From 0 to 25K in 9 months: How we did it?

I’m a serial entrepreneur who has founded 10+ initiatives and 3 startups. These have reached ~3M paying users throughout the years.

When people say – it’s hard to get from 0 to 1 - I disagree. You just have to validate the idea well.

I’ve done it several times, but no other project has had a stronger start than the current one: Team-GPT.

• 9.5 months
• ~$200K in revenue
• 25K users
• 3K active businesses
• Used by: Maersk, Johns Hopkins University, Yale University, Shift4, Charles Schwab, and more.

I wrote a bit about this in the From 0 to 10K users and I suggest you start there.

Here’s how we did it.

The experiment

GPT-4 came out, we tried it, and we were crazy about it. For a couple of days, we tried to incorporate it in our work but talking to ChatGPT was a very personal process: always 1-1 with the AI.

We really hated the lack of collaboration as it was hindering out ability to adopt the technology as a team.

With a strong product team already in place, it took is a couple of days to come up with an internal tool, which allowed us to collaborate with the AI.

The idea was simple: collaborate in chats, organize in folders, save prompts for future use.

I had just read the book ‘Pretotype it’ by Alberto Savoia (first engineering hire of Google). It contained a strong idea: the XYZ hypothesis, which states you should test product ideas with: ‘At least X% of Y will do Z’.

As a data scientist with 1.3M AI students, I loved this: a hypothesis testing approach for validating product ideas! Let’s put it to work!

Our initial hypothesis?

Once you have a hypothesis, it must be tested with real clients:

1. Create a website (landing page)
2. Add a payment option
3. Promote it to someone…

Who do you promote it to?

We had several targets in mind:

• Marketing teams and agencies (who create a lot of content)
• IT teams (who write code)
• AI startups that are very open to AI tools

At a price point of $29, paid one time – it sounded like a no brainer.

And it was!

We hit $300 in sales in the first two weeks…

Biggest surprise?

We didn’t even promote it. People found us organically on Google. We had a simple WordPress website, optimized for the words ‘collaborate in ChatGPT’, ‘ChatGPT for teams’ and several others.

And yes, we got lucky with the name (team-gpt.com cost $10 back then), and I understand this will not be the case for many others – but this is our story, and I can’t change it. I’m just grateful it happened.

They say ‘build it and they’ll come’ is the wrong approach… because they never come. Usually true.

In this case, they came before we built it.

There was no product, yet we were making sales. There is no stronger indicator that you should build it.

Soon, we raised the prices to $39/lifetime, $49/lifetime, $10/team of 10/m.

Right now, we charge $2/user/m, charged in bulk for 10 people, so $20/m for a team of 10, scaling up to $1000/m for a team of 500. The pricing is simple, yet effective.

While there was the potential to charge even higher prices, we decided to keep growing the user base at a lower price point.

Product development

After the first clients came, we had 2 choices:

1. Refund them and start thinking about the next steps
2. Propose that we refund them and build it immediately

We asked… and the people wanted the product, not the refund.

So, we built it on top of the internal tool.

We used a tech stack of:

• An open-source project that we started building on
• Clerk for client authentication (this literally brings the ‘team plan’ out of the box)
• MongoDB Atlas (a scalable cloud database that allowed us to take any number of users)
• Lemon Squeezy for payment processing

The first version was out in a couple of days, built entirely by Ilko and Yavor (the other two co-founders of Team-GPT), and it was a system with many bugs, broken interface, practically no mobile version, and some of the worst tech debt in the universe.

But a good old startup saying goes like: ‘If you don’t hate the first version of your product, you launched too late’.

The product was working and most importantly, it was solving a big pain: collaboration in ChatGPT.

9 months later we’ve got a great system with hundreds of features, based on customer feedback.

We’ve also built a separate B2B app, that caters to Enterprise clients for whom the initial tech stack was suboptimal.

And remember - you can always change the tech stack if you need to. Just don’t overkill from the start.

Wouldn’t ChatGPT build it? They did!

Truth is – every business is a limited time opportunity. Then it iterates and reinvents itself.

I had just read the story of Ruslan Leteyski, who had made $10M+ from an app that enhances the checkout experience in Shopify. They reached $600K ARR before Shopify shut his operation down and that’s a pretty good outcome.

This showed us that no matter if OpenAI does it in 5 months or 2 years, there is no point in us missing out on this opportunity.

