Problem
"Chat With Us!" is UMD Libraries' live chat service — embedded across every digital property the library provides, staffed by specialized subject librarians who take shifts answering research questions one-on-one. With 5–10 librarians and staff online at any given time, the service was mostly idle — typically only one or two patrons using it at once. Usage had stayed flat for several years, which both wasted the staffing dedicated to keeping it online and put the future of the service itself at risk.
increase the usage of the live chat service?
Framing the Problem
As the UX designer, I couldn't directly influence how librarians answered questions, and I couldn't control whether users wanted to start a conversation in the first place. What I could work on was the medium — the chat service itself — by increasing its visibility and lowering the barrier to using it.
Design Strategy: a decision-making equation
I framed the question of "will a user start a chat?" as a simple value equation:
Likelihood of use = Gain − Effort
If the result is positive, users are more likely to engage.
Of these four lenses, Effective and Efficient are properties of the librarians and the conversation itself — outside the medium. Discoverable and Usable are properties of the interface, so that's where the redesign focused:
- Find the service → increase discoverability
- Use the service → make it more usable
Design
Each change below pushes one side of the equation — easier to find, or easier to use. (The audit behind them leaned on classic usability heuristics; the equation set the priorities.)
One name, everywhere (Discoverable)
The service appeared under several different names across the website. Consolidating on a single name — "Chat With Us!" — gave users one recognizable label to look for, regardless of where they landed.
Show when it's live (Discoverable)
The chat box now clearly indicates whether the service is live or offline, so users know upfront whether starting a chat is worth their effort.
Meet users where they expect it (Discoverable + Usable)
Recognition beats recall on both sides of the equation, so this drives two changes.
Discoverable — aligned the widget with conventions users already know from chat services like Facebook Messenger: anchored at the bottom-right corner of the screen, consistent across pages.
Usable — standardized the widget so that every element a user needs is visible the moment it opens, rather than hidden behind extra clicks or exploration.
Prevent mistakes before they happen (Usable)
Added consistent, contextual guidance to head off mistakes:
- A persistent notice at the top of the chat warning users not to close the tab and lose their conversation.
- Clear instructions on the file-upload page so users know what formats and sizes to expect, replacing inconsistent upload limits with a single set of consistent guidelines.
Make offline a starting point, not a dead end (Usable)
When the service is offline, the widget now surfaces quick links to the most common alternative resources, alongside the schedule of upcoming live hours. The layout was redesigned to be scannable, so users can see at a glance what help is still available.
Impact
After the redesign launched, chat usage grew ~30% year-over-year.
Because the widget is embedded across every UMD Libraries digital property, the redesigned experience reaches users at every entry point into the library system — so the gain compounds across surfaces rather than being isolated to a single page.
Reflection
In product design there's often a lot you can't control — the quality of the service being delivered, or whether users feel motivated to engage with it. The work is to identify what parts of the system you can influence and pick the highest-leverage points. The value equation made the medium itself the obvious lever, and Nielsen's heuristics translated that lever into specific moves.
The same framing also points to where the next round of work lives: Effective and Efficient — the conversation itself. Surfacing expected response time during the wait, or pairing the widget with lightweight post-chat feedback, would let the same equation drive improvements on the service side, not just the medium.
