Avast SafePrice

Redesigning a shopping extension so it helps at the right moment instead of interrupting all the time.

Company: Avast
Role: Product designer, UX redesign
Team: PM, e-commerce experts, devs, 1 designer, UX research
Result: Chrome rating 3.4 → 4.2

Context

SafePrice is a browser extension that finds better prices and discount coupons while you shop online. It had been live for about two years when I joined, and its key metrics, churn above all, were missing their targets. I owned the UX redesign. The business side was handled together with our e-commerce experts.

The problem

The service itself was simple: show a relevant offer or coupon. The flow wasn't. Coupons appeared repeatedly at the wrong moments, people didn't know when or how to use them, and the frustration ended in uninstalls. The UI was also dense and visually unbalanced, with unclear interaction signifiers.

Previous version of the extension — The version I started from.

What I changed

After a flow analysis, user research, usability tests and a competitor review, we ran a short workshop with the whole team to agree on the must-haves. Then I iterated on the flow with the PM, business experts and developers until it held up. The final flow came down to four rules:

Show less, more deliberately. Clear conditions for when each notification type may appear, weighted by how well each e-shop's coupons had performed and by user feedback.
Right moment, right message. For example, coupon codes are available exactly at checkout, not on the product page.
User controls the noise. Settings to choose which notifications show up and when.
Pull, not only push. People can open the extension and search for deals on purpose, a strong use case from research.

Final user flow diagram, blurred — Final user flow, based on the analyses and A/B tests (intentionally blurred).

With the flow approved, we finished the UI on top of the pattern library by Lukas Dubay. We made it lighter and more playful than before, with casual copy, illustrations and small animations.

Selected UI components — Selected final UI elements.

Validation

Testing a shopping tool needs realistic shopping data, so I kept prototypes simple and used them for several formative tests. We planned buffer time before launch to fix flow gaps, copy ambiguities and interaction timing. Before release we ran multivariate tests of different flow and UI variants with power and casual users, and picked the winner together with our data analyst.

Users' rating in the Chrome store went from 3.4 to 4.2 in four months. Engagement improved steadily in our metrics and eventually led to revenue growth.

What I took away

In e-commerce, frustrated users leave fast and rarely come back. The only real strategy is an experience that never annoys, even if that means leaving out features that would be useful but would crowd the moment. Balancing what helps the user against what meets the business goals was the most interesting part of the project.