How the app that was sold to Pinterest is made
January 5, 2022
Listen to the episode
Vochi is an app for adding effects to one’s videos. At the end of last year, it was acquired by Pinterest.
In the 11th episode of SubHub podcast, we talk to Vochi founders Ilya Lesun and Anna Buglakova. They told us:
- how it occurred to them to launch an app in such a crowded category as Photo & Video,
- what makes Vochi special,
- how they develop their app,
- what process of implementing new video effects look like,
- how they manage to use cutting edge technology on mobile devices,
- and how they’ve got acquired by Pinterest.
Listen to the episode in Russian or read the most interesting parts just below in English (translated automatically with DeepL).
How Vochi differs from other video editors
Anna: We try to reinvent the principle of video creation on the phone, not copying the desktop experience, but creating something more native, simpler, so that absolutely anyone can do cool things on the device for their videos. Our main trick is that we know how to do cool effects for individual objects in the video. To do that we use some neural nets inside our app.
How Vochi came to be
Ilya: The right point is that there are a lot of apps in the store. At that time, it was the most saturated category in the store (photo and video – ed.), with the largest number of apps there. I think the situation is the same now, although Health And Fitness has reached the same amount.
But there is one nuance: when I looked at all the applications in the Store, I was scared that everything is too the same, all carry the desktop experience into the cell phone, just simplify it a little bit, on the one hand.
On the other hand, it’s a bunch of apps that do filters and full-frame effects, effects that are pretty low quality.
The third thing is that at the time, the apps were really all the same. If we’re talking about video, you’ve got basic functionality, you’ve got filter overlay, full frame effects, music overlay, and little manipulation. Yes, there were apps that let you do a little bit more, yes, there were apps that specialized in conditional filters, did a very cool job, but I kept thinking that something was missing, some kind of personalization in terms of what you could edit, where you could edit.
The other point was: based on my background, I realized that a lot of the things that we used to do in professional video editors and motion design programs that we spent a lot of time on, you can already do it on your phone in preloaded presets, with the power of the same smartphone. This thought didn’t leave me until the last one. So Vochi was born as an idea, an editor in which you can select any object, work with it, apply effects of the same quality, at least comparable in quality with professional production.
There’s a lot of technology under the hood. There’s not just a few neural networks, there’s a lot of work on computational photography… There’s a big technological stack under the hood.
It was clear from the start that it was a very competitive market, and it would be very hard to grow there. But there was a lot of hope that we were trying to come out with a new product, with a new vision, with a new, unique feature that has blossomed into a big range of effects that have different technology under the hood, different quality, different capabilities.
Anna: I think it’s very important that there was a period when there was a boom of photo and video applications, but first of all it was a boom of photo applications, then TikTok had just appeared. Video started to get a lot of attention, and it started to grow wildly. The super power players in video were there, but they were pretty narrow or still complicated.
Why we got it somewhere was because it crossed over with another boom. The second boom was on all sorts of AI and AR stuff, where they were trying to shove technology everywhere, into every app. I think it just worked out at that point. As far as I remember, back then there were just the very first rumors about TikTok, the first people posting there. We started to make it, it wasn’t that much spread in the markets yet. At that time, all the projects that tried to put some very high technology into a device were booming.
How to hire if you’re a startup
Vitaly Davydov: In Belarus, in Ukraine, in Russia, there are a bunch of cool, big companies that offer a lot of money, good offices, and so on. How do you find a person who can program for iOS, knows machine learning, knows computer vision, and will work with you?
Ilya: When people came to us, they were all of a certain stack.
Example: if it was a guy from the ML team, he would come only with ML, and gradually he would start to understand graphics. When Vasya came in as the first iOS engineer, then he became head of mobile, then he became STO, but he came in with just a cool, classical knowledge of SWIFT and Objective-C. There was no Core ML background, there was nothing about graphics, shaders. It was all gradually being acquired.
