OVERVIEW
[TODO]: add an overview
TEAM
Growth Buy
TIMELINE
10 weeks
ROLE
Product Design Intern
SKILLS
Experiment design, Buy journeys,
[todo]: add more
THE PROBLEM
THE PROBLEM
Conversion with free users drops ~5% around month four, and it mostly never recovers.
Conversion with free users drops ~5% around month four, and it mostly never recovers.
There is a large segment of free customers that use Confluence, where about half of these users have never started a trial of a paid plan. Of the ones who do, the large majority never touch a paid feature before the trial ends.
The bigger problem was timing. After 3 months, customer conversions tend to drop around 5%. Somewhere around the 4 month mark, most free customers who were going to become paying customers already have. The rest grow accustomed to working around the free plan’s limits.
My team worked inside Growth, on the Buy side, where the group is responsible for the trial and purchase journey once someone's interested in Confluence. I spent 10 weeks there owning design for one experiment aimed at that four-month cliff, working with a PM, a few engineers, and a content designer.
There is a large segment of free customers that use Confluence, where about half of these users have never started a trial of a paid plan. Of the ones who do, the large majority never touch a paid feature before the trial ends.
The bigger problem was timing. After 3 months, customer conversions tend to drop around 5%. Somewhere around the 4 month mark, most free customers who were going to become paying customers already have. The rest grow accustomed to working around the free plan’s limits.
My team worked inside Growth, on the Buy side, where the group is responsible for the trial and purchase journey once someone's interested in Confluence. I spent 10 weeks there owning design for one experiment aimed at that four-month cliff, working with a PM, a few engineers, and a content designer.
THE REFRAME
THE REFRAME
The project was called Auto-upgrade, and my PM eventually agreed that was the wrong word for what we wanted.
The project was called Auto-upgrade, and my PM eventually agreed that was the wrong word for what we wanted.
The title of the experiment included the word “Auto-upgrade” and initially, the experiment led with that intention. During the first scoping conversation with my PM, I asked a question that set the foundation for the experiment: how do we get customers to upgrade without making it feel forceful instead of something they chose? I reviewed past experiments and noticed a pattern where customers who'd been auto-upgraded in the past converted to paying customers at a worse rate than customers who upgraded manually. There was misalignment with the data justifying the project and the thing its name described.
My PM agreed that we should avoid auto-upgrading. What we were actually building was closer to a strong, opt-in recommendation. Using a clear CTA like ‘Get started’ would be the clear next step, and we needed to include an easy way to opt out of the upgrade.
The title of the experiment included the word “Auto-upgrade” and initially, the experiment led with that intention. During the first scoping conversation with my PM, I asked a question that set the foundation for the experiment: how do we get customers to upgrade without making it feel forceful instead of something they chose? I reviewed past experiments and noticed a pattern where customers who'd been auto-upgraded in the past converted to paying customers at a worse rate than customers who upgraded manually. There was misalignment with the data justifying the project and the thing its name described.
My PM agreed that we should avoid auto-upgrading. What we were actually building was closer to a strong, opt-in recommendation. Using a clear CTA like ‘Get started’ would be the clear next step, and we needed to include an easy way to opt out of the upgrade.
THE FIRST DESIGN
THE FIRST DESIGN
The first modal told customers they'd already gotten Premium, before they had.
The first modal told customers they'd already gotten Premium, before they had.
Upgrade prompts were presented to users primarily within modals. The first modal I designed leaned on language to deliver the upgrade prompt. I framed it to say their workspace had been given free access to Premium. Technically, their workspace wouldn’t be upgraded on the backend until they clicked the primary CTA. This avoided the auto-upgrade concern and framed the trial as a gift for the customer.
Crit feedback came back with concerns about potentially disrupting the user’s workflow through the surfacing of the modal. This was the same underlying concern I had with auto-upgrading users. The content of the modal was good, but if it was going to interrupt someone's work, it needed to connect to something they were actually doing, rather than show up disruptively.
Upgrade prompts were presented to users primarily within modals. The first modal I designed leaned on language to deliver the upgrade prompt. I framed it to say their workspace had been given free access to Premium. Technically, their workspace wouldn’t be upgraded on the backend until they clicked the primary CTA. This avoided the auto-upgrade concern and framed the trial as a gift for the customer.
Crit feedback came back with concerns about potentially disrupting the user’s workflow through the surfacing of the modal. This was the same underlying concern I had with auto-upgrading users. The content of the modal was good, but if it was going to interrupt someone's work, it needed to connect to something they were actually doing, rather than show up disruptively.
THE CONTEXTUAL FIX
THE CONTEXTUAL FIX
Four specific moments could trigger the modal, and only the first one a customer hit would show it.
Four specific moments could trigger the modal, and only the first one a customer hit would show it.
The fix I explored was contextual: tie the modal to something the customer was actually doing. I mapped four moments where a free customer might specifically feel the ceiling of the free plan: creating a third whiteboard, approaching a storage limit, adding close to the free user cap, or a natural pause after a stretch of active work in a Confluence doc. Whichever of the four a customer encountered first would surface a modal offering a trial because of that limit. After that, they wouldn't see it again to avoid repetitive disruptiveness.
The fix I explored was contextual: tie the modal to something the customer was actually doing. I mapped four moments where a free customer might specifically feel the ceiling of the free plan: creating a third whiteboard, approaching a storage limit, adding close to the free user cap, or a natural pause after a stretch of active work in a Confluence doc. Whichever of the four a customer encountered first would surface a modal offering a trial because of that limit. After that, they wouldn't see it again to avoid repetitive disruptiveness.
THE HARDER CALL
THE HARDER CALL
This audience had spent four months proving they'd never trigger any of them.
