Scaling a startup is never linear. Ektora began as a weekend experiment on a Raspberry Pi and, within two years, grew into a SaaS platform serving 1,790 paying customers, generating $321,194 in revenue, in one of the most competitive niches in tech: social media automation.
Two founders. Ten thousand euros of our own money. No outside capital, ever.
This is what it taught us about technology, markets and ethics, and why we ultimately chose to shut it down.
MVP PHASE · micro-entreprise · one-off licence, avg order $116 | ||||
| Dec 2018 | $5,000 (mo $5,000 · 55 cust.) | |||
| Jan 2019 | $22,000 (mo $17,000 · 154 cust.) | |||
| Feb 2019 | $36,500 (mo $14,500 · 121 cust.) | |||
| Mar 2019 | $55,800 (mo $19,300 · 175 cust.) | |||
| Apr 2019 | $72,800 (mo $17,000 · 142 cust.) | |||
| May 2019 | $106,800 (mo $34,000 · 305 cust.) | |||
| Jun 2019 | $130,800 (mo $24,000 · 203 cust.) | |||
| Jul 2019 | $147,800 (mo $17,000 · 138 cust.) | |||
| Aug 2019 | $159,800 (mo $12,000 · 96 cust.) | |||
| Sep 2019 | $169,400 (mo $9,600 · 76 cust.) | |||
| Oct 2019 | $176,600 (mo $7,200 · 59 cust.) | |||
| Nov 2019 | $181,400 (mo $4,800 · 43 cust.) | |||
| Dec 2019 | $186,200 (mo $4,800 · 39 cust.) | |||
| Jan 2020 | $191,000 (mo $4,800 · 35 cust.) | |||
| Feb 2020 | $193,400 (mo $2,400 · 22 cust.) | |||
| Mar 2020 | $205,300 (mo $11,900 · 110 cust.) | |||
| Apr 2020 | $206,500 (mo $1,200 · 10 cust.) | |||
| May 2020 | $207,460 (mo $960 · 7 cust.) | |||
SAAS PHASE · Ektora SAS · subscription at $99/mo | ||||
| Jun 2020 | $213,180 (MRR $5,720 · 58 subs) | |||
| Jul 2020 | $222,280 (MRR $9,100 · 92 subs) | |||
| Aug 2020 | $234,760 (MRR $12,480 · 126 subs) | |||
| Sep 2020 | $251,620 (MRR $16,860 · 170 subs) | |||
| Oct 2020 | $272,860 (MRR $21,240 · 215 subs) | |||
| Nov 2020 | $298,050 (MRR $25,190 · 254 subs) | |||
| Dec 2020 | $321,194 (MRR $23,144 · 234 subs) | |||
| Legal form | Société par Actions Simplifiée (SAS), France |
| Registered | 21 April 2020, capital €10,000. Closed since. |
| Before that | micro-entreprise, 2018 to Apr 2020: $207,460 and 1,790 customers |
| The $321,194 | gross, in USD, across both structures. Neither set of French accounts shows that figure on its own. |
| Team | two founders, fully remote, no employees |
| Funding | none. Bootstrapped from first sale to last. |
| Waiting list | 14,000 signups at SaaS launch |
| Unit cost | under 30MB of RAM per running agent, on commodity VPS |
| Revenue / head | $160,597 |
| MRR growth | 34.5% month over month, Jun to Nov 2020 |
Ektora's mission was simple: help professionals grow their Instagram audiences efficiently, securely, and at scale. Instead of paying for ads, users could gain visibility with a single click, powered by automation algorithms that replicated authentic engagement. In a world where paid social advertising was becoming prohibitively expensive, Ektora offered a cost-effective edge.
We weren't the only ones chasing this market. Competitors like Jarvee, Ingramer, Nitreo, and Kenji.ai had been around for years. But most played the same game: superficial automation, easy for Instagram to detect.
Our differentiator was technical depth. We built Ektora to mimic human behavior at the signal level, making our automation far harder to flag. Where others cut corners, we engineered authenticity.
Our CTO spent months on how the mobile app actually talked to its servers. What that work found was the channel the app used to stream constant telemetry back about the device and its environment. Nobody else in the open market was reading it.
Instead of bots that just did things, we designed bots that lived.
Every step generated telemetry identical to a real user. To Instagram, our agents were indistinguishable from humans. No two sessions looked alike, and each one was consistent with the environment the user claimed to be in.
Behind the scenes, we focused on simplicity and resilience. The frontend ran on Intercooler.js, now HTMX, and talked to the backend through a message queue, so the agents were never exposed to the network. This kept costs low while letting us scale fast, thousands of users on inexpensive VPS fleets.
Security.
When a subscription ended, the container was destroyed and the user's data went with it. No retention was the default, not a setting. We held as little about our own customers as the design allowed, which is the one habit from Ektora we kept without changing a thing.
When Instagram couldn't tell our AI agents from their users, we knew we had built something that worked. We are less sure now that it was something we should have built.
Two things nearly killed the business long before we chose to end it, and neither of them was competition.
Payment processors. Stripe and PayPal did not want our category. Being technically excellent in a niche the payment layer refuses to serve means your revenue can be switched off by a risk team that has never read your code. Distribution risk and payment risk are the same risk.
Platform dependency. Every Instagram release could invalidate months of reverse engineering. Each new action took days of traffic analysis to template. We were fast enough to keep up, but we were renting our product from a company that would have preferred we did not exist.
Ektora worked. That was never the problem.
The truth is that Ektora was not a venture we actively chose. It was a project that happened to work well in a competitive, constantly shifting market. We could adapt to Instagram's algorithms and solve the technical problems as they came. What we could not do indefinitely was stay comfortable operating in a grey area.
We were tired of playing pirates. Tired of the fine line between innovation and compliance, and of building something whose whole defence was that nobody could see it. As entrepreneurs who value integrity, we decided it was time to move on and work on something where transparency is the point, not the risk.
So we turned it off while it was still growing. That is the part people find strange, and it is the part we would repeat.
The same two founders now build THINKPOL, an intelligence platform for law enforcement and national security, on a 30 billion post archive of the grey web. Same skills, right side of the line.
— Mat Gillard & Lucas Dominguez