It’s more about, when you’re looking for people, you’re selling them a vision that people believe in. But a vision just as a vision is not a working story. You have to give and salary for a period of time enough to cover a person’s basic needs and not embarrass them too much by previous stakes. And option. You’re talking about having an option that, under certain circumstances, might later make you an income that you wouldn’t get at the same big company.
Third, in our case the factor that our colleagues at Bulba Ventures have a very cool reputation in the market worked. That was the main factor why people came to us. We also had a tricky thing: from the very beginning, we called ourselves Top Secret Project Bulba Ventures and used that in all our communications right up to the launch, before the public press release. This helped a lot of people get interviewed and re-hired. Due to the fact that we had quite an open culture in the team, we learned how to call people – friends, acquaintances, former colleagues… The referral between people worked very well for us.
Anna: I would also add here about hiring. It seems to me, when you call a person – it doesn’t matter what salary, what position – to a startup, especially if we’re talking about giant companies, our CTO is from Viber, many guys from our mobile team are from Viber. We have UX designers from EPAM who came in. Often you can find very cool professionals at these companies who haven’t been given enough air and opportunities for self-realization.
For example, it’s wildly important for UX designers to be around the product, to lead it, to see it grow. To be a little closer to it than just – outsourced a few screens, gave them away, never found out what happened to them again. It’s wildly important for the mobile team, too. When you come to a startup, you can do not only your own little piece of work, you can try absolutely different professions, specialties, and specializations. That’s what a lot of people go for, because in the first year we couldn’t compete on salaries until the first big investment. But we were able to give a lot of this opportunity to implement our ideas, to participate in decision-making, in product discussions, and so on. The Bulba name helped a lot, too. That’s why we got the cool guys from the first year on the team who are still with us today.
Server or device
Vitaly Davydov: It always seemed to me that the biggest challange in developing modern ML algorithms is how to make it work on iPhone 10, iPhone 8, on such devices.
Ilya: Yeah, that’s true. We pretty quickly came up with an approach that allowed us to process it all very quickly. On older iPhones, we learned how to process it almost without losing quality, we learned how to process it as fast as possible by overclocking the processor as much as possible. Other apps will be slow in the background, but Vochi will work fine.
From the start we had a plan to make it work on the device, work fast enough, but it was only about neural networks. Neural nets work on object extraction fast enough. The next step was to make effects that contain a lot of different mathematical and visual operations, and we had to learn how to do that quickly, too. There’s a completely different approach with each effect, we even have some effects flying around in the app just… We technically reduced their processing speed because they work too fast. There are effects that are slower, and we cannot run them on the old devices, because there is not enough power.
How do you come up with the effects?
Vitaliy Davydov: The process is interesting. Where does the idea of the effect come from? What state do you have to be in to invent a new video effect? How does it go afterwards to realization, how is the roll-out pipeline of new filters arranged in your company?
Anna: In the beginning we didn’t have an effect designer in our team, we were looking for one for a long time, we spent several months.
The very first effects were purely mathematical: a white line that runs around the object, the background becomes black and white, the outline is tinted – nothing super complicated. With the first designer, we were already starting to try on some effects outside of the application. He was making some prototypes, we were looking for ideas.
When we were a very small startup, we had spectacular brunches every two weeks. We’d bring croissants, coffee, all the guys on the team would bring their effect ideas, found anywhere on the internet-whatever they liked.
We were putting together a big base from that. But back then we couldn’t do everything, we didn’t have enough technological expertise. We couldn’t compress the whole thing well enough and put it in a device. We’ve always had this marketplace of what we have to work in real time.
As our team grew, the number of ideas that were in the backlog grew, the number of competencies grew. New developers came in, they brought their competencies, which we were able to apply to improve the effects work. We plugged in new frameworks. When you start a new technology inside the whole effects process, you often have a huge number of ideas, which were in the backlog, unlocked, you can implement them today.
