With over 500 million unique visitors and 2 billion impressions per month, Linktree’s massive scale presents both challenges and opportunities for experimentation.

A strong culture of experimentation

Rob Erdin, Senior Director of Engineering - Platform, who oversees all the data teams at Linktree, emphasized the strong commitment from leadership to cultivate a data-driven culture.

“The expectation is that every change we make should be traceable to its impact. Our leadership has always shown a keen interest in experimentation, and there’s never been a need to explain its benefits. Instead, our discussions revolved around how we can experiment more, faster, and concurrently.”

The build vs. buy decision

Rob noted that despite the culture being in place, missing capabilities made it difficult to have objective conversations about product roadmaps.

“We needed to determine whether we should do more of the same or not and see the before vs. after results of changes we were making.”

Linktree realized that they needed modern experimentation tools to prioritize product decisions and unlock new avenues for growth.

“The build vs. buy decision came down to having the speed of acquiring these experimentation capabilities as soon as possible.”

Rob Erdin

Senior Director of Engineering - Platform

Migrating from legacy feature flagging

Linktree had been using a legacy vendor for feature flags but encountered obstacles when it came to expanding their experimentation capabilities.

Rob explained, “We were using a legacy vendor for feature flags, but lacked the experimentation capabilities we desired. They had limitations in their experimentation capabilities, and historically, were slow to introduce additional features.”

Linktree wanted a unified platform that could assist with their feature flagging, experimentation, and analytics requirements. “Feature flags and experiments are closely related, so it made sense to have them on the same platform.”

Two key benefits identified in combining feature flags and experimentation:

• Cost optimization.
• Reduction in operational overhead (no more juggling between multiple tools).

The migration process was straightforward and seamless. Rob’s primary focus was ensuring that they had the same segments available in Statsig as they did in their previous tool.

He added, "With Statsig, we gained the confidence to roll out changes, analyze the results, and continuously iterate with conviction.”

Making key product tradeoffs

Rob explained that Linktree faces unique challenges that complicate experiment setup.

“Our customers have separate profiles, each with its own set of visitors. This complexity requires us to create flexible solutions for handling scenarios that go beyond a single type of user.”

“What particularly impressed us was the way Statsig handles IDs, offering us the flexibility needed when setting up experiments to meet our requirements.”

Rob Erdin

Senior Director of Engineering - Platform

Linktree needs to make crucial trade-offs when developing features. These can have a significant impact on core business metrics.

For instance, their product-led growth (PLG) strategy involves using the ‘profile page’ as an acquisition channel for new users.

“We need to test certain changes in two ways. When we make changes to the user profiles, we need to understand the impact on visitors. When visitors view a profile, they often click through and create a Linktree account for themselves, making this a crucial surface for acquiring new users.”

Rob explained the nuances, “The core challenge here is that we need to ensure changes to profiles have a positive impact on our users (Linkers). We need to measure the impact of experiments on both users and visitors.”

“We may make a change to profiles that will allow us to convert more visitors into users. However, we also need to measure if this change results in a decrease in the retention of existing users.”

Another key use case is optimizing referrals and the SEO engine, where Linktree sets up experiments targeting users who arrive via Google or specific referral sources.

Achieving big wins with small changes

"Our overarching goal is to conduct more experiments because we have evidence that the number of experiments correlates with the success of our product."

Rob Erdin

Senior Director of Engineering - Platform

At Linktree’s scale with half a billion unique monthly visitors, even minor optimizations can directly influence revenue. Rob explains, “When we identify a significant pain point for a user, a simple change can yield substantial growth in core metrics.”

“Linktree receives 500 million unique visitors per month and generates 2 billion impressions each month. Each of these impressions can trigger multiple data points that Statsig can ingest, enabling us to monitor various behaviors and make informed decisions.”

Example: Replacing the Linktree logo with a CTA in the footer

Linktree’s user profile page is a key acquisition channel for new users. They conducted a test by adding a straightforward ‘Create Your Linktree’ button at the bottom of a user’s profile (instead of just the logo) – nudging visitors to easily create their own profiles.

Previously, there was no clear call to action, but the addition of the button resulted in an immediate increase in visitor engagement, leading to more new sign-ups.

These seemingly small adjustments can have a substantial downstream impact given the scale at which Linktree operates. They were able to increase daily sign-ups from profile logos by a factor of 10, while also improving activation rates. Overall, they doubled the daily sign-ups from profile referral traffic.

Building trust with warehouse-native experimentation

Rob noted how warehouse-native experimentation has made their lives easier:

“There is no longer a need to ingest data, and we don’t have to constantly monitor the ingestion pipelines for failures or anomalies in event volumes.”

Warehouse-native experimentation also increased trust across the entire organization.

"Having the peace of mind that the metrics we're using in Statsig are computed on the exact same data we use for our product analytics and reporting increases our confidence in the experiment results."

Rob Erdin

Senior Director of Engineering - Platform

Approval workflows

Rob also emphasized the importance of avoiding mistakes in config management, as they can lead to bugs or unintentionally turning off core features.

To address this concern, Linktree utilizes change reviews on Statsig, requiring approvals for feature flag changes and ensure appropriate enterprise-grade controls.

Working with the Statsig team and the road ahead

Rob shared his positive experience working with the Statsig team.

“Collaborating with the team has been fantastic. We’ve received an excellent level of support from Vineeth and team. They have a deep understanding of experimentation, and can quickly determine if something is feasible and provide clear feedback.”

“Our ambition for the next year is to increase our experimentation velocity and work on expanding the culture of experimentation — making sure teams understand the metrics they move."