News

Tumblr boosts conversions by +4.7% with Just Words’ AI Messaging Engine.

Discover how Tumblr achieved a +4.7% lift in push notification performance using Just Words’ AI-powered Message Intelligence Platform. See real-world results across 5 high-impact campaigns with dynamic content optimization.

🎯Tumblr Taps Into the Future of Messaging with Just Words

Tumblr, one of the world’s largest creative communities, set out to reimagine how it communicates with millions of users through push notifications. With daily touchpoints like “Recommended Topics” and “Intro Trending,” even fractional lifts could shift the trajectory of engagement and long-term retention.

But Tumblr’s team wasn’t just looking to A/B test copy.
They wanted to evolve how messages are created, selected, and scaled—automatically.

🧠 The Old World: Static Tests. Gut-Driven Copy.

  • Slow Cycles: Testing was gated by setup time and engineering priorities, leading to limited experimentation.

  • Fixed Variants: At most, a few handpicked notifications could be tested in parallel. And once the test ended, so did the learning.

  • Low Insight Density: Sequential, single-channel tests made it hard to extract clear insights or transfer them across journeys.

⚡ The New World: Always-On Message Intelligence

With Just Words, Tumblr unlocked:

  • Instant Velocity: Copy variants were auto-generated, plugged in, and tested without waiting on new briefs or dev cycles.

  • Dynamic Optimization: Instead of static A/B tests, messages were continuously refreshed, ranked, and improved across campaigns.

  • Learning Loops: Tumblr didn’t just learn which message won, but why, and for whom, unlocking personalization at scale.

Approach

  1. Building a self-improving Message Pool

    Just Words created push notification content across 6 different notification types, according to Tumblr’s brand and tone.



  2. Auto‑Tune Optimization

    Just Words dynamically reallocated send volume toward top‑performing variants in real time—surface champions early, maximize impact. The system auto-archived variants in this pool based on past performance and recommended new variations from learnings over time.



  3. Transparent Reporting

    Every week, we reported the performance of winners and consistency in lift week over week.


  4. Extended Rollout

    After identifying winners in each test, we consolidated those champion variants into a single, larger experiment spanning all five notification types.

Results at a Glance

Notification

Variants Tested

Winners Found

Launch Rate Lift vs. Control

Recommended Topics

10

4

+23.11%

Intro Trending

19

4

+32.43%

Orbitz News

22

4

+31.85%

Recommended Blog

12

4

+33.89%

Crushes

20

5

+4.12%

Aggregated Open‑Rate Lift

By rolling out the five per‑notification winners in one combined test, Tumblr achieved a 4.7% lift in open rates across all notification types - a significant win given their millions‑strong user base.

“Integrating Just Words into Tumblr let us rapidly iterate on notification copy without tying up engineering resources for every change. It enabled our editorial team to own the process and test 10x more copy variations in just a few of months than we previously managed in a six-month period.”

— Andrii Nikitiuk, Engineering Lead, Tumblr

Key Takeaways

  • Scalable Gains: Double‑digit lift in launch rates across four of five campaigns, followed by a 4.7% aggregate open‑rate increase.

  • Speed & Agility: Champion variants emerged in under two weeks, enabling quick reinvestment of learnings.

  • Full Transparency: Direct dashboard links ensured continuous verification and stakeholder confidence.

Ready to unlock similar engagement lifts for your company? Chat with us

© 2024. Choice AI Inc. All Rights Reserved