least often, which were the most and least complicated to help, and which might be missing. We also asked participating Gurus to share advice for new Gurus still learning how to best help merchants.

So many great insights came from these interviews. They allowed us to pass feedback along to our Quality Assurance and Training teams to create more training for new Gurus and to understand what continuing education resources we could provide our more tenured Gurus. Advice for new Gurus included:

• Go into each interaction fresh and without any biases so you can support it properly. Conversations will be very different merchant to merchant.
• Let the merchant have the space. You might think that you need to have low handle times (the time spent on a chat), or wrap everything in a bow, but let the conversation evolve.
• Establish emotional states before starting troubleshooting. Instead of asking how they got where they are, establish how they’re feeling and how they’re doing.
• See their question, and really understand it. If you don’t understand it, ask them. The idea of asking questions is under appreciated.

Building on these personas, we worked to map out a specific support experience. For the map itself, we included the different phases of a Support journey:

These phases included:

• Feeling — How does someone feel during a Support experience? Where do the peaks and valleys form?
• Thinking — What’s going through a person’s mind during a stage of the journey? For example, when they click a button, they might think: “Did that work? How do I know that this actually worked?”
• Doing — What is a person’s environment as they work through a specific task or decision? For example, are they sitting at their desk with the company website open on their laptop?
• Needs — What does a person need at any moment in a support experience? This can help us to understand expectations and assess whether or not we’re meeting user needs.
• Opportunities — Where can we, as a company, reduce the valleys and increase the peaks for support users? To answer this question, it is best to work in cross-functional teams to avoid bias, and to ensure that the right teams are addressing issues in the best ways. For example, is this truly something that support can help with, or is this an issue with Shopify’s product?

We were able to come this far because of the incredible team that we work with. We had a lot of different backgrounds and areas of expertise from teams like Knowledge Management, Documentation, Support Research, Support Technology, and UX. We also had a couple of current and former Gurus providing their feedback. Because of this, we’re at a point where we have a strong base to work from, with a solid set of personas, and a detailed journey map to move personas through support scenarios.

Moving forward, we want to interview merchants to have them go through a specific support task and document what they’re feeling, thinking, and doing at each stage. We would eventually like to build a library cataloguing different journeys and then look for ways to improve the merchant experience.

Authentication

Finally, we were asked this past summer to understand what authentication means for support users, so we can insert authentication points to the Help Center in ways that set us up for future UX opportunities.

With authentication, users log in to the Help Center, which tells us about them and the stores their account is associated with. It also enables us to collect a range of information about a user’s session behavior. Combined with account data, we can use authentication to understand and improve user journeys through support. It also lets us expedite live support contacts by passing information about the user from their authenticated self-help session to live support agents.

To develop authenticated experiences, we worked to understand the current user journey to live channels. This helped us to decide where we could add authentication points without introducing undo friction to the chat, email, and phone funnels. We worked with the Support Data team to ensure we collected performance data when the authentication points were released.

After applying authentication points to the Help Center, we completed an analysis of the current state of authentication. We assembled existing data and research resources related to authentication, and worked closely with UX to study the user’s journey through to live channels. We shadowed Gurus who use live channels and interviewed them about their experiences, using the interviews to map the live channel experience for Gurus. We also worked closely with Support developers and Support Data to understand upcoming technology changes related to authentication.

Pairing a current view of authentication with what opportunities will open with technology changes, we helped to build an exciting roadmap for Support user authentication and personalization. The first phase of this work is in development, and (hopefully!) future projects will build on the personas and journeys work to route authenticated users to the information they need, exactly when they need it.

With every project we complete, we’re learning more about the value of UX thinking, and we’re showing its value to stakeholders. The user feedback form helped us to understand user needs in more detail, pushing technical writers to deliver even more valuable support content for Shopify’s users. Support personas and journey maps helped us to understand user journeys through Shopify support, empowering us to identify and communicate user pain points so we can improve support systems. Authentication pushed us to understand and design Support UX as part of Shopify’s broader product offering, setting the infrastructure for a more unified relationship between Shopify and Shopify support.

Each of these projects has seen increased buy-in from key teams and stakeholders. We find ourselves with a growing list of UX work, ranging from user research requests (like card sorts, user interviews, and surveys) to design work (including mock prototypes and design system development) and collaboration with product owners and developers. We’re involved with everything from data strategy to planning and conducting experiments. The timing is good: as more people use support, and as Shopify continues to expand its product to new markets, the need to understand how support can help different users is only going to get more complex.