If you’re in any sort of publication profession or podcasting service you’re probably very familiar with transcription. The ability to convert spoken words into text is valuable for keeping searchable records of conversations. It’s also become a common tool for repurposing content online, letting you easily turn audio into blogs and visual posts.
It’s hard to believe that only a decade ago, it was still common to transcribe audio completely by hand. Today, an AI with a clear audio file can produce a transcription at 95 percent accuracy almost instantly. But despite leaps and bounds in AI, would it surprise you to know that most transcription services still rely on people to type out speech manually?
Rajiv Poddar is an Indian engineer and the founder of Scribie: a manual transcription service he founded in the late 2000s. Over 13 years in a rapidly-developing industry, he created a proprietary four-part transcription system he thinks could be the most effective on the market.
While he thinks he is poised to take on the best in the industry today, Rajiv spent his first decade barely staying afloat. After years of developing his product, he watched helplessly as larger, VC-backed businesses passed him by. He needed a change both in his business and personal life and he needed it fast.
In a gambit to open his mind by force, Rajiv took a fateful trip (in both forms of the word) to ingest magic mushrooms in 2019. He believes his first psychedelic experience not only changed him profoundly as a person, but also helped him uncover the missing element to his business that rocketed him to $1 million in annual revenue.
Here’s his story.
The Bedroom Project That Lasted a Decade
Rajiv already had a successful coding career under his belt by the time he started Scribie in 2007. He’d worked for Nokia networks back when it was still called Lucent and was an IT Analyst at Tata Consulting Service for about three years.
In 2007, after a previous venture failed, Rajiv started Scribie in his bedroom as a last ditch attempt to keep the lights on.
“There was a lot of demand for recording plugins in 2007 though we only had Skype,” he says. “I figured most customers would pay good money for transcriptions.”
Rajiv was right. Back then, customers were paying big money for transcription services. When he started Scribie, he charged roughly $90 per recording (about $40 per hour). That’s a far cry from the prices you see now.
Despite transcription technology’s dramatic growth over the next few years, Rajiv kept the project small and didn’t even buy office space until 2013. He’s the kind of guy who only needs a bedroom project and his computer to stay happy.
“I started in my bedroom, got married in 2010, and moved to a different bedroom,” he jokes. “In 2013 I got an office but still worked out of my bedroom most of the time. I had nothing better to do and it was paying my bills so I kept working on it. I never stopped until 2019. I would work all day coding and experimenting with things. Everything that a coder likes to do. I had a great time.”
However, in 2019, Rajiv suddenly wasn’t having a great time. He’d had personal problems crop up and transcription services from big tech and VC-backed enterprises were using AI to cut into his slice of the pie. He just couldn’t seem to scale the business he had been running for over 12 years.
“One of the problems we faced was we only had a manual service,” he says. “However, whenever there is human touch there is a chance of error. We completed our system in 2013 I was still trying to scale it in 2015. The problem was I was trying to use engineering to scale a manual system and it was unsuccessful. Somehow our competitors had scaled the manual side better than we had.”
He needed a big change and he thought he knew how to do it.
Doing ‘Shrooms to Unlock His Potential
Desperate to expose himself to new ways of thinking, Rajiv decided to try magic mushrooms in 2019.
“I’d read articles on people connecting to themselves through ingesting psilocybin,” he says. “The mushrooms that produce this chemical grow by the thousands in the wild in Kodaikanal and there is a large underground market.”
Intrigued by what he saw about its effects, Rajiv began to research this supposed “miracle drug.”
“Microdosing has been big in The Valley since 2007. There was this paper which came out in 2006 about it which I’d read. I always thought in the back of my mind I should try it,” he says. “Recently a book came out called How to Change Your Mind. I recommend anyone who wants to do mushrooms to read that first. It gave me a very good base and it has excellent research. There’s a systematic method of doing psychedelics. You have to be knowledgeable and it may not work for everyone.”
Satisfied with his findings on the subject, Rajiv went on a fateful trip to Kodaikanal to pick mushrooms and have his first psychedelic experience. He says it was exactly what he needed.
“The thought process brought on by the mushrooms made me realize what was preventing Scribie’s growth wasn’t an engineering problem. It was a human problem. I was trying to treat the humans who transcribe as bots. As an engineer you want to do that,” he says. “Humans are humans. You have to treat them as humans. Be compassionate and allow them to improve. That is the switch that flipped in my head.”
Rajiv claims psilocybin didn’t just change how he viewed his business. It helped him make a slew of other changes in his life.
“I have done it once more but that first trip had most of the benefits,” he says. “There have been a lot of changes. I stopped vaping, I stopped drinking, I lost 10 kilograms, I’ve started doing yoga and meditating.”
“Humans are humans. You have to treat them as humans. Be compassionate and allow them to improve.”
Rajiv made changes to his business model and then COVID hit. With the advent of Zoom, everyone wanted to transcribe their video calls and demand for his service went through the roof.
How Transcription Works Today
Despite huge leaps in AI, Rajiv claims the best transcription services today still need humans to reach the highest levels of accuracy.
The reason automated transcriptions can’t quite measure up today is that they still depend heavily on audio quality.
“Zoom’s auto-transcriber has about 95 percent accuracy,” says Rajiv. “If there’s background noise that can be a confounding factor, however, manually we always make sure it’s 99 percent.”
