Little Language Models

I’m at a client’s offices for a day of interviews and information gathering. I’m all ears, and while I’ve prepared a list of questions, I especially try to pay attention to what comes out in the in-between spaces, in the silences, smiles, and hesitations.

Some of the questions I ask:

• What’s your role?
• How does the website support your work?
• Who does this site serve?
• How is content created?
• What does the user come to the site for?
• What does your ideal future state look like for their business and content?
• What content issues have you seen firsthand or heard about from sales, customer support, or other teams?
• What’s the most valuable content on the site? What’s missing? What’s confusing? How do you know?

I try to do the initial interviews 1:1 rather than with multiple people to avoid them affecting each other’s responses, and to promote open discussion, which in some work cultures, can be difficult to do with leadership or other coworkers in the room. I’ve written about the ROI of user research for content, but I’d be happy to write an article on how I do information architecture interviews, what I look for, and when I know my job is done (and I can move on to next steps like the audit and redesign). Let me know if that’s something you’re interested in, and whether you’d like me to focus on the perspective of the stakeholder, participant, or the person who asks the questions.

After initial interviews, it’s common to have a few people join discussions, for example, if I’ve asked for a walkthrough of common publishing workflows for content managers, or if I’m sharing my findings after reviewing an experience from the client’s platform.

Let’s look at the example of a client who offers hotel management software.

The platform serves different property types, including hotels, hostels, Airbnbs, and vacation rentals. During this quarter of our ongoing information architecture partnership, we’ve decided that my focus should be on improving the platform’s information architecture for large hospitality companies that operate multiple brands and locations.

Mid-stay move was one of the client’s newest and most requested features, allowing hoteliers at multi-property hotel groups to quickly transfer guests between rooms as needed.

In the past, users had to duplicate entries, and in some cases, start from scratch, resulting in increased potential registration mistakes, difficulty coordinating between staff, and sometimes affecting guest stays.

After interviewing users working at large hotel groups and reviewing product analytics, it’s obvious that something isn’t working.

The feature is supposed to work seamlessly as a drag & drop between rooms, but many users still don’t even know it exists.

After reviewing the experience, I have some hypotheses on why that might be.

I may have learned this from talking to users or stakeholders, or looking at analytics, but the scenario in a shared room often sounds like this, with A and B representing my client’s employees:

“But it’s right there on the website!”

I’ve heard this refrain of frustration from many teams. And it is frustrating, I empathize.

Companies handle this in different ways.

Some companies blame users. Behind closed doors, users are considered stupid, impatient, uneducated, and digitally illiterate. The company employees are expected to know better than the user what the user needs and be able to create that with limited, often nonexistent, input from the user.

Some companies play the blame game internally, with sales blaming marketing, marketing blaming engineering, engineering blaming product, product blaming design, design blaming QA, QA blaming customer support, and more often than not, shifting blame across layers of ICs and management in different departments. Solutions generally revolve around adding: introducing new technologies, processes, and features.

But some companies, maybe my favorite type, get curious. They try to understand why users may be struggling with mid-stay moves. They don’t view the feature as a Jira ticket marked complete, but try to dissect the experience as a whole, reviewing steps that need to happen before and after the actual drag and drop.

The company asks:

• What in the sequence could be confusing?
• What could be going wrong?
• Is this section where users would look for guest transfers between rooms?
• Does the language we’ve used match how users refer to this task?
• Is in-app search performing as expected for feature-related terms?

Leading with curiosity, they may discover:

• The language they’ve used doesn’t match how users refer to this task at work. They’ve called it Reassignment, but users refer to it as Room change or Mid-stay moves.
• In the platform control panel, Reassignment is a subtask under Revenue and reporting tools. It lives as a child item under Revenue because room changes are part of how one of the key features, AI-driven algorithmic pricing based on demand, weather, trends, availability, is calculated, so it made sense to place it under there. But in users’ day-to-day, this task was part of their Maintenance and housekeeping standard operating procedures, so they weren’t looking for it under Revenue.
• They’ve relied too heavily on the control panel, not offering other ways to discover this information that match users’ paths, such as in the bookings view, Guests pages, or guest experience templates most chain hoteliers rely on. (They’re thinking of content vertically, not horizontally.)
• Since the feature language doesn’t match natural language (how users refer to this task), the in-app search doesn’t return relevant results and it doesn’t take users to the Mid-stay moves page.

Many other factors may be impacting the experience, such as interaction design choices, page speed, and how the feature was marketed, but I focused on information architecture since that’s my area of expertise.

“But it’s right there on the website!”

If you hear or find yourself saying that, get curious.

Turning these findings into recommendations and implementing changes is the simple part. It’s a fairly straightforward process of research, auditing, and developing new content structures aligned with key goals, but one that immediately affects findability and user satisfaction.

For this example client, I’ve focused on improving the hotel management software’s IA for multi-property hotel groups. Small, but meaningful changes in the platform, such as improving the mid-stay moves IA, can result in hours saved for large hospitality companies that operate multiple brands and locations, and more importantly, help them provide an impeccable guest experience, while also avoiding enterprise churn and deepening the group’s relationship with my client.