Next comes the issue of prioritization. The more opportunities you have, the more you want to implement everything, and you have a limited number of hands, you have to decide what will go into the extension today, what will not. Then a story helped us a lot… We were just beginning to monetize, we learned how to test the creatives at the purchase – we began to understand what users like, what they do not like. This helped us a lot in prioritizing. Plus the community in Instagram started to grow, we started to get good signals about what users liked and what they didn’t like. So we got some minimal prioritization.
Now the general flow looks like that, we have a backlog of ideas that are collected in different ways: the same brunch we still sometimes hold, people from the team can just throw it all to our spectacular manager, maybe some designer brought it in, we can find it in creative, test it in advance. After that, we have a prototype done by a designer. We have some super-hardcore designers who work in the effects team. They also know how to write code in After Effects, and they can make any 3D thing you need. They make a prototype, lay it out, speculate on that effect, and discuss it with the developers. They have bi-weekly meetings where the developers already tell you how realistic it is, how many months of research, if necessary, how it will work.
Further on, the more you do these effects… We’ve come to the point where we don’t write each effect separately now, but write engines for them. Often even the designer himself can make a bunch of other effects on the same engine with minimal changes. He can use the settings inside the application to change the assemblies that come with it, he can change the speed, the trajectory, something else, and sculpt something completely new out of it.
Acquisition of the first 10,000 users
Vitaly Davydov: How many users are there in the app?
Anna: Right now we have about 400,000 MAUs, but we are not buying right now, we took a break. At times when we had a boom, when the effects were flying around, there were more than a million MAUs.
Vitaly Davydov: How do you get the first 10,000 MAUs?
Anna: We’ve had user engagement since before User acquisition. When we first launched the app, we had the very first Soft launch in Canada, which we looked at how well our user flow was working. At that time we had already poured a couple of thousand dollars into it, just to check where it was going, where it was leaving, how users were getting through.
And so it happened that we filled a couple of thousand dollars into the first batch of users, we bought about 1000 users. So it turned out that they began to come. In the same Canada, the organic users began to come. There was some minimal virality in the app from the very beginning, even with the very first versions. Then we had quite a high-profile announcement of our launch in the CIS countries. As it happened, we got a lot of support there, and the app went viral. Then we constantly had a CIS audience, which came organically. We downloaded the app on November 26 and on December 31, we had 100,000 hits. They were all organic, except for the first 1,000, which we poured to test how the app worked.
Then we had the Arabic boom. We had a lot coming in from Saudi Arabia, UAE, also organic. Some Influencer… (mentioned us – ed.). Our first resource was because some of the creators post us somewhere. Since we’re new, there was a lot of excitement, it’s a new technology, it looks different, we got a bunch of users. And they, thus, we had them all the time. We never had a moment where we didn’t have 10,000 users in the app. Always some tens of thousands of MAUs we had. Then the moment came, it was a year ago, we realized that we had to scale, that organically we were growing, but very slowly.
Accordingly, you can’t accelerate very much on your own, without money, without investments. We had started looking for UA six months before, but we were not very good at it. It so happened that we didn’t find the specialist who had just magically started everything and it just flew away. Our partners from Genesis Investment, now Flyer One Ventures, told us, a little bit even insisted, that guys, let’s try to buy.
We talked to Ilya; he was not very interested in doing this, but I was quite interested. It turns out that the first purchase on the flow had to put myself. In about a month and a half of experimentation-the guys from the Foundation helped us a lot at the time, helped us with consultants, with the guys who did our own experiments. We set our first ambitious monetization KPI, which also helped a lot. We then reached out to our team, since we didn’t have a UA department, and asked everyone to get involved in buying, in creating creatives, in everything else. We got the buy in for the first quarter of 2021, and it started growing steeply, scaling, with a good ROI. That’s how it all started.
Nikita Maidanov: Tell us a little bit about monetization: how was it arranged all this time while you were buying traffic?
Anna: It’s hard to say how monetization was set up, because it was never set up the same way every month. It’s constant operations, tests. We had a weekly subscription, we came to it through tests. That one-week subscription had a pretty good retention that was already plateauing in week 14, give or take. This one-week subscription gave us good LTV at the user. Those subscribers who were our pro-users, they had very good product retention. Retention from video retention to video retention, from month to month, they also leveled off by the fifth month, they stayed with us all the time, and at a pretty high mark.
