It's no secret that we're planning new software and will be launching it soon. So I decided to put together a list of things I want to watch out for—because we've already been through one SaaS project. That project is Editee, and I honestly think it taught us absolutely everything we know today.
When Štěpán and I started Editee (formerly called Deeply.cz), we knew absolutely nothing about the SaaS business as such. We started completely from scratch—we built an MVP, launched it, and managed to get it off the ground. After all, we had and still have extensive business and marketing backgrounds, so the foundation was there 🙂. Along the way, we dealt with many challenges where I now know we wouldn't make the same mistakes again. But as the saying goes, hindsight is 20/20.
That's why I wrote this list. These are 14 specific things I want to avoid, watch out for, and definitely not repeat. So let's get started.
1. No unlimited plan and strict cost tracking
This is probably the most pressing issue for us right now. We discontinued Editee's unlimited plan because it simply stopped making sense for us.
When we founded Editee (then still called Deeply), an unlimited plan seemed logical. The prices of language models and image generation weren't nearly as high as they are now, and the overall computational demands were manageable. But since then, prices for some models have increased fifty- to a hundredfold. The computing power we need today is incomparably greater, and so are our costs. In return for these increased computational demands, however, we provide outputs that are many times better than ever before.
Through ongoing monitoring, we found that some users were using the service excessively—far beyond what we could cover with a reasonable margin. We had addressed this in our terms and conditions, where we reserved the right to change the plan and defined a so-called fair use policy. We didn't enforce it for a long time, but eventually we had to—because some people were using bots, deliberately overloading the service, repeatedly sending the same prompts in chats, and so on.
Out of thousands of users, this involved only a few dozen people. Everyone else, of course, remained completely unaffected. But those few dozen problematic customers generated most of our monthly costs—we discovered that roughly 30–40 people actually consumed more resources than thousands of other users combined over an entire month. Because of them, we had to maintain better servers, pay more for computing power, and they disproportionately increased our operating costs at everyone else's expense.
I remember a similar situation from when I worked on the Rousinovsko.net wireless internet service project. We had a few “heavy downloaders" who were capable of placing an extreme load on the fiber connection. They paid exactly the same as everyone else, but slowed it down for everyone and generally ruined the service itself. We had to apply a fair use policy and limit their speed. Naturally, they didn't like it and left—but then they came back after realizing there was no better alternative.
The lesson is clear: I wouldn't offer an unlimited plan in any future SaaS project. Especially if content generation, computations, or servers create variable costs. I guarantee that there will always be people who abuse it. It's not just in the Czech Republic—it's everywhere in the world; it's human nature. It's always only a matter of time before it happens.
2. Security from day one
From the very beginning, I would definitely implement every form of security available to us. I won't disclose exactly what we do or how we do it, which logs we monitor, or which background metrics we evaluate—I obviously don't want to reveal our security architecture.
But I'll mention one example: someone created a bot that bypassed the standard Google reCAPTCHA v3 and began creating thousands of fake registrations within a single hour, deliberately flooding our database. Honestly, it had never occurred to me that bots could bypass standard reCAPTCHA so easily—but they can. We had to secure the entire registration process with additional mechanisms.
Later, I researched the issue online and discovered that bypassing CAPTCHA systems has historically been a fairly well-documented problem. As early as 2014, Google researchers showed that machine learning algorithms could defeat their own reCAPTCHA system with an accuracy of over 99%. It has only gotten worse since then—today, entire services offer CAPTCHA solving as a paid service for just a few cents per solution. That's why it's important not to rely on a single layer of protection and to have additional verification mechanisms in place.
So this is definitely something to watch out for.
3. Protection against DDoS attacks
Then, of course, you need truly robust protection against DDoS attacks. Some people are simply bored, and we've had to mitigate these attacks many times.
Unfortunately, DDoS attacks are an increasingly common problem—according to Cloudflare reports, the number of DDoS attacks worldwide has increased by tens of percent annually in recent years, and today they are one of the most common types of cyber threats faced by virtually every online business.
Thanks to the strategy we have in place, we've always mitigated them very quickly. In most cases, there wasn't even any downtime. But once, the service was unavailable for several minutes because, for some reason—even though we had mitigated the attack—Linux on the load balancer crashed completely. I had to restart it, and then it came back up. But we've experienced quite a lot of DDoS attacks, and it was very unpleasant.
