Matteo Franceschetti on Eight Sleep’s Hiring, Execution and Sleep Technology

Matteo Franceschetti, co-founder and CEO of Eight Sleep, tells Harry Stebbings that competitive tennis and other sports made recovery important to him. He began asking why people spent a third of their lives on a passive mattress while technology improved so many waking activities.

Franceschetti imagines using the hