How I work
I look for important problems we can improve with a small, well-chosen change. Then we find out whether it helped.
Big problems. Small solutions.
I weigh the potential impact against the effort involved. Sometimes a missing piece of information is getting in everyone’s way. Sometimes we need to understand the problem better before changing anything.
Start with five questions
Who is this for?
What problem are we trying to solve?
What are these users trying to accomplish?
What does the business need to achieve?
How will we know whether it helped?
Being specific about the audience became especially important as we explored personalization. The same experience might serve different people and goals.
Choose the right way to learn
An A/B test compares versions of an experience to see how a change affects a defined outcome. It is one option. A few usability sessions, watching people try a design, can reveal where they struggle before we invest in a live-site experiment. Interviews, customer questions, or a technical investigation may point us toward a different next step.
I often shared Nielsen Norman Group’s UX Research Cheat Sheet and built my own guide from it. I also helped designers choose methods, structure studies, and interpret their findings.
How I helped colleagues run their own research →
Useful evidence can come from everyday work, too. Ask Ulta questions exposed missing product information. A visit to retailers with design and development colleagues helped us explore an in-store kiosk before designing it.
The Ask Ulta story → · Sometimes the research is a field trip →
My four-part test brief
When an experiment is the right choice, I write down:
01 · The problem
What is getting in someone’s way? What evidence tells us it matters?
02 · The proposed solution
What change do we believe will help, and why? This is the solution hypothesis.
03 · The primary success metric
What measure will determine whether the test met its goal?
04 · The secondary success metric
What else do we need to understand about the effect of the change?
Afterward, I keep the result, what we learned, and the next action separate. An improvement in a secondary measure doesn’t turn an unmet primary goal into a win.
Make the work clear
Product, design, development, analytics, and business partners bring different knowledge. A useful process makes responsibilities and handoffs clear, from the first question through testing and implementation. I want people to know what they own, what they need from one another, and where to raise a concern.
When I share the work, I lead with the answer or recommendation, give the few reasons that support it, and keep the detail available. People who are short on time should still leave with something useful.
At Ulta Beauty, I put this into practice by mapping the Digital Optimization Team’s process from discovery through implementation, with owners, decision points, and handoffs made visible.
See the process map and the story behind it →
Keep learning after launch
After a test, we might implement a change, try another version, do more research, or leave the existing experience alone. A recommendation to implement is different from a change that has shipped.
Launch is another point in the cycle. Even a successful experience has room to evolve. Did it help? Where are people still struggling? What should we investigate next? Continuous improvement means making changes for a useful reason, while giving people clear choices and control along the way.