Claire Ferrari, PhD
All work

02 · Verizon · 2026

Making selections inside a chat

A 193-person test showed vertical lists were 10 seconds faster than carousels, and set the standard for choosing a line in Verizon's AI chat.

Org and Year
Verizon, 2026
Role and Team
Lead managing two junior researchers
Methods
Unmoderated interaction testing, n=193
Partners
Conversational design
Two Verizon Assistant chat screens, labeled A and B, asking which line eligible for international travel to start with. A: each line is a row with a device image, the line owner's name, the phone number and an arrow, so one tap selects it. B: the same rows with radio buttons, the first line selected, and a Confirm button below.
Two of the tested line selection patterns: tap-to-select rows (A) and radio buttons with Confirm (B).

What selection pattern helps people pick the right line or device in Verizon Assistant, on a phone screen?

Context

Verizon Assistant, Verizon's AI-powered chat, had no standard template for choosing a line or device in chat. The goal was to find the layout, controls and level of detail that worked best and turn them into a standard.

Approach

I led the research design and delivery over multiple rounds, managing two researchers: one focused on accessibility and one on mixed methods. We ran unmoderated interaction tests with 193 participants, comparing vertical layouts with carousels, different selection controls, and levels of information density up to Level 4: name, number, device image, plan details and pricing.

What changed

Vertical layouts beat carousels on speed, and the most detailed rows gave people the most confidence. The conversational design team built the findings into its design system: vertical layouts on mobile across plan management flows, and a baseline rule requiring a device image for every troubleshooting line selection.

People preferred confirming a selection because it prevented mistakes, and the pattern with radio buttons and a confirm step scored 58% confidence and 41% ease of use.

193participants
10 secfaster median task time, vertical vs. carousel
70%confidence at the highest detail level