That marginal level of accuracy makes all the difference to manual transcription services that often have no way to compete financially with AI transcribers from tech giants. Today, industry prices are a far cry from the $40 per hour Rajiv used to charge.
“I believe Amazon’s automated transcription averages 0.002 cents per minute,” says Rajiv. “They can’t make money at that margin. I think they either want to rule the market or they want customer data. That business model only works out if you are massive like them.”
Rajiv claims that because he can guarantee such a high level of accuracy, internally he doesn’t spend much time measuring it. He says a much more important determining factor for most transcription services is customer satisfaction.
“We say that even if the transcript is not 99 percent accurate, you can come back to us and we’ll make sure that it’s right. And so we don’t see a lot of customers complaining about the accuracy,” he says.
Scribie’s Proprietary Four-Step Transcription Model
According to Rajiv, what makes Scribie so different from its competitors are two factors: its heavy reliance on human capital for transcriptions and its signature four-step transcription process.
“We break audio files into small pieces in four steps,” says Rajiv. “Most other systems just require two people. The first person is usually low-skilled and the second one is highly skilled. We used to use that system but then added a third and fourth person.”
While Scribie still uses four people to transcribe today, they have evolved with the times. They now do an initial transcription with an AI before moving it down the line.
“AI came into the picture around 2017,” says Rajiv. “Our first transcription is done by AI, the second is a lower-level transcriber, the third is higher, and the fourth is usually great. Finally, we do a quality control step. That’s done by the people who are the best according to our system.”
The Scribie transcription system is all about encouraging transcribers to reach the “top-level” where the money is good and the hours are short.
“We try to optimize how much money our transcribers will make at the top level,” says Rajiv. “Usually you just spend one hour at the top for the most money.”
According to Rajiv, Scribie has roughly 7,000 transcribers actively transcribing on a month-to-month basis. It uses its large pool to its advantage by pushing transcriptions to those who are available and letting them select based on what they want to do.
“We have a very low-touch system for writers and we make it easy to onboard,” says Rajiv. “Whenever a transcriber logs in they get to choose what they transcribe. We sort the files by who is most likely to do them. We have a good system to tell who is good and who is not. As our transcribers work more and more, we collect data on them. We can figure out who is good at transcribing which accents and speaking styles.”
Here is the “human element” that Rajiv has integrated into the service. While he used to focus on trying to make his transcribers accurate no matter the audio quality or accent, Scribie now focuses on playing to each transcriber’s strengths.
“From an engineering perspective I wanted consistency but to do that with all different kinds of people is impossible and inefficient,” he says.
In 2019 Scribie was just Rajiv and two contractors in the Philippines managing operations. Today they have a six-person team with an office at a coworking space in California. Rajiv says at last check Scribie had 39,000 transcribers who had passed their preliminary test with 7,000 who were active every month.
For both finding leads and new transcribers, Rajiv says he has had systems in place from the beginning.
“We get most of our leads through organic search even though we haven’t done anything for it,” he says. “I think that’s because we are very old and have done some blogging over the years and Google rewards that. Most of our customers come through that channel. We’ve never reached out to transcribers. We created a referral program to bring them in that’s been in place since 2006.”
Rajiv is proud of his system today and thinks he has all the pieces in place to go head to head with anyone else in the industry.
“That business thing has fallen into place,” he says. “We hit $1 million in revenue in 2020. Now in 2021, if our system is better, we should be able to beat the others. Now we get to find out who wins.”
How Rajiv Wants to Expand Scribie
Rajiv thinks Scribie today has passed the threshold for exponential growth.
“Once you cross $1 million you realize something has happened. The next step is $5 million,” he says. “I think we need new management. A fresh way to do it. A new wave is coming. We have a couple of competitors that are growing rapidly and our rate has been low. Now is the time to catch up. I’m excited.”
While operating completely bootstrapped over the years, Rajiv has looked into venture capital before with little success. He thinks his mindset at the time got in the way.
“We did try to raise funds in Silicon Valley once. At that time I went around to all the major places and made it to an interview with Y-Combinator. The president, Sam Altman, was on my panel. Everything went well and at one point Sam asked me, ‘What do you think this could be?’ I think the question meant, ‘Do you think you can be a unicorn?’ I thought not at the time, but I wonder what would’ve happened if I’d said yes.”
These days, Rajiv thinks he really could be a unicorn and may have reached a size where venture capital would be helpful.
“I now know why VC hasn’t worked,” he says. “Maybe with new management we will try. Funding does do things. With funding it wouldn’t have taken me 10 years to reach the point I’m at now. Funding accelerates your growth. Our competitors may have scaled so fast because of VC as well.”
“If you don’t love it, don’t do it. Do something you enjoy.”
Rajiv’s advice for new founders is this:
“There are times when you question yourself, ‘Are you working on your project because you love it?’ If you don’t love it, don’t do it. Do something you enjoy,” he says.
Finally, he believes that bootstrapping is a great mindset for a founder to have.
“I believe in bootstrapping,” he says. “When you are bootstrapping you need to be very frugal, and frugality is a great tool. It’s a good thing to be as frugal as possible.”
It’s hard to change someone else’s mind and it’s perhaps much harder to change your own. Scribie’s success shows the benefits of seeking new insight and perspective by any means necessary. If you want to make a great company, you need to think about all the possibilities.
Want to share your bootstrapping story with the world? Enter your details in this Typeform and we’ll be in touch.