EP177 · Economy · first published 2023-02-26
Databases and MariaDB | Patrik Backman | Negotiator 177
Open Ocean's Patrik Backman explains how MySQL became the internet's bookshelf and why MariaDB had to be founded once Oracle acquired Sun — and with it the world's most popular open database. The episode works through MariaDB's SPAC listing, whose pricing window closed before the company reached the market, and sets out the metrics a growth company is valued on: ARR multiple, Rule of 40, Net Revenue Retention and churn. Portfolio cases include Truecaller, Supermetrics, Leadoo, Nosto, AppGyver and MindsDB, with a side trip into the ChatGPT enthusiasm of early 2023. The host discloses that he is himself a small investor in Leadoo. Published 26 February 2023.
Databases and MariaDB | Patrik Backman
Summary: In episode 177 Sami Miettinen interviews Patrik Backman of Open Ocean about open source databases, how SaaS companies are valued, and early-stage investing. Backman was at MySQL from 2000 to 2008 and now sits on the fund side behind MariaDB. Published 26 February 2023.
Disclosure. The host, Sami Miettinen, states in the episode that he is himself a small investor in Leadoo, which is discussed as one of the portfolio cases. He is also a partner at Translink Corporate Finance, and Translink and Open Ocean share an address at Mikonkatu 1B in Helsinki.
A database is a bookshelf
The single best explanation in the episode is Backman’s analogy, and it is worth taking first: a database is a bookshelf. Data is organised into books, the books go on the shelf, and the shelf gives the data back quickly. When a web page needs to display something, it is taken off that shelf — and new data is written back onto it.
This explains what MySQL became. In the late 1990s Michael “Monty” Widenius and his co-founders decided to put the database on the internet for free as open source — inspired, Backman says, by Linux. The goal was a database built for the internet: as fast, reliable, efficient and free as possible.
The timing landed. The internet was growing, and the internet is full of text and data that has to be shoved somewhere. In Backman’s words, all the early Amazons and Facebooks and YouTubes put their data into MySQL and pulled it back out in order to show anything on their pages. Millions of users by the early 2000s, because it was the fastest, the easiest — and free by default.
MySQL from 50 people to a billion-euro exit
Backman joined MySQL after university in the early 2000s. The company had been founded in 1995, Finnish-Swedish, and had 50 people at the time. When he left in 2008 there were 600, revenue just under 100 million and an exit of roughly a billion.
A piece of cultural history runs alongside: MySQL’s internal discussion happened on open channels worldwide — on IRC — and managers were criticised there from the ground up. Backman draws the line straight to Slack, which Miettinen describes as workplace democracy: you can communicate regardless of title, where email is rigid in exactly that respect.
Why MariaDB had to exist
The chain of events is short and its logic matters.
MySQL was sold in 2008 to Sun Microsystems. Sun did badly, and Oracle bought Sun. Backman’s reading of Larry Ellison’s motive is direct: Oracle wanted Sun precisely because MySQL was so popular as a database worldwide. Oracle was — and still is — the largest database company in the world, making tens of billions in sales from large companies running important things on its database.
A counter-reaction formed in the market. Customers, users and former employees took the view that a commercial and at times hard-edged operator could not be allowed to take an independent, free project into its own hands. That was the opening, and MariaDB was founded.
The mechanics are open source’s basic logic: take the last stable open MySQL version and improve and productise it. The versions have since diverged. Backman’s argument for independence is practical rather than ideological: Oracle has a reason not to develop MySQL into a competitor for its own large systems, so it is constrained in where it can take the platform. MariaDB can take its product wherever the need goes.
He is honest about the relative scale: MySQL is still more popular. MariaDB is the alternative — around 20 million users, one of the ten largest databases in the world.
A SPAC whose pricing window closed
This is the most interesting financial mechanism in the episode, and Miettinen raises it head-on: an HS Vision analysis had left the impression that the company was more or less worthless after listing.
Backman does not comment on the company’s internal affairs — he is no longer on the board, and a listed company has to handle its own disclosure — but he explains the mechanism.
The SPAC was priced a year before the listing, when markets were very high. A large number of investors had committed money. By the time the company reached the market late in the year, multiples had come down. The investors concluded, entirely logically, that the same multiple of 14 was not going to be available, and took their money out.
Backman’s conclusion is that nobody did anything wrong:
Nobody broke anything on purpose here, it just went that way.
Miettinen puts the same point structurally — it is in the nature of a SPAC that the pricing is fixed in advance, and if the pricing window breaks, the mechanism simply does not work. There was no dilution, because no shares were issued, but the hoped-for funding did not arrive either.
The figures, as Backman gives them: the valuation a year earlier was around 700 million, a multiple of 14 on roughly 50 million of revenue. At the time of recording the market valuation was 250 million, or five times. He says it plainly, in an owner’s voice:
That is of course a disappointment to us and in our view far too low.
