High churn, low LTV
High churn, low LTV
"What's that?" - Murf AI
Gen AI companies are often looked at with skepticism for their inability to scale past consumer/prosumer use cases, which are ailed by high churn and low LTV. Murf AI isn’t one of those companies.
Murf is the runaway leader of the text-to-speech space not just for prosumers, but also for businesses. What separates the wheat from the chaff in this space (and Murf from the rest) is a GTM orchestration that (a) proactively triggers self-serve conversions and (b) serves flaming hot enterprise opportunities to their sales team. On a platter.
This is the story of how the generative AI company Murf uses behavioral AI to penetrate the enterprise market while putting their self-serve conversions on anabolic steroids.
Come say hi to us at PLGTM: Okay, so this is fresh off the press. Sort of.
We're hosting a first-of-its-kind conference for Go-To-Market teams at the Mezzanine in New York City on June 27 🎉
Read about the speakers, sessions, and the agenda for the day here. If you’re in town and would like to attend, use the code Toplyne50 to get 50% off on your tickets when you sign up. ⚡️
PS: We’re giving away a handful of tickets for free to our subscribers - drop a “PLG FTW” in comments to enter the raffle 🎟️
Meet Murf AI 🤝
Entertainment animation agencies creating entire TV series episodes using AI-generated voices… authors creating a series of fantasy fiction audiobooks with little to no production costs… YouTube influencers creating rap videos using AI-powered speech - these stories are all very real in 2023 and they’re all powered by Murf.ai.
Synthetic speech company Murf lets users generate “human-like” voiceovers without needing to buy recording equipment or hire a voice artist.
Founded in October 2020, Murf saw its ARR rise 26x over the next two years while synthesizing more than one million voiceover projects along the way, in various speaking styles and tones. With a library of more than 120 human-parity AI voices across 20 languages, Murf raised a $10 million Series A round in September 2022, on the back of their blazing growth.
What they didn’t know at the time, was that their growth so far was about to pale in comparison to the volcanic surge that was coming their way.
“Post the announcement of DALL-E and ChatGPT, we started seeing 3x more website visits than we were, three months ago.” - Chaitanya, Head of Growth at Murf
The trouble with growing 3x in 3 months...
Murf tracks the behavioral data of users on Mixpanel. On the GTM end, they use Customer.io and Sendinblue to trigger emails to their customers prompting conversions - both self-serve and sales-driven.
Prior to using behavioral AI, Murf built cohorts on Mixpanel to qualify users to trigger conversion prompts to. They tested a few hypotheses:
Hypothesis 1: Anyone who’s created 10 minutes of audio
Being a content creation tool fundamentally, Murf’s north star metric is “minutes to content created.” Qualifying anyone who created 10 minutes of speech on Murf broke at volume. Way too many users crossed the threshold, and once past the 10-minute mark, they all looked the same.
Hypothesis 2: Anyone who’s used 5 or more different voices
While this logic was good at disqualifying users who weren’t ready to convert yet, it failed to surface the conversion crème de la crème.
Hypothesis 3: Anyone who shares while on the free trial
In Chaitanya’s words: “If a user shares while on the free trial… that's a very high interaction, but a very small percentage was doing that. The percentage was similar to our conversion rate. Then again, it's tricky to qualify.”
Murf experimented internally with cohorts defined on Mixpanel against these and various other actions they hypothesized to be correlated to conversion. Although individually each of these experiments positively impacted conversion rates, they were discovering a bottleneck with each, beyond which there was no meaningful difference.
“Given that it’s a software tool, there are a bunch of things that users could do. As we got into more metrics, it got trickier and trickier. How many metrics, which metrics - where do we stop? So yeah, then it was very clear that this is a data science problem.” - Chaitanya, Murf
And so in September 2022, an AI company Murf turned to AI for help in a very unsurprising turn of events.
Murf integrated behavioral AI into their GTM stack.
Skyrocketing conversion rates by 7x
Murf's problem was two-fold. On one hand, they needed to focus their relatively small sales team on the leads most likely to convert, rather than waste time, energy, and emails chasing leads who weren't qualified to convert yet.
On the other hand, Murf didn't want to spray and pray by email blasting the many thousands of new users who sign up every day. For many reasons:
1. Murf's sales bandwidth is a bottleneck. While their sales teams looked to book meetings, they didn't want to book too many meetings either and swamp them with leads who in reality had shown no intent. Not too many, not too few.
2. E-mail marketing fatigue is real. E-mails triggered to users who, through their behavior, show no intent to convert can lead to decaying open rates over time, and tune them out of the channel completely.
"Our (growth's) job is to drive Calendly invites. That's our success metric. You can run an email campaign, or in the future, run an in-app campaign, the idea is to drive Calendly invites. We can't drive too many and we can't drive too less," adds Chaitanya.
Murf built two orchestrations through the behavioral AI engine.
Playbook #1
Goal defined to the AI: Free-to-enterprise sales-led conversions
Integrations into the AI engine: Mixpanel (product analytics), Customer.io, and Sendinblue (e-mail engagement)
GTM touch: E-mail sequence asking users if they'd be interested in talking to sales to learn more about Murf.
Impact: Within the first two months of running the playbook, the AI engine increased Murf's enterprise conversions by 10%, leading to an immediate 20% lift in revenue closed for the sales team.
Playbook #2
Goal defined to the AI: Free-to-pro self-serve conversions
Integrations into the AI engine: Mixpanel (product analytics), customer.io, and Sendinblue (e-mail engagement)
GTM touch: E-mail sequence prompting a discount code to users during Black Friday if they upgrade to the pro plan.
Impact: In the duration that this playbook ran, Murf ran an A/B test to validate the impact of the AI engine, against a cohort defined on Mixpanel. The conversion rates among the AI cohort outperformed the Mixpanel cohort by 7x.
AI for AI: Up next for Murf
With these two playbooks orchestrating their GTM, Murf is able to be proactive about penetrating enterprises at a higher ticket price. Their top of the funnel growth rates continue to hold.
And so, while acquisition is not a burning problem for Murf (it's an understood problem statement), a problem with a lot of unknowns is churn reduction and account expansion.
Playbooks 3 & 4 loading ⏳, but more on that later.
PLG FTW
How is Murf's GTM different from competitors like Deepgram and Eleven Labs?