Harry Brodsky built a screen-recording app that became popular with teachers during Covid. Then, he pivoted his ICP and expanded the feature set. Now, Kommodo is bringing in a five-figure MRR.
Here's Harry on how he did it. 👇
I got started in building mobile apps for higher education. There was a clear need for faculty to record their lectures, and the process was complex, so we created a simple screen-recording app in 2019. When COVID hit, lots of school teachers started using the app to record their lectures and upload them to places like Google Classroom and Canvas.
In the process, we realized that every team — not just teachers — has knowledge that needs to be captured, documented, and shared. The teachers were recording lectures, but the same problem existed in every company: onboarding, SOPs, product demos, bug reports, client walkthroughs. Video was the most natural way to capture all of it, but there was no tool that took you from recording to searchable, reusable documentation in one place.
So, my cofounder, Khanan Grauer, and I pivoted the product into Kommodo.ai.
Kommodo started as a better way to do screen recording — unlimited, no watermarks, no time caps. But it's evolved into something much bigger. Today, we're an all-in-one platform where teams can record their screens, automatically generate step-by-step guides and SOPs, capture meetings with AI transcription and summaries, and search across their entire video libraries using an AI assistant. Think of it as the place where your team's video knowledge lives, gets organized, and becomes actionable — without needing five separate tools to do it.
We're bootstrapped with a five-person team. We're trusted by over 100,000 users, with a five-figure MRR.
The initial product was just an iPhone mobile app. It let you import a PowerPoint or PDF file, then broke them into pages. The user could record on each slide and talk through it. Then, we'd create a video link.
Now, we also have an Android app, desktop app, and a Chrome extension.
Our stack is a combination of GCP, Cloudflare, and Digital Ocean.
Next.js (React/TypeScript)
Firebase (auth, database, analytics)
Electron desktop app
Chrome Extension (Manifest V3)
Stripe for payments
GCP / Digital Ocean / Cloudflare for hosting & CDN
FFmpeg for video processing
OpenAI for AI features
Analytics have been our biggest challenge.
There's no off-the-shelf tool that works for us — our usage patterns are too specific to the platform. So we had to build our own visibility layer to understand if changes are actually working or quietly breaking things for users.
The first version taught us how little we knew about event tracking. And then a new problem: once you have the data, what do you do with it? We spent a lot of time trying to find causation in the numbers — what's actually driving behavior versus what's just noise.
We haven't fully figured it out. It's still a constant learning environment.
If I could go back, I'd instrument everything from day one. Don't wait until you need the data to start collecting it.
Our business started with a generous offer of unlimited recordings for free users. The bet was that 1 to 2% of users would subscribe and cover the rest. This is similar to the WhatsApp model (before the Meta acquisition) of free texting while monetizing power users.
But it didn't work. It brought in lots of organic traffic to get the flywheel started, but as a monetization approach, we had to pivot.
Today, we let anyone record up to 15 videos for free. We are still experimenting with different monetization models. Another experiment we ran was gating content after 60 days.
Lesson learned: It helps to grow revenue by catering to power users. These tend to be closer aligned to our ICP. Upselling users who don't have any intention of subscribing does not work.
The number one way people find Kommodo is through a link — someone shares a recording or a guide, and the recipient sees the product in action before they ever visit our site. Our users are our distribution.
Beyond that, we focus on one thing: solving real problems. Customers talk when something actually helps them. It's a slow strategy, but it compounds.
We also built a set of free tools — no install required — like a quick screen recording tool, a teleprompter, and an SOP-from-video generator. These attract people who are actively searching for a solution, not just browsing. It's high-intent traffic that converts better than anything we've paid for.
Speaking of paid — we tried ads. They didn't work. At our price point, the cost of acquiring a customer through ads exceeded what they'd pay us. So we stopped and put that energy back into the product.
I handled support myself for a long time, and it turned out to be one of the most valuable things I did.
Every conversation is a free user interview. You learn how people actually use your product — not how you think they use it. I made a habit of asking two questions every time:
How did you find us?
Why do you need us?
The answers shaped our messaging, our SEO, and our roadmap more than any strategy session ever did. You can't outsource that kind of listening.
Books and podcasts are super valuable, but they alone won't help you figure out what to do or how to be successful. Every success story is unique. Therefore, I created a framework that we use to sort of become a "learning machine."
I start the day with 1 question: "What did we learn from the last 24 hours?" If the answer is blank, it means we don't have mechanisms in place to learn.
Take analytics, for instance: If we shipped feature X, what is the outcome? How many users tried it? Does it work? Does it crash? Is there repeat usage? Does it trend up or down? Have our customers reached about it? Did anyone complain or compliment? If you get analytical about steps you take on your platform you can start to surface insights.
The next question is: "What action are we taking from these learnings?" The action should be something that builds on what you learn.
So, for instance, feature X has a negative effect on conversion because we offered something for free and that was enough. After 24 hours of data, we may decide to give it another 24-48 hours or gut the feature. This helps us remove things that people don't use.
Having a constant pulse of learning every 24 hours helps boost velocity and focus on the important stuff. Do what is working - and stop everything else.
From here, I plan to grow the business by serving our customers, figuring out distribution, and shipping products to help our customers be more productive.
Check us out at kommodo.ai!
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。