This audience had spent four months proving they'd never trigger any of them.
Contextual triggers assume the customer will eventually do the thing that trips the trigger. This audience's patterns argued against that. During 4 months of regular use, these users typically didn’t interact with any existing upgrade points. In this case, we had enough evidence to show that waiting for a behavior that, for this group, had already proven itself absent, would likely be ineffective.
Contextual triggers assume the customer will eventually do the thing that trips the trigger. This audience's patterns argued against that. During 4 months of regular use, these users typically didn’t interact with any existing upgrade points. In this case, we had enough evidence to show that waiting for a behavior that, for this group, had already proven itself absent, would likely be ineffective.
I shipped the broader modal anyway, and left the contextual version for future experiments.
I shipped the broader modal anyway, and left the contextual version for future experiments.
I handed off the first version instead: a modal shown to this audience regardless of what they were doing in the moment. I kept in mind crit feedback and avoided claiming that the upgrade had already happened. It was a more assertive approach than I initially expected to take, but for this audience, it was also the one more likely to actually reach them based on past data. Additionally, we would be able to take the learnings from a more assertive approach like this, and develop stronger future experiments.
Instead of throwing out the contextual version I designed, I documented it in the handoff as a recommendation for a different, more engaged audience, or a future run of this same experiment once this segment's been worked through.
To keep some of the goals of tying value to the trial, I explored modals that promoted AI features. This supported company goals and would showcase Premium value to users.
I handed off the first version instead: a modal shown to this audience regardless of what they were doing in the moment. I kept in mind crit feedback and avoided claiming that the upgrade had already happened. It was a more assertive approach than I initially expected to take, but for this audience, it was also the one more likely to actually reach them based on past data. Additionally, we would be able to take the learnings from a more assertive approach like this, and develop stronger future experiments.
Instead of throwing out the contextual version I designed, I documented it in the handoff as a recommendation for a different, more engaged audience, or a future run of this same experiment once this segment's been worked through.
To keep some of the goals of tying value to the trial, I explored modals that promoted AI features. This supported company goals and would showcase Premium value to users.
WHAT SHIPPED?
WHAT SHIPPED?
The modal would go out to every free customer that fit our criteria, with a full trial lifecycle mapped from day 0 to day 30.
The modal would go out to every free customer that fit our criteria, with a full trial lifecycle mapped from day 0 to day 30.
Because of limited time and it’s place on the roadmap, I handed off my work to another designer.
The experiment would go out to eligible free customers on both Confluence. Those eligible needed to be active for four or more months, have a specific number of enabled users, and have never trialed.
Beyond the modal itself, I mapped and designed the full trial lifecycle which included, a welcome moment on day 0, a nudge around day 15 to use Premium features, a review prompt on day 25, and countdown reminders through the final three days before the trial ended. Customers without payment details on file got a modal at the end recommending a downgrade to Standard instead of immediately being placed back on the Free plan.
Because of limited time and it’s place on the roadmap, I handed off my work to another designer.
The experiment would go out to eligible free customers on both Confluence. Those eligible needed to be active for four or more months, have a specific number of enabled users, and have never trialed.
Beyond the modal itself, I mapped and designed the full trial lifecycle which included, a welcome moment on day 0, a nudge around day 15 to use Premium features, a review prompt on day 25, and countdown reminders through the final three days before the trial ended. Customers without payment details on file got a modal at the end recommending a downgrade to Standard instead of immediately being placed back on the Free plan.
WHAT I TOOK FROM IT
WHAT I TOOK FROM IT
Growth design taught me to make a call I could defend, and to leave the better idea findable for whoever came next.
Growth design taught me to make a call I could defend, and to leave the better idea findable for whoever came next.
Growth design follows the same lifecycle as any other kind of design work, but it's faster, messier, and carries a lot more ambiguity. Working on a growth team pushed me to think about how a single design decision could shape our learnings. Leaning on data from past experiments built up a repository I could point to. It helped me understand our customers, and it helped me inform and defend my own decisions. Balancing what different stakeholders needed, what would actually serve the user, and the ambiguous nature of growth design made me a stronger designer. I learned to make decisions I could justify with data, scope solutions to fit the constraints of a specific project, and document the ideal experience so future work had a foundation to start from.
Growth design follows the same lifecycle as any other kind of design work, but it's faster, messier, and carries a lot more ambiguity. Working on a growth team pushed me to think about how a single design decision could shape our learnings. Leaning on data from past experiments built up a repository I could point to. It helped me understand our customers, and it helped me inform and defend my own decisions. Balancing what different stakeholders needed, what would actually serve the user, and the ambiguous nature of growth design made me a stronger designer. I learned to make decisions I could justify with data, scope solutions to fit the constraints of a specific project, and document the ideal experience so future work had a foundation to start from.
A huge shoutout to my mentor Charlotte, my manager Stephen, and the enter Growth and Education team designers that supported me throughout my internship
OVERVIEW
The ACLU is one of the most recognizable non-profit organizations in the country, centering their work around the preservation of Constitutional rights and United States laws that are guaranteed for everyone in it. Along with it’s National site, there are also more than 50+ separate affiliate websites, one for every state.
TEAM
Design Systems
TIMELINE
3 months
ROLE
Product Design Intern
SKILLS
Component design, Documentation,
Accessibility, Engineering handoff
OVERVIEW
The ACLU is one of the most recognizable non-profit organizations in the country, centering their work around the preservation of Constitutional rights and United States laws that are guaranteed for everyone in it. Along with it’s National site, there are also more than 50+ separate affiliate websites, one for every state.
TEAM
Design Systems
TIMELINE
3 months
ROLE
Product Design Intern
SKILLS
Component design, Documentation,
Accessibility, Engineering handoff