Best Buds: Data-Rich Accounts
Project review for a strain quiz aimed account acquisition.
STRATEGYACQUISITION
Results
In the first six months post launch
50,000 data rich accounts were created - a 20% increase in account creation
50% increase in strain favoriting
70% quiz completion rate
Each week, over a million Leafly web sessions start with a Google search for a specific strain name: “purple punch strain”, “wedding cake strain”. That search brings the user one of Leafly's 5,000 strain pages.
Background and problem to solve
Leafly's goal is to increase revenue through expansion of their marketplace. Retailers can post menus online and link menu items to strains. Retailers that have those menu items linked will show up on strain pages, and roughly 5-6% of users who visit a strain page will visit a dispensary page in the same session.
However, once a user goes to a dispensary page, conversion remains low. There were prior efforts to optimize conversion, but ultimately user feedback consistently pointed to feelings of being overwhelmed or not interested in ordering on line.
Furthermore, due to the nature of cannabis content on the site, Leafly faced major restrictions on advertising. About 70% of the web traffic is fed through an organic source, and 90% for the strain pages. This effort is also aimed at optimizing the cost of customer acquisition (and re-acquisition) through Google.
Once we can get a customer through 'owned' channels, data showed that they are more likely to convert and hit other key results.
“I like that you can filter the results especially because looking at this it says that there are 1,240 results within 30 miles and that’s kind of overwhelming”
“There’s a lot of data going on and a lot of things being shown it’s kind of overwhelming when you’re on the site”
“There’s a lot on Leafly it seems a little overwhelming”
In the words of our users...
Choice paralysis was common feedback from multiple UX studies
And from the famous jelly coupon experiment, we know that more is not always better.
OVERWHELMING, YOU SAY?
How might we allow customers to know if this strain is something they’ll like and want to buy within seconds of landing on a strain page?
Our hypothesis is that by immediately answering the question of “will I like this strain?” we can increase confidence in the rest of the content on our page, including the shopping moments which will yield more orders from strain pages and more direct traffic as users begin to rely on Leafly for trusted advice.
What question do we need to answer?
Analyzing landing page sessions, there are 30x more sessions that land on a strain page over a strain list.
Keyword analysis showed that keywords like "wedding cake strain" or "gelato strain" would bring users to a strain page, and keywords like "sleepy strains" or "sativa strains" would bring people to a strain list.
Based on this, we concluded that in order to meet the users where they are, we shouldn't answer "which strain might I like?" but instead "will I like this strain?"
What information do we need?
In order to compare a user's preferences to a strain and provide helpful, relevant, and accurate scores, what data do we need, and what order should we present it in?
Strain information we have
Each strain has a set of metadata we can leverage to compare and score how well it matches to a user's preference.
Using filter interactions
Based on filter interactions from our primary strain list, we can see that the effects and effects users want to avoid are most important to users. This is also underlined by user feedback.
"Condition" refers to user-reported conditions users can claim the strain helped with. However, legal concerns around using these claims made us choose to not include it in our first version.
Calming & energizing is not exposed in filters, but based on qualatative data, we knew it was similar to how people talk about strains. it as well as the consumption method type - which would help unlock future efforts of matching preferences to products.
The quiz questions
Developer feedback was that in order to ensure a reasonable score, we needed at least two effects captured - this is also the only question we required.
Rounds of user testing
Validating our hypothesis and unearthing new perspectives.