We experimented a lot, tried different approaches to onboarding, to paywalls. Scam pivots never worked for us, although we tried different ones, we tried both harder pivots and softer ones to understand what the conversion limit is, how much you can convert. In the end, we had a pretty fair peewall, we had onboarding that described the product pretty well, there was a little warm-up questionnaire that asked what kind of content the user preferred to do, which also helped us gather information, understand how to modify the app further under the users.
What was the question about monetization? It was. A constant iterating entity within the app.
Vitaly Davydov: What’s more important – the number of users or revenue in B2C?
Anna: I don’t like such flat metrics. The number of users – it sounds nice, we can say we have millions of videos in the app. But it seems to me that what’s more important here is the number of users which you brought in, they made your first video, they came back for the second one. If you go to B2C, you want to invest, you want to work, you think whether you need this option, the most important metric here is to understand how much the users are attracted to it, whether they come back for that incredible unique feature that the product gives you, at least every month.
Ilya: Yes, I agree with Anna that the engagement story is more important for B2C, if it’s a venture story. There are a lot of people talking about it. The last person who talked about it publicly was Andrew Chen. He says, work on engagement metrics. Okay, you can do it without monetization, okay, you don’t have revenue, you don’t have a growth point in terms of number of users, but you have engagement metrics on the cohort you’re working with, on the target audience you’re already working with, they’re great. If you show, prove that there are audiences in other markets that are just as cool and could potentially have the same engagement metrics, then the product is important. A product becomes valuable when it has cool engagement metrics, not when it makes a lot of money, or it gets a lot of users who fall off pretty quickly.
Nikita Maidanov: Tell us about the exit poll, how did it happen?
Ilya: Suddenly for everyone. Any startup has a question: it develops further by itself, raising the next rounds, increasing monetization, revenue, reinvesting it, or it becomes part of another company, there are two branches: either you transfer your technologies, experience, developments of this company, you just deal with integration, or you make your product part of the ecosystem of another company, you just get access to resources.
We were lucky: we were immediately offered that we would become part of the ecosystem of another company, which gives a lot of resources to grow.
Vitaly Davydov: How long is such a deal structured? How much time passes from the offer to the end?
Ilya: It happened quickly for us. I have not seen similar deals. Both in terms of speed and what was offered in the deal. I hadn’t seen one. Even when I was talking to other venture fund guys, angels, funders that exited, completely different. Quite often differently. The clearest example, mostly deals in the same CIS, is the story about acqui-hiring and the subsequent integration of the team and technology into the buyer’s product. Purely product exits, when the product stays inside, it stays the same, it’s engaged in monetization – there are very few such stories.
Vitaly Davydov: So the future of Vochi is not that it will be pushed into Pinterest, but that it will be an independent product?
Anna: We will remain a truly independent product. We’re not saying anything about the press release that Pinterest had right now. There’s really a challenge right now to empower creators, to help creators unleash their creativity. And they’re very interested in the topic of video. Basically what Vochi is about, what we talked about at the beginning of our post, we want to reinvent the story. We want to make it so that everyone, no matter what background, no matter what skills they have, can make cool videos on a device. The most important thing to understand about this deal is that it’s public information, we had a very cool match about our shared future vision of how mobile video should evolve. I guess that was the spark that then ignited this whole speed, all the processes that were going on. We just swiped right, it turned out to be a cool mash-up, and then there was just this cool motivation on both sides to make some kind of big story out of it.
Ilya: Yeah, that had a big impact on the speed of how it all happened. There was a need for Pinterest to make it happen quickly, because in any deal, in any investment, it’s no secret to anyone that there’s an exclusivity term, within which you can’t sign other offers or communicate with anyone. Due to the fact that the two parties had a match happen, everybody was very motivated to get it all done quickly. That’s why it happened quickly.