Inside Rewind’s path to 170 Series A offers
Co-founder and CEO Dan Siroker on building a viral AI product in public
👋 Hi, it’s Kyle and welcome back to Growth Unhinged. Every week, I take a closer look at the playbooks behind the fastest growing startups. Today I’m announcing a new series about the winding path from zero to one. First up: AI darling Rewind.
Current subscribers: 39,412. Join the crew ☝️
From the outside Rewind has all the trappings of an overnight success. While that’s not entirely untrue – the product did attract 300,000 waitlist signups – it’s not the full story, either.
I met co-founder and CEO Dan Siroker in April 2021 as he was contemplating how to monetize the precursor to Rewind, a meeting recording and transcription product called Scribe (which was backed by OpenAI CEO Sam Altman and First Round Capital). The seed for the idea came out of Dan’s experience losing his hearing in his 20s. He gained a superpower with his first hearing aid and subsequently yearned for ways to use technology to give others these types of superpowers. Dan eventually narrowed his focus to human memory, which peaks around age 20 and then declines afterwards. “If there’s a hearing aid for hearing, what’s the alternative for memory?” he wondered.
Out of the gate “Scribe was doing OK,” Dan told me. “Most founders wouldn’t have pivoted from Scribe to Rewind.” But he thought he could build something that would really take off – and Apple’s new Silicon M1 chip was his way in.
We discussed how Dan and the Rewind team built a viral product, how he approached early-stage monetization and how Rewind leverages product-led growth (PLG) to maintain momentum. Keep reading for bonus insights into why Dan fundraised in public – and turned down a $1B+ valuation.
From pivot to viral AI hit
After Apple Silicon came out, Dan and team decided to reimagine their product as a Mac app called Rewind. They started working on it in Fall 2021, spending six months on product development then another six months on user onboarding and product finetuning. All told, it took about a year of building in stealth until Rewind’s public launch on November 1, 2022 (yes, they won Product Hunt).
The team was mostly generalists at that point, consisting of eight people who were almost all software engineers including “two designers who could code.” Each of them had to learn a new tech stack (Apple) given this shift in product direction. (Dan recommends hiring people who are adaptable at this stage and avoiding overspecialization too soon.)
In those early days Dan and team manually onboarded hundreds of people. They created a separate Slack channel for each and paid close attention to every part of the user’s experience. Thankfully for the team, attracting these early users was relatively easy. Many were initial Scribe customers who had actually come to Scribe looking for a product like Rewind. Others were early-stage founders or people in Rewind’s orbit who they thought would give good feedback.
Through this high-touch approach, the Rewind team “got rid of every single small issue that we learned, which is what gave us confidence it would be a hit.” The team were specifically looking at both qualitative and quantitative signals of product-market fit:
Qualitative → the things they’d hear during manual onboarding
Were new users getting not only to an ‘aha’ moment, but to a magical experience?
In Rewind’s case this was typically when a user could go back to something they had thought they had lost forever (we’ve all been there…)
Quantitative → the data around product usage
How many days a week were people actively using Rewind?
How much would people record passively?
Dan had two tips for founders building new products:
Look at cohort-based retention of engagement and usage. You want to see the curve flatten out or plateau. The product needs to be sticky enough for at least some users, although it won’t need to be sticky for all users.
You might not know ahead of time which users it’ll be sticky for. For Rewind, the product has been stickiest among people with ADHD. Dan noted that this wasn’t preordained; the Rewind team stumbled upon that group accidentally.
“For about three weeks we were the hottest AI company out there until ChatGPT launched,” Dan recalled.
Twitter was Rewind’s biggest launch channel, aided by Dan’s pre-existing Twitter following. Product Hunt was effective as well – Rewind was #1 product of the week.
Dan attributes their viral success partly to the controversial aspect of what they were doing. “The product is a little weird and even a little creepy at first blush,” he said. He recommends leaning into the parts of your product that feel different because “that’s what people will want to talk about.”
His advice for generating a buzzy product launch:
Tell a story around your product. Dan’s personal story of losing his hearing in his 20s resonated with people as did narratives about using AI to augment memory and Rewind’s ‘why now’ of the new Apple Silicon chips.
Be transparent in your thinking about why you are doing what you’re doing.
Create a kickass demo. “You have to be able to show and not just tell. Some part of your demo needs to feel magical. For us, it was the basic idea of rewinding time and that visceral feeling that comes with it.”
Deciding to monetize from day one
Many PLG startups launch with a free product, then figure out how to monetize it. But Dan knew from the beginning that he wanted Rewind to be a paid subscription. “From a privacy perspective, we knew that charging for it would be a feature and not a bug. It would signal that we weren’t going to sell your data,” Dan said.
Rewind quickly attracted more than 300,000 waitlist signups (their initial goal was a mere 1,000). The team wanted to galvanize this wave of interest before people forgot and moved onto the next thing. They decided that the only way to use Rewind would be to pay for it. This would screen out any ‘tire kickers’ who weren’t serious and would allow the team to focus on building a better product for the people who were willing to pay for it.
The upside: a lot of revenue growth very quickly.
