Marina Ruiz Available · Los Angeles

Item No. 1 · Weedmaps · Conversion and growth

Onboarding 100K new users a month, 20% better

The largest cannabis marketplace had its biggest conversion problem at the front door: high intent, no vocabulary, and a bounce before the first purchase.

Role

Lead product designer, Conversion and Growth

Surfaces

Web, iOS, Android

Scope

Onboarding funnel and brand page consolidation

Method

Shipped behind A/B tests with PM, engineering, and data science

Weedmaps is the largest cannabis marketplace, and 100K+ new users arrived every month with high intent and no vocabulary. I designed a guided onboarding funnel that replaced open browsing with a short preference flow feeding personalized recommendations, then shipped it with PM and engineering through the experimentation program across web and mobile.

Five Weedmaps onboarding screens in sequence: interest selection, dispensary-type preference, a curated list of recommended dispensaries, a brand carousel, and location permission.
Scroll to see the full five-screen flow →The full flow, one path: first question to recommended dispensary.

Line 01

This is an ecommerce problem, not a cannabis problem

Strip the category away and the shape is familiar to anyone running a DTC funnel: a shopper with zero product vocabulary. The same person who lands on a fragrance, skincare, or fine jewelry site knowing what they want to feel but not what to ask for.

Advertising restrictions meant the platform could not buy its way out with retargeting, so every bounce was expensive and mostly unrecoverable. That is why this was staffed as a conversion initiative, not a delight initiative.

Line 02

Ask about the person, not the product

Users who cannot answer "what product do you want?" can answer "how do you want to feel, and how experienced are you?" The flow required zero category knowledge, stayed short, and made each answer visibly narrow the outcome, so it read as progress rather than a form.

Returned

“What product do you want?”

A question a first-time visitor structurally cannot answer.

Exchanged for

“How do you want to feel?”

Zero category knowledge required.

ProductPerson

What it cost

Every added question is friction before value. We were asking newcomers to invest before showing them anything.

Two onboarding screens: a product-type preference question, then a results screen recommending local brands that carry the selected products.
Scroll to see both screens →One answer, one step later: personalized brand recommendations.

Line 03

Funnel to a decision, not a catalog

The recommendations ended at a small, curated set of nearby dispensaries and products instead of dumping users into filtered browse. For a zero-vocabulary user, relevance beats completeness, and the bounce data on open browsing backed it.

What it cost

Showing less risked feeling limited, and it reduced the surface area available for merchandising.

Desktop WYLD brand page on Weedmaps: product grid, profile header, and quick-action tabs consolidated onto one indexed page.
Scroll to see the full desktop page →The desktop brand page, same hierarchy, sized for longer research sessions.

Line 04

Make every answer compound

Onboarding responses fed the affinity models the data science team maintained, so the flow was not just a first-session fix, it was structured preference data improving recommendations platform-wide. This is the decision I would defend as the most senior one: it turned a UX patch into a data asset.