His reasoning is that the share price does not account for the 20 million user base and what that brings for the future. Miettinen’s response is healthy scepticism — “this kind of eyeball valuation worries me” — to which Backman replies that this was a waypoint.
What a buyer looks at: four metrics
Miettinen runs through the metrics Translink uses in valuations when selling companies. This is the most directly usable part of the episode if your own company is for sale.
The ARR multiple. Annual recurring revenue. A complex valuation compressed into one multiple. Miettinen stresses to listeners that ARR is vastly more valuable than revenue: once the customer relationship is won it repeats into the future, whereas an ordinary company has to make the sale again every year.
Rule of 40. Profitability plus growth. If the sum of those two parameters does not clear forty, a large part of the buyers walk away. It works as a sanity check on company quality: you can deliver growth, profitability, or a combination.
Net Revenue Retention and churn. Customer retention and its cruder counterpart, customer attrition. As a rule of thumb, five per cent churn is already very good.
Gross margin. Miettinen says he pushes on this often, because small companies have no CFO and no great spreadsheet discipline, so profitability gets calculated oddly — database costs, for instance, get left out.
Backman confirms the list for a mature SaaS company doing five million plus. But he adds the investor’s view, which is materially different:
As a VC investor we have a somewhat different perspective. If it goes really well, it ought to grow at least a hundredfold.
And here comes the sharpest observation in the episode about growth investing. The first tenfold can usually be argued from the metrics within a few years. The hardest part is the second tenfold — one million to ten is a different discipline from ten to a hundred. That forces a difficult judgement about what the team looks like, whether it can lead the company in that direction, and whether there is the ambition to hire people beyond the core team who can take it from ten to a hundred.
Open Ocean’s investment focus
Open Ocean comes in early: typically when there are about five people and a first version of the product has reached the market. The focus is all of Europe, and the targets are companies building data software — something innovative and intelligent around data inside the product, but productised so that it is easy to adopt and therefore efficient to scale.
Miettinen uses Slack as his example in SaaS training: Accel put in 1.5 million and Salesforce bought the company for 25.4 billion dollars. Backman takes it as a fair picture of the best case: the product can be very simple at the start — “people messaging each other” — but if it is built into a broadly useful product for businesses, it can be commercialised widely.
A note on the figures given in the episode. The publicly reported enterprise value of Salesforce’s Slack acquisition was about 27.7 billion dollars, and Accel’s first investment came in a Series A of roughly five million dollars. The difference is the ordinary imprecision of live conversation and does not change the argument — Accel’s total return on Slack has publicly been reported in the billions.
Truecaller: why it scaled in India
Of the portfolio stories this is the best, because it is market-specific rather than generic.
Truecaller shows you who is calling. When Open Ocean met the company more than ten years earlier it had a million users and a million numbers. At the point of investment, three million. At the time of recording, hundreds of millions — revenue approaching 200 million and a listed value of a couple of billion.
Miettinen asks why this did not become Fonecta. Backman’s answer is that the first version of Fonecta Caller was in fact licensed from Truecaller, and Fonecta held exclusivity in Finland — Truecaller was free to operate everywhere else.
And then the actual reason:
The reason it was so popular in the Middle East and India was that there the recipient pays for the call.
When answering costs money, you do not want to answer an unknown number — and in those markets there was no other way to know who was calling. The product’s value was not in the technology but in a billing model that made it necessary.
The rest is platform-economy arc: five years non-commercial, users started building profiles, it was used even for dating, and it became a broader identity platform. Commercialisation began only about five years before the recording.
Supermetrics: a t-shirt and profitability from day one
Supermetrics is a data integration tool. In practice: if you do digital marketing on Google, Facebook and LinkedIn and want to see how a campaign went, it brings the analytics data into a simple place — Google Sheets or Excel. Backman notes that a business user does not want it in a database; a database is too complex, a spreadsheet is not.
The origin story is the funniest thing in the episode. A decade or so earlier Google itself had no tool for getting Google Analytics data into Google Sheets — two products of the same company, and it could not be done. Mikael was at an event where a Google representative said that whoever solves this gets a t-shirt from me.
It took a few months, and then Mikael was walking around town in that t-shirt.
The company was profitable from day one, and the founder’s own reasoning, per Backman, was simple: a company’s job is to make a profit. The most telling detail is that Open Ocean’s investment is still sitting in the company’s account — the money was never needed. The investor came in as an owner and a partner, not as a financier.
The scaling shows in two directions: there were about five data sources at the start and roughly a hundred now; the destination was Google Sheets, and now also Google Data Studio and Snowflake. Recurring revenue around 60 million, headcount just over 300.