The downside: if people didn’t get to a magic moment in Rewind, they would churn. Dan would rather give these people a chance to try the product and get to the magic moment before being asked to pay.
The early pricing came from rigorous A/B price testing – remember, this was all before the company had transparent pricing on the website. For everyone who was invited off the waitlist, Rewind randomly assigned them one of five price points ranging from $10 to $50 per month. From there, the team monitored conversion rates, engagement and retention.
Dan noted that conversion “was pretty good” across each of the five price points, although pricing did make a big difference in terms of product engagement. Surprisingly, the people who spent $20 per month had far higher engagement than those who spent only $10 per month. This led the team to set initial public pricing at $20.
Rewind continues to iterate on its pricing and packaging. They’ve moved to a freemium model, which Dan said has been effective. And they’ve tested a usage-based paywall as well as two types of paid plans (“Basic” at $10 per month and “Pro” at $30 per month), both of which they’ve rolled back.
The v1 usage paywall, for example, didn’t work because it was dis-incentivizing usage. Rewind had capped the number of “rewinds” (the free plan got 50 total, the Basic plan got 10 per month), which was the value metric that correlated most strongly with retention. This caused people to monitor and self-police their rewind usage, which hurt their product experience and future monetization potential. Lifting the usage paywall and removing the notion of tracking rewinds during the evaluation stage dramatically increased rewind usage by 3-4x, Dan mentioned.
Rewind now has a reverse trial motion, which I’ve written about previously. New users get to experience the best of Rewind Pro for their first 30 days. If they don’t convert during the trial period, users start to experience gates around their search results (only the top three results are shown to free users), AI generated emails (free users only see a teaser) and AI summarization.
Dan recommends that other founders run price testing through the van Westendorp method. This is a series of survey questions where respondents tell you a price that they consider to be:
Too cheap where they’d doubt the quality of the product
A bargain where it’s a great buy for the money
Pricey where it’s not out of the question, but they’d have to give thought to buying it
Too expensive where they would not consider buying it
These van Westendorp tests can be run with existing users or paid panelists who don’t use the product. Dan told me that his team used SurveyMonkey’s paid panelist tool, which gets responses back quickly, comes with out-of-the-box qualifying features and only costs about $1 per response. “If you do the analysis right, you get this great graph around the user psychology of your product.” (Here is Rewind’s real-life van Westendorp chart.)
How Rewind adopts PLG
Rewind continues to grow by leaning into the powerful word of mouth conversations around the product. Specifically, people will take a screenshot of Rewind and share it on Twitter (X). “The weirdness of our product is a viral thing that keeps growing. It’s not a one moment in time kind of thing,” Dan noted.
The Rewind team has done some limited paid ad testing. While Dan said that those were profitable, they’re not a focus right now.
Instead the team has been emphasizing viral loops to show off the product. One example is a one-click way to share a draft meeting summary, which includes a footer denoting it was “powered by rewind.ai” (complete with a hyperlink to Rewind’s website).
Rewind also leverages their fast shipping cadence to create a constant drip of small launch events. “You can fall victim to thinking you don’t ship until you have something big,” Dan cautioned. “We use all of these events as a way to talk about what we’re doing. Some people just subscribe to us because we ship frequently.”
Thanks for reading Growth Unhinged! Support my work by subscribing, liking and sharing.
Creating a VC feeding frenzy
I had to ask Dan about the notorious public Series A pitch deck and fundraise. Why did he do it and how did he choose from the reported 170+ funding offers?
Dan emphasized that his main goal was to find a way to build trust with customers, and he believed that the best way to build trust was to be transparent. A public fundraise could show that Rewind was doing well financially (3+ years of runway) and had nothing to hide about its future plans.
His secondary motive was simple: to save time on VC conversations. As a buzzy AI startup, Rewind had been inundated with investors who wanted to meet. Dan figured he could put the deck online and only meet those who were serious about investing. The business had three years of runway before the fundraise, which mitigated any potential downside risk of fundraising in public.
Rewind’s pitch deck went viral. It had millions of views and attracted “well over 1,000 preliminary offers” including 170 formal fundraising offers. 22 of these offers were at a $1B valuation or higher – yet Rewind chose the offer from NEA at a $350M valuation instead.
Dan picked NEA’s term sheet because he sensed that they had a clear view of the market and they were in it for the long-term. NEA has 15 year fund cycles rather than the industry-standard 10 and the firm typically continues to invest at IPO, Dan pointed out. Choosing a more modest offer gives Rewind room to grow its valuation over time, too, rather than be faced with a potential down round in the future.
As we wrapped up our conversation, Dan offered one closing piece of advice for aspiring startup entrepreneurs and operators. “Most founders are afraid of failure, but they should be equally afraid of success. If we had stuck with Scribe instead of pivoting to Rewind we would likely be in a much harder situation.”
Dan believes that if something isn’t really taking off, perhaps there is a chance it will work eventually but it probably won’t. Being OK with a company that’s just doing “good” will stop founders from pursuing 100+ other ideas, some of which might be aided by tailwinds around the market, technology and timing. If you can build a startup on easy mode or on hard mode, why settle for hard mode?