Customer.io's Founder Wrote the Shutdown Letter Before the Database Came Back
Colin Nederkoorn spent six hours of a twelve-hour outage drafting the letter that would end his company. It is one of nine mistakes he catalogues from Customer.io's run to $10 million ARR.
On this page
Six hours into a twelve-hour outage, with millions in ARR on the line and no path to recovery, Colin Nederkoorn started drafting the letter telling Customer.io’s customers the company was finished.
I’m sitting there at about hour six of twelve, thinking about the letter I’m going to write to our customers… we’re probably going to have to shut down the business.
Colin Nederkoorn, founder and CEO, Customer.io
The cause was not a hack or a bad deploy. It was a database the company had chosen years earlier because it was technically excellent, whose vendor had since been bought by Apple and had stopped renewing support contracts.
That story is the centre of a talk Nederkoorn gave about the stretch from $1 million to $10 million in ARR, and it belongs to a category he names outright: things that don’t make the beer taste better.
Things that don’t make the beer taste better
The phrase is Jim Koch’s, the founder of the company behind Sam Adams. Nederkoorn’s reading of it is an allocation rule: spend as little effort as possible on anything that does not improve the thing customers actually value, and as much as possible on the thing that does.
Three of Customer.io’s early decisions failed that test, and all three were infrastructure choices made for reasons that sounded like engineering discipline at the time.
Beefy servers in a data centre, chosen as more cost-effective, partly so European customers could have data outside the US. Then the fibre got cut at the Monsignor Langelois Bridge and the servers were unreachable. “I don’t even know if it got cut above the bridge or below the bridge.”
Amazing technology, one vendor, acquired by Apple. When the cluster fell over there was no support to call. Migrating off it meant rewriting the entire backend.
The gap between version 0.x and 1.0 was large enough to force an application rewrite, and every new engineer needed training in it. “Customers didn’t care about that. They just wanted to use a CRUD app.”
The database recovery came from a cold LinkedIn outreach that reached someone experienced from the acquired team. What replaced it inverted the architecture: every customer now gets their own MySQL database. Nederkoorn draws the lesson that a distributed store does not remove single points of failure so much as consolidate them. “Now every customer has a single point of failure, with backups of course. But you can’t take out our entire customer database.”
The AWS layer sitting in front of the Quebec servers is the only reason the fibre cut was survivable — customer data queued there rather than being lost. The eventual move was to Google Cloud, and the conclusion was blunt: “Our customers don’t care where we host. They care that it’s super reliable.”
The criteria that came before the product
Customer.io was picked, not stumbled into. Before writing anything, Nederkoorn and his co-founder wrote down what kind of company they wanted.
- Build SaaS. The business model was a starting constraint, not an outcome.
- Sell to our peers. People whose problems the founders would recognise.
- Attach it to revenue. “So that if there was a downturn, people wouldn’t turn it off.”
- Make it technically hard. Explicitly to keep it interesting for the founders.
The fourth criterion produced the product’s defining constraint. The original tool emailed people based on what they did — or did not do — inside an application, and the hard part is the negative. “If you look at analytics products, especially in this era, 2012 when we started, a lot of them are doing sampling. And you can’t sample in order to send an email message that’s perfectly targeted.”
The launch target was deliberately tiny: five companies paying $10 a month, in April 2012. Preparing this talk, Nederkoorn checked what happened to those five.
One of the original five is still a customer. Twelve years on, that account is worth more than $100,000 in lifetime value. “It blew my mind that they’re still around and still on this revenue chart.”
It took two years to reach $1 million in ARR. The product that carried the company from there to $10 million was not fundamentally different — a transactional email endpoint and a newsletter send, both reusing the segmentation already built for triggered messages. “That fundamental thing that we did for companies was what helped us scale from one to 10.”
Three people mistakes
The second category is the one Nederkoorn is least forgiving of himself about, and the first item in it is a hiring mistake with a subtle failure mode.
Scrappy doers and systems builders are both correct, at different times. The problem is that the transition is invisible from inside. “When it’s time to scale up and that scrappy process starts breaking, they won’t be able to tell you that. They’re going to continue doing it that way.” The only detector he trusts is external: “you talk to people whose businesses are a little bit ahead of you… and then you realise all of the things that you’re doing wrong.”
The second was refusing to call sales what it was. Customer.io wanted to be MailChimp — wholly self-service — so when prospects asked to talk to someone, the founder took the call himself, understood their business, and sent them elsewhere if the fit was wrong.
I didn’t realise at the time but this was sales. If you’re there now trying to figure out how you’re going to reinvent how people buy from you, stop. Just call it sales is my advice.
The attempt to scale that instinct produced salaried “onboarding specialists” taking orders. It took years and several incarnations to arrive at account executives and a chief revenue officer.
The third was structural blindness inside customer accounts. Growing customers hired a VP of marketing who ripped Customer.io out; the company had no relationship deeper than the original buyer and never saw it coming. Shopify, its largest customer in 2013 and 2014, cancelled around 2015. Customer success exists because that kept happening, and the team is about 35 CSMs today.
The strap ratio
The last category is the founder’s own isolation, and it starts with a near-death that shaped a decade of decisions. Late in 2012 Customer.io nearly ran out of money, and Nederkoorn tried to raise across November and December.
“VCs go on holidays, angels go on holidays, everyone’s on vacation from Thanksgiving to the New Year. So it’s near impossible to get anything done then. I didn’t realise this.” The experience made him determined never to depend on investor money to stay alive — and, he now thinks, held the company back.
strap ratio = total capital raised ÷ ARRNederkoorn’s own coinage. Below 1 early is a warning, not a badge.
His argument is that a product which must be up around the clock needs enough people to run it, and that a ten-person team carrying a mission-critical service is a burnout machine. “Back when we were 10 people, everyone was burning out. Everyone’s exhausted all the time.” At the 250 people the company runs on today, “it feels really comfortable”.
The isolation extended to governance. He resisted a board for five years — “I had read these horror stories of VCs firing CEOs” — and structured party rounds so that nobody held enough to threaten him. By 2017 he had changed his mind entirely: “I really wanted a boss, and I wanted someone else looking at the business from the vantage point that I was looking at it.” Quarterly board meetings are now, he says, his favourite thing to prepare for.
Customer.io’s GetLatka profile records $70 million in revenue at the time of this talk, against $18 million in May 2021 and $30 million in January 2022 — and a single $35 million institutional round in March 2022, at a $690 million valuation, after a decade of the party rounds he describes.
The question that arrives at ten million
The talk stops at $10 million because that is where a different problem starts, and Nederkoorn thinks it is the one founders are least prepared for.
At about 10,000,000, you could sell. If you’re bootstrapped and you get to 10,000,000, incredible — you can sell the business and it can be life-changing for you. But if you keep going and you keep building, there has to be some other reason.
His own reason has two parts, and neither is the number. “The opportunity to make this a much bigger outcome for everyone on the team,” he said, “and I get to enjoy working with a really talented team of execs around me.”
He is candid that talks like this rarely stop anyone making the same mistakes. “Most of the time I listen, I’m like, yeah, I’ll remember that. And then I make the mistake myself. I think the best I can hope for is that you feel a sense of affinity or closeness with me when you make these same mistakes. And if you avoid them, great.”
Sources — Colin Nederkoorn, founder and CEO of Customer.io, speaking on stage; recording captured 5 September 2024. Revenue, headcount, funding and valuation figures from the Customer.io profile on GetLatka, with dates as recorded.
