YEAR
2023-24
ROLE
Product strategy, data analysis, UX & motion design
PLATFORM
Web and mobile
AI Writing Industry Overview
QuillBot’s Paraphraser has long been a go-to companion for writers, known for simplifying complex ideas and enhancing clarity with just a few clicks. But as more and more large players (primarily OpenAI's ChatGPT) having more resources at their hands and having a global presence in multiple markets came up with their solutions (a.k.a., the LLM era) in the field of AI, it was obviously a challenge for existing players to sustain themselves in the domain. This new market situation led to a downfall in overall paraphrasing traffic, which ultimately led to a reduced top line.
Advent of ChatGPT & LLMS
Paraphrasing tool trends
😰The Angst!!!
The Problem
Amidst the noise, when the dust settles, the writer, at the end of the day, is not able to extract optimum value from the products they use. Writers tend to browse multiple AI products; hence, extracting value (from a product) swiftly becomes highly important for them in their workflow.
Entice user interest with serendipity
Writers—especially academic students and professionals—often navigate multiple AI tools under tight timelines. In fast-paced workflows, discovering value swiftly becomes essential. Amid the noise of features and interfaces, many users never get to experience the full potential of the tools they use.
To address this, the product now highlights the value of premium paraphrasing modes in a more contextual, unobtrusive way. Instead of burying users in forms or locking everything behind paywalls, targeted “sneak peeks” are surfaced based on writing intent and target audience. These timely previews help users make more informed decisions while staying focused on the task at hand.
In an environment where attention is scarce and deadlines matter, this approach ensures users encounter value when it matters most—during their actual writing process.
What data spoke?
The premium modes have only 20% adoption among free users.
By default, users land on “Standard” mode selected. We observed a mere ~27% adoption across premium modes amongst experienced users (7+ days). These users exhibited a default bias, thus losing out on betting writing outcomes possible by accessing premium modes.
Hypothesis
Solve #A
To address the needs of users who are ultra new (0-7 days), we sought to increase the adoption of premium modes by proactively delivering value. The challenge was to tap into the users’ organic workflow without seeming intrusive.
Solve #B
To bolster experienced users (7+ days), we challenged the default bias with well-timed variable rewards that shift focus back to the premise of premium modes.
Solve #A • Catered towards ultra new users
No sooner does the user paraphrase than they are prompted proactively with a preview of how it may sound with an actual alternate mode sample. This approach also gives a clear-cut head-to-head comparison amongst the different premium modes and the incumbent output, giving users optimum flexibility to reach the best outcome efficiently. These sneak-peeks are triggered by our model, where the user's writing context is at the helm.
This resulted in a whopping 🎉 ~6.5% increase in premium conversions and a 19% increase in free users premium mode exploration.
Solve #B
We also explored a few gamification concepts that were never implemented. The explorations included gift boxes and button animations, as depicted below.
Tapping into the organic user's flow
🗣️ “As a writer with 💎 QuillBot premium, I add text, then I think about my writing mode, and I go ahead with.”
We saw the above user story playing out most times as per our UXR. An average 💎 premium user takes 3-5 seconds to click on the “Paraphrase” CTA after adding text. Therefore, we injected a recommendation unit on the writer’s text-input side to cash in on that sentiment.
Awareness during Season SALE (augmented via motion graphics)
We conducted an A/B test where the control had a static creative on the homepage and the test variant had an animated sequence (depicted below).
🎉 The test variant yielded ~4.5% better performance in conversion during the sale.
PARTING THOUGHTS 👋👋👋
We drove a net 40%+ conversion through the initiatives aforementioned and beyond.
God speed. ⚡