4. Rigorous codebase testing
When we were initially prototyping Editee, we didn't test much of anything. I admit that testing was never something I did extensively—especially because I didn't want it to slow me down at the beginning.
But that was the wrong approach. Today, we rigorously test the new application—and every other application we'll build—not only the backend but also the frontend. Tests can reveal that, while developing a new feature, you changed a method and broke something somewhere else. You'll detect it in the tests instead of shifting the burden onto customers, hunting through logs, or relying on Sentry monitoring.
The whole concept of automated software testing began gaining traction around the turn of the millennium with the rise of agile methodologies. Today, it's effectively an industry standard—and if you ignore it, as we did in the beginning, it will catch up with you sooner or later, and you'll eventually have to deal with that technical debt.
Definitely test your software; with AI, it's easy.
5. Rapid MVP development and immediate production release
We do everything quickly. Even now, we're prototyping the new application, launching an MVP, testing it, moving it into production quickly, and ironing out the bugs along the way. We also continuously improve everything as we go.
Definitely don't spend too long thinking about what you're doing. I've seen many entrepreneurs develop software for several years. I recently read a Facebook post by someone who wrote that he had invested 5 million Czech crowns in developing a SaaS application and spent three years building it—only to launch it, discover that nobody wanted it or that he didn't know how to sell it, and shut the project down a month later. Three years of work, an investment of five million Czech crowns, and zero results.
The statistics confirm this—according to CB Insights data, “no market need" has long been one of the most common reasons startups fail. Roughly 35% of startups fail because they created a product the market didn't need. And the longer you develop something without feedback from real users, the greater the risk that this will happen to you too.
I therefore recommend rapidly developing an MVP, getting the product in front of people as quickly as possible, and testing whether there is demand for it. Those people will then guide you themselves—they send feedback, tell you what they want changed, what they don't like, and give you better ideas. We built some of our features entirely based on customer insights. But be careful: always think everything through carefully. Some customers would happily redesign your entire business model and try to convince you that you're doing it wrong, even when metrics from thousands of customers completely contradict their beliefs—but they don't see that.
That's why you, as the business owners, should always be the ones making the final assessment.
The main goal is to make money. If you aren't making money and are constantly just tweaking things, you'll go bankrupt sooner or later.
6. Don't wait for your money to come back—use high-ticket sales
This was actually Štěpán's idea, and it works brilliantly. The principle is simple: use a webinar to sell a larger package—in our case, it was a larger marketing academy package—and include a free subscription to our service as a bonus. This allows us to generate high-ticket sales and completely recover the money invested in advertising.
It's an important principle because if you wait for your advertising spend to be recouped solely through monthly subscriptions costing a few hundred crowns, it can take months. This way, you recover your investment almost immediately and have additional cash flow for scaling.
7. A better hiring process and thorough candidate screening
Unfortunately, this is an area in which I've had many bad experiences. Sometimes I took people at their word—and that didn't pay off at all.
One specific case: a programmer told us he was good and showed us a codebase—we didn't even know whether it was actually his, but it looked good. The result? After a month or two, he stopped communicating completely. He was supposed to finish a task by a deadline he had set himself, but he missed it entirely. We called him, messaged him, and I even sent him a message through the official data mailbox system—just so he would let us know what was happening. Eventually, he admitted that he was also working for someone else and probably couldn't manage everything timewise. And the code he ultimately delivered was one enormous disaster. It cost us roughly 150,000 Czech crowns in wages, and we got absolutely nothing. We could have completed the feature much faster and more cheaply.
Another case involved a salesperson who cold-called customers for us. On the side, he began selling completely unrelated services—some website development packages—and used the marketing database we had provided to him in good faith. We found out entirely by chance from one of our customers, whom he had called and offered something for 2,000 euros that had nothing to do with us. On top of that, he used our name and pretended that we were providing the service.
The lesson is clear: screen people thoroughly. Conduct multiple rounds of interviews if necessary, assign test tasks, and verify that the person can genuinely do what they claim. Make sure you have NDA agreements in place and record sales calls—always trust, but verify. I like to trust people, but sometimes they disappoint me.
And most importantly—if you see that someone isn't working out, replace them quickly. The longer you wait, the more it costs you—undelivered work, wasted time, missed opportunities, or, heaven forbid, problems within the team.