Fresh news at the time of recording was that the company had two chief executives: Anssi Rusi had joined at the start of the year, with Mikael focusing on product and technology and Anssi on the commercial side. Backman’s justification is quantitative — when a company grows to this scale and keeps growing worldwide, there is simply not enough time in one person’s day.
Leadoo, teletext, and a call back within five minutes
Disclosure. Miettinen states here that he is himself a small investor in Leadoo. He says he put in ten thousand or so and wishes the company well for slightly selfish reasons too.
Leadoo does website conversion: an interactive bot you can tell what you are looking for by clicking, instead of having to type into a chat window. It steers the user into revealing more about themselves, which helps win the sale. At the time of recording, roughly six million in recurring revenue and international expansion under way into Sweden and the UK.
Backman found the company on teletext. More than ten years earlier he had set teletext as the start page on the first mobile devices that had internet — for the NHL scores — and it had stayed there. In 2019 the front page said that intelligent bots had now arrived on websites.
Then comes the part that makes the story worth telling. Backman emailed founder Mikael Koski, who called back within five minutes: “Hello, hello Patrik. Can we meet tomorrow?” Within twenty-four hours Koski had built a presentation about the company which Backman heard afterwards had not existed at all. The speed of reaction was the signal.
Miettinen tells his own version of the same idea and calls it his personal commercial DD: he looked up who Leadoo’s largest customer was, called them — and it turned out to be the father of his daughter’s good friend — and asked whether this was genuinely any good. The answer came back after a few enquiries: yes, this does seem to work.
Nosto, AppGyver and MindsDB
Nosto. Ten years of working with the founders. Five people at the start and about 100,000 in annual recurring revenue; 30 million at the time of recording. The product is one technical line of code on a merchant’s site, after which it tracks everything visitors do and dynamically adapts content, selection and offers to what it understands about the user. Backman’s example: if you are looking at car tyres, it can offer you a pump. The same logic as Amazon’s “people who bought this also looked at this”, but integrated directly into the storefront.
AppGyver. Sold a couple of years before the episode to SAP, very nearly Europe’s largest software company. The product lets anyone build a mobile application with intelligence and data from the company’s systems built in — an enterprise app in hours or days. Backman singles out founders Marko Lehtimäki and Henri Vähäkainu as an example of vision carrying: “this is how it has to be done.” The conversation widens to how much of this capability sits in Finland — Miettinen mentions Unity, and Backman notes that Applifier is part of Unity.
MindsDB. Machine learning inside the database. Backman’s explanation is concrete: if the database has tables of weather — sunshine, temperature — and elsewhere ice cream sales, which usually correlate, you can build predictive models of what next year’s ice cream sales might look like. The product is not for consumers but for database developers who want to bring predictability into their own code. The day before the recording it had been announced that Benchmark — one of the leading US investors — had come in and led the A round. MindsDB had already integrated ChatGPT.
ChatGPT in February 2023
The episode was recorded in the first months of ChatGPT enthusiasm, and it shows.
Backman’s own demo was at the Christmas table: he asked, live, for a thank-you speech in verse, with rhymes. Six stanzas came out, and it could then be extended iteratively — add the succulent Christmas ham, add these three cats under the table, add this good white wine. The speech could be delivered, quite convincingly, at the end.
But he also tells the failure, and it is the best single anecdote in the episode. At the same Christmas table they asked why Messi had not won a European championship:
The chat replied that unfortunately Argentina had not qualified for the final tournament.
Backman’s business idea for listeners is concrete, and should be read in its 2023 context: anyone who wants to be an entrepreneur and can code a little should found a professional services and software shop that uses ChatGPT to write the code. Traditional coding projects can be done at least twice as fast, and you can then ask three-quarters of the price and go and compete.
B2B or B2C
A small but useful correction near the end. Miettinen says he used B2C examples for a long time until he noticed people treating them as the business model to aim for. In reality SaaS companies operate mostly in B2B, and his experience is that B2B SaaS is considerably easier to sell than B2C SaaS. Backman’s explanation: in B2C the competition turns fierce quickly, and then you have to spend outrageously on marketing and content. Netflix was fairly alone at the start; now there are dozens of competitors. WordDive is named as the Finnish B2C example.
What stays with you
Three things.
Open source is a competitive position, not a principle. MariaDB’s reason to exist is not ideological but structural: Oracle has a commercial reason not to develop MySQL into a competitor for its own systems. An independent fork sells what the owner cannot promise.
A SPAC’s pricing window is a mechanism, not a scandal. The price is fixed a year before listing. If multiples fall in between, the rational response from investors is to redeem, and the company does not get the funding it sought. The outcome looks like a failure, and in funding terms it is — but the cause is timing, not operations.
The second tenfold is a different discipline from the first. One million to ten can be reached on the metrics. Ten to a hundred is reached only if the team has both the capacity and the ambition to hire people better than itself — and this, Backman says, is the hardest judgement in venture capital.