And we knew ChatGPT would not copy us – they have their own plans, nor would they shut us down – we were bringing fresh users, paying for their services. In fact, we were capturing a segment which OpenAI was not serving, making them money, and popularizing the technology. We were not a threat – but a partner (users are paying OpenAI directly for the model usage). Every other business built on top of GPT models was in the same boat as us and that’s why the OpenAI APIs exist.

It took OpenAI 9 months to release a team version and it is still pretty basic. Here’s how ChatGPT for teams compares to Team-GPT.

This gave us 9 months to think about what we want to pivot (if we needed to).

Was a pivot needed?

We were well aware that we can’t be just a collaboration wrapper on top of ChatGPT.

3 days after we released the first version of Team-GPT, I was already on my way to San Francisco, going around meeting people. I met with people from OpenAI, Google DeepMind, AI startups, literally anyone who could give me information. The main question?

‘Where is this going, what does the future look like for ChatGPT’

After realizing that ChatGPT’s success was a shock even for OpenAI, I knew they weren’t going to rush to compete in this SaaS realm. Time was on our side. Google was catching up in terms of models. Amazon invested $4B in Anthropic – to compete on the foundational models, too. The world was on fire and the big players did not really care about the small gains of SaaS software.

They all cared (and still do) about AGI.

In the last 9 months we’ve been conducting many customer interviews and focused on building the best experience for our users. We listened to their feedback and we incorporated it – and they love us for it!

You read that right – adoption was the key.

Adoption is the key

For a product to be used – it needs to be understood and adopted.

What could we do better than OpenAI?

Education.

Based on our team’s previous experience, we knew we had to bring something more to the table – and we did.

We focused our whole product around adoption:

1. The online course

I’ve created 30+ courses, with some of them having 600,000+ students, 130,000+ reviews (see the The Data Science Course).

So, we rushed to build an online course. I disconnected completely for 2 weeks and did what I do best – an online course.

ChatGPT for Work: The Interactive Course was built in record time (thanks to using ChatGPT of course).

And we released it for free.

A free online course, built by a top instructor - all inside the software.

Every new user of Team-GPT would go on the platform and be urged to take the course.

And when they take the course?

They’ve learned how to communicate with the AI properly.

Not only this.

They’ve learned how to use Team-GPT.

And since the product is rather good – they simply stick with it.

This became a major part of our marketing efforts.

Users come for the course, stay for the software.

Or… they come for the product. They stay because we taught them to talk to the AI.

Many of our users prefer Team-GPT over ChatGPT because we helped them see the value.

And that’s powerful.

2. Prompt library

Nurturing the usage was key from Day 1. How do we help users prompt better?

By preselecting useful prompts for them and providing them for free as a part of the product.

The prompt library kickstarts the users’ prompting journey and supercharges their prompts.

The feedback we got was that people also use our prompts to learn to prompt better (more education).

Once they are proficient, users can also create their own prompts and store in their ‘Workspace library’.

3. Tips and tricks

Each chat boasts tips and tricks in the middle (see above).

Every time you start a chat, a tip is shown which further helps your learning.

The more you use the product, the more tips you see (it’s like a computer game).

4. Adoption reports

So, we are very good at adoption (nurturing the usage of the software).

How do we know? We measure it.

And since we were already measuring the adoption, we realized there is nothing better than showing these reports to our users, too!

This greatly helps all admins and managers nurture the adoption of the software throughout their organizations.

Much of our adoption success is due to our ability to make users use Team-GPT.

And this is where we are so aligned with our clients (the businesses).

Everyone knows ChatGPT is a real deal.

But Team-GPT cares about adoption.

And adoption is exactly what companies want at the end of the day. Not only to pay for the software but to make sure all employees are AI-enabled.

User acquisition

While user acquisition is crucial for a 0-25K users, I’ve covered this well in another one: From 0 to 10K users | Team-GPT Getting Ready to Scale, so I’ll spare you a lot of it here.

In fact, most of it was carefully following the plan outlined in this 16-min onboarding video for our sales and marketing hires: Sales and Marketing Strategy: How to Achieve our Revenue Goals??

We are experienced marketers, and we know there is 1 thing that’s most important: MOMENTUM.

All marketing campaigns need to follow one after the other.

And when they are planned well, they create a chain reaction.

My favorite channel? The one that brought the first users: SEO and content marketing.

We saw the opportunity and grabbed it.

We’ve invested more than almost any other early-stage startup in SEO and this was a bet, which is paying off tremendously.

Today we are getting 1,500 clicks and 25,000 impressions on Google every day.

We are ranking for thousands of words and are on first position for many of popular ones:

In fact, ChatGPT releasing their team plan has not hurt us.

