Inbenta Hit $14M in Revenue on a $160K Monthly Burn — Here's the Math
Asked for Inbenta's gross revenue churn in February 2019, Jordi Torras said 40% — a number that couldn't possibly be true. Untangling it exposed the engine behind a $14M business losing only $160K a month.
On this page
Halfway through a February 2019 taping, Nathan asked Jordi Torras the question he asks every guest — what is your gross revenue churn? — and Torras said “around forty percent.” Nathan refused it on the spot: “God, I was gonna say there’s no way it was forty percent. You wouldn’t have 26 million dollars raised.” He was right, and the misunderstanding turned out to be the most useful minute on the tape. The 40% wasn’t churn. It was Inbenta’s growth rate, and pulling it apart exposed the engine under a $14 million company that was only losing $160,000 a month.
Torras had answered a different question. Inbenta closed 2017 at about $10 million and ended December 2018 at $14 million — 40% growth — and when Nathan asked him to split it, he said half came from new accounts and half from customers Inbenta already had. Pressed for gross revenue churn on its own, the revenue that walks out of the existing book before any upsell is counted, Torras said 5%. Expansion on that same book ran 25%. The arithmetic closes exactly: 20 points of new logos, plus 25 points of expansion less 5 points of churn, is the 40% he had blurted out.
The company in one line. Half of the $4 million Inbenta added in 2018 came from customers it already had, on a meter that bills for usage rather than seats — which is why a $14 million business was losing $2 million a year instead of $20 million.
Where the extra $4 million came from
Net revenue retention of 120% is a claim; Nathan made him name the pricing axes it runs on. Torras gave two, and both are structural rather than salesmanship.
Inbenta charged in tiers of sessions — a customer’s customer asking a question starts one. When a client opened a new country, the bill moved without anyone selling anything. “Some of our customers move international,” Torras said, “that creates more sessions, more usage of our technology.”
Accounts that started with the AI answering questions on a public website could point the same engine at their call-centre agents. A second use case, sold to a buyer already live in production.
One number on the tape never reconciled. Nathan tried to rebuild the revenue from the two figures Torras had already given — 300 customers at a $30,000 average annual contract value — and got $750,000 a month, about $9 million a year. Torras corrected him to $14 million and moved on. By the show’s own closing summary the average customer was paying “three, four grand a month,” nearer $40,000 a year. The likeliest reading is the one Torras gave when he first quoted the $30,000: “that’s where we are now” — the price of a new deal, not the average of a book that expansion had already pushed a third higher.
Eighty cents to buy a dollar
The other half of the story is what new revenue cost. Torras put Inbenta’s acquisition efficiency at 0.8, fully loaded: “marketing, outbound sales, inbound sales, SDRs, account executives, demos, the whole thing.” On the $30,000 average deal that is a willingness to spend roughly $25,000 to sign it, and about ten months of that contract’s revenue to earn the cost back — before gross margin, which stretches it further. Against the CAC benchmarks founders disclose on the show, sub-year payback at enterprise contract sizes sits at the comfortable end.
$0.80fully loaded cost of adding one dollar of new ACV, February 2019
Let’s say adding one dollar of new ACV is costing us 80 cents.
Jordi Torras, founder & CEO, Inbenta
The $160,000 in the episode title
Which brings us to the number the episode is named for. The title reads “$160k New Burn Monthly”; on the tape it is net burn, and it is not a bookings figure at all. Inbenta lost about $2 million in 2018, Torras planned to lose roughly the same in 2019, and Nathan divided by twelve out loud to land on $160,000 a month. Torras had a comment about the arithmetic — “you are good at dividing by 12” — and none about the number.
For a company with 165 employees, 300 enterprise customers and $14 million in revenue, that is a remarkably thin loss, and it follows directly from the two mechanics above: growth that mostly arrives on the existing book, bought at 80 cents on the dollar. Torras expected break-even in the second quarter of 2020, which Nathan framed as a year and two months away. Torras agreed.
The funding ladder
Nine years before the first term sheet
The frugality isn’t a policy Inbenta adopted; it’s the shape the company grew into. Torras started it in 2005, but not as a software business — “we started more like a consulting business, a professional services company,” and that business paid for the product. “For many years the company was bootstrapped, particularly during the time that we were a consulting business, and we were self-financed. That consulting business actually allowed us to finance the very first version of the software.” The first beta shipped in 2010. Torras moved to California in 2012. Outside money didn’t arrive until 2014.
- 2005 · Founded Barcelona, as a consulting and professional-services business that self-financed the product.
- 2010 · First beta The natural-language engine ships, built by a team of linguists, software architects and customer-care specialists.
- 2012 · California Torras relocates to the US to grow the company.
- Apr 2014 · Series A $2M — the first outside capital, nine years after founding.
- Sep 2016 · Series B $12.6M, the largest round GetLatka records.
- Jun 2017 · Venture round $2.2M.
- Aug 2018 · Series B $6.66M, recorded under the same label as 2016.
That last round came in through a convertible note, for a reason worth repeating because it has nothing to do with valuation games. One investor — Scale Capital, out of Chile, partly funded by a government organisation — had to invest in two phases because its approval process ran longer than the round did. It was, as Torras put it, an approval-process and regulation matter rather than a strategic one; the note converted a few months later. He also mentioned, without much ceremony, that one of Inbenta’s investors is also one of its customers, a communications company kept at arm’s length as two separate legal entities.
The totals don’t quite agree. Torras said “26 million” raised to date. GetLatka’s four recorded rounds add to $23.5 million. Neither the tape nor the dataset resolves the $2.5 million gap.
What the years since look like
The dataset captured Inbenta’s run rate at $14.2 million in February 2019, matching what Torras said on the tape. GetLatka then recorded $32.9 million in December 2023 and an estimated $53.3 million in October 2024 — roughly 3.8 times the taping figure.
Headcount is the more interesting series. The 165 people Torras described in February 2019 had fallen to 144 by December 2019 — the company got smaller on its way to the break-even it was aiming at — and didn’t pass the 2019 mark again until 2021. It reached 194 by September 2023 and the dataset still had it at 194 in October 2024. Revenue nearly quadrupled; the team grew about 18%. That is what a usage meter and a 120% net retention rate look like six years later, and it is the same argument Torras was making in 2019, just with more evidence behind it. The current figures sit on Inbenta’s GetLatka profile.
Torras was 50 at the taping, married with a son and two daughters, sleeping about seven hours a night, reading a book called Whale Hunters, and naming his own product as his favourite tool for building the company. Asked what he wished his 20-year-old self had known, the founder who moved a Barcelona company to California didn’t reach for anything about strategy or capital: “I should have learned English better when I was studying.”
SourcesNathan Latka’s February 2019 interview with Jordi Torras, CEO of Inbenta; GetLatka dataset rows for revenue and headcount through October 2024, where the October 2024 revenue figure is flagged an estimate; GetLatka funding records, 2014–2018.