8. More aggressive marketing and quickly turning off underperforming creatives
Alongside the strict cost tracking I mentioned at the beginning, I would pursue more aggressive marketing. Most importantly, if something isn't working—such as ad creatives—turn it off immediately and don't burn through your budget.
There were times when we went on vacation or simply didn't pay much attention for a few days—and in the meantime, some creatives stopped working. On Facebook and Instagram, you have to rotate them frequently because audiences quickly become “blind" to seeing the same ads over and over. Štěpán manages this very aggressively today—he turns ads off, launches new ones, and if something doesn't convert or the numbers are poor, it gets removed immediately.
Of course, the reverse also applies—if something works and makes sense, scale it and increase the budget. But that's a topic for a separate article.
9. Greater emphasis on SEO and optimization for language models
I'd like to focus more on SEO—ideally, I'd have a dedicated team or at least AI agents set up to handle it continuously. I don't really have the capacity for it, but I try to work on it regularly, and even what I've done so far brings us quite a lot of customers. Thanks to SEO, we have fairly good visibility.
Then there's a relatively new area—SEO for language models, known as LLM SEO/AEO, and so on. It involves optimizing content so that your product or brand appears in the responses of AI chatbots such as ChatGPT, Claude, Gemini, and others. This trend is gaining momentum because more and more people are searching for information directly through AI instead of using traditional Google searches. I definitely want to devote more time to it than I have so far.
10. Better time management and quarterly planning
Another important thing is planning features, tasks, and work overall—ideally by weeks, quarters, or even six months in advance, including for the entire team. Without that, everyone does whatever they want and chaos ensues.
Of course, software development isn't always straightforward. I know from personal experience that I've set a deadline and then missed it by a week because I hadn't fully considered certain dependencies or thought of something else along the way that needed to be completed or reworked. With the arrival of AI development tools, it's somewhat better today, but it's still not 100% predictable.
Nevertheless, it's good to have a plan. The people who work for you need to know where things are headed and what will be done. It also helps me personally—when I break something down into smaller steps, it's much easier to get started. When I have one huge task in front of me, it can even be a little demotivating. But when I break it into small pieces, decide where to start, and get into it, I can continue without any problems. I think a lot of people struggle with this.
11. More rigorous card checks and preventing trial abuse
If you offer a trial—whether it's five days, seven days, or any other type—make sure it's secured so that people can't repeatedly abuse it.
Stripe allows you to store a so-called card fingerprint, which prevents the same card from being used multiple times. We have it set up so that if someone uses the same card for another trial, the new account is automatically blocked and the trial can no longer be activated.
We used to be very lenient about this—and we discovered that one user, for example, had dozens or even a hundred accounts with similar email addresses, used the same card for all of them, and repeatedly took advantage of our trial. In effect, he was using the service for free indefinitely. Once we identified the pattern and secured the system, the problem disappeared.
Trial abuse is a fairly common problem in the SaaS world. According to industry estimates, as many as 10–15% of free-trial registrations may come from users who register repeatedly just to avoid paying. That's why it's important to have mechanisms in place from the outset to prevent this behavior.
12. Don't operate a business in the Czech Republic—use a Merchant of Record abroad
This may be a controversial topic, but my view is clear: I wouldn't recommend operating a new business in the Czech Republic, especially if you have a digital product with a global reach.
First, everything is constantly changing—and by that I mean laws and procedures. Second, there's the unbelievable bureaucracy imposed by the authorities. Constant audits, constant inquiries from the tax administrator. And that's despite the fact that we have everything in order and our accountant handles everything diligently. But compared with the United States, for example, the bureaucracy is absolutely insane and disproportionate.
VAT legislation is complex, the tax system is constantly changing, and tax rates change—it's all brutally complicated. Even my father, who runs a construction business, says that dealing with the authorities is becoming worse and more complicated year after year.
That's why I recommend using a so-called Merchant of Record (MoR) for new companies. Essentially, we'll now process everything through a US entity, and the payment gateway will handle all tax obligations on our behalf—VAT, digital taxes, and so on. I discussed this with our lawyer, who said we should be careful about digital tax legislation—it could be argued that the place of supply is in the Czech Republic or the EU, which would require VAT to be paid in the relevant country. A Merchant of Record handles this from a legislative standpoint.