On the contrary, they’ve generated demand, bringing even more traffic to us.

Social

Once you have so much content and you are good at it, it is a shame not to promote it all around the place.

Our main success is on YouTube, Instagram and LinkedIn (this article included).

We’ve published hundreds of short videos.

The rule us: reuse, reduce, recycle - or as we call it in marketing: repurpose content.

We have repurposed so much of our content and are posting it all around the place.

Here’s what our YouTube metrics look like.

Instagram and TikTok are hit and miss but we did have a fair share of viral videos, reaching 6,752 followers.

We don’t spend much time on this, we just publish content we already have.

Finally, LinkedIn.

This one takes the most time for me, but it is crucial for our long-term success. LinkedIn is the professional network where most of our users check if we are real people – and that’s really the case when you sell to businesses.

I’ve been regularly posting for about 1 year and here’s how many people I’ve reached.

Now LinkedIn has put more eyeballs on our stuff than YouTube and Instagram!

I’ve never had a viral post (my best posts get ~15-20K impressions, while the average is ~2K).

I’m not in a rush to grow faster - I keep the content relevant to our audience and align it with the business goals.

Retargeting

Retargeting is extremely underrated.

I don’t know why people think ads are expensive.

Ads out of the blue are.

But once someone lands on your website, it is rather cheap to show them an ad (retarget/remarket them).

We have spent a bit below 920 EUR to bring back 41K users and show our ads 5.5M times.

We are spending around 5-10 EUR/day for 10K impressions, which are RELEVANT to the people we are targeting.

We are aiming to show a certain ad < 5 times per person, and we’ve got 30 variations that we shuffle every 10 days.

The main point: make people remember you, without being annoying.

25K users: Digging in the data

One of the biggest questions is not how many users you’ve got.

It is how active they are.

First – how do you measure it? We are using June for this – took us a couple of hours to set up and the rest has been a breeze. It’s a bit simplistic but does the job.

Second – what do you measure?

When it comes to GPT usage, the relevant metric is WEEKLY BUSINESSES. We (and OpenAI) consider someone an active ChatGPT user if they have used it once per week.

Our weekly active base is growing fast, showing that they actually use the software.

Right now, we’ve got around 3K weekly active businesses using Team-GPT.

The top use cases, based on self-reported data:

• Marketing
• Education
• IT

These are the use cases where people immediate see the value from using GPT models. The early adopters of the GPT technology.

The Enterprise opportunity

Do you remember when we spoke about what we would do if OpenAI introduce ChatGPT for teams?

We had another hypothesis.

Enterprise clients would not want to share their data with OpenAI.

This is how we built Team-GPT Enterprise – the ability to deploy Team-GPT on premises or on private cloud and hook it up with any model (incl. Microsoft Azure, Google Bard, LLaMa, Mixtral, and more).

As teams from bigger companies were starting to use us, we were getting the following question:

‘Do you offer a private deployment of your software?’

And yes – we started offering it.

Enterprise deals are not easy, though.

We have a pipeline of 20+ companies that are seriously interested in the product above, with 10+ inbound (SEO-driven) inquiries each week.

But it takes time to convert and Enterprise client – 6-12 months on average.

The good thing?

While they are considering the Enterprise option, we hook them up with the ‘$2/user/m’ product. Today we’ve got 20+ companies with 60+ active users (4 with 250+ users), negotiating a private instance to switch to.

We know that the big opportunity lies in Enterprise, and we are gearing up to scale our sales efforts. The negotiations are never easy but once they convert, we will be into the serious business of Enterprise software.

The future: Time to seize the opportunity we’ve been given

I’m a big optimist about the future of Team-GPT… as every startup CEO should be.

All signs are positive, and we’ve done most of the things right (many mistakes were made but much less than with my previous startups).

We've got a strong team.

It’s getting easier to acquire new users.

It’s also easier to attract top talent.

We know where the opportunity lies – and clients are finding us on Google.

Our next goal? Find the right funding partner for our big plans. We’ve got a the core team in place and are ready to receive a big financial boost.

10 years from now, most work will be done by AI agents, while humans set goals, monitor and approve their actions.

Team-GPT will be the place for this human-AI collaboration.

??

Thanks to everyone involved!

Ilko Kacharov

Yavor Belakov

Simona Dobreva

Katya Vaptsarova

Maria Valchanova

Jeko Kolev

John Ozuysal

Margarita Arsova

Marta Kostova

Marta K.

Alexandar Galabov

Camplight | Your Digital Venture Builder