A Merchant of Record will also solve our tax issues worldwide—not only in the United States, but also in Australia, Canada, and other countries.
And there's another benefit of doing business in the United States: we have a credit card with unlimited 1.5% cashback. When we spend millions on advertising every month, getting 1.5% back in actual cash is not an insignificant amount. In the Czech Republic, I was offered a credit card with a “CZK 500 discount at Datart"—which is honestly completely useless. The cashback system works brilliantly in the United States, and you can reinvest that money in the product, pay for flights, cover a car lease—anything.
13. Pay yourself first
This is our golden rule: pay yourself first. Pay yourself before anything else, otherwise you, as the founders, will gradually lose motivation.
Everything works better when you can see that the company is profitable and can afford to pay yourself. That's what you live on and what drives you forward. I often see entrepreneurs who are just starting out and pay themselves almost no salary for two or three years—and then it affects their mental health and simply everything else. Their savings disappear, they have no money, they can't function, and they can't even pay for basic necessities.
Yes, the purpose of business is to create added value for customers, identify their pain point, and solve it. But the other side of the coin is that you're also doing it for profit—and ideally for more than you would earn as an employee. If you aren't getting any money out of it and are burning it on costs, employees, and everything else, then the point of running a business disappears. You'll either go bankrupt, burn out, or lose interest—and that will affect the product and everything else.
We've followed the pay yourself first strategy from the beginning, and it's absolutely fantastic. Of course, it requires you to keep the entire business profitable so that it works over the long term. But if you don't make money from your business, you'll lose motivation and become trapped in a vicious cycle.
14. Don't waste time on marginal matters—delegate
The final point, but definitely not the least important. I often dealt with marginal matters that I could have delegated—to a virtual assistant, support staff, other programmers, or simply someone hired on Fiverr for a small amount of money.
Many times, I've discovered that I could solve a problem for a few hundred crowns on Fiverr that would otherwise have taken me an extremely long time. So delegate marginal matters whenever possible and focus only on the important ones—the things that move both you and the business forward.
I know that, as an entrepreneur starting out, you have to handle everything—accounting, marketing, sales, and a million other things. But gradually try to delegate the most pressing areas.
At the beginning, Štěpán and I discussed what each of us disliked doing, and we immediately avoided those areas. I told him that I didn't like customer support or making calls to customers. So almost all customer support was taken off my plate right from the start. I only handle what remains and what nobody else can resolve—and even that, I gradually try to hand over to colleagues.
By no means am I saying that customers are marginal to me—they definitely aren't, and they must always receive a response and be satisfied. If something doesn't work, they want a refund, or they can't make a payment, then of course it has to be resolved. But with thousands of customers, I don't think I need to handle it all personally. My responsibility is to supervise the people who handle these matters and make sure they do so conscientiously and well.
When you have a larger company and more people, you can't do everything. You can't program, handle marketing, and manage support—you can't be a jack-of-all-trades, because then you do everything and ultimately accomplish nothing.
A specific example: Štěpán used to prepare various reports manually—he extracted data from accounts and calculated how much we were spending on what. It took him a great deal of time. Now a virtual assistant does it for a few hundred crowns, saving us hours that we can devote to things that genuinely move us forward.
And when I listened to recordings of our salespeople talking to customers—how perfectly they did it, how beautifully they communicated with them, and how obvious it was that they enjoyed it—I knew I could never do that job as well as they did. Because I wouldn't enjoy it nearly as much. I know that I enjoy programming, long-term planning, and developing the company. And I'm good at those things. As Štěpán and I agreed—I would code, work on the product, and do the things I enjoy, while he would do the same with marketing. It's great; we complement each other perfectly. We delegate everything else.
So in business, choose the thing you're good at and enjoy doing. And whatever you don't enjoy, will fail at, or will neglect—delegate it. I know that even though money can sometimes be an issue, you can use virtual assistants and similar services. They can do many things cheaply and save you time for what truly matters.
What should I say in conclusion?
These are all the things I wanted to mention and want to avoid in our next business. I wrote them down partly as a reminder to myself and partly to share them here on the blog, because they may help you someday—whether you're building SaaS software or any other type of business.
If I had to sum it up in one sentence: track your costs, secure everything from day one, test your code, move quickly, pay yourself first, and delegate whatever you can. And most importantly—don't repeat mistakes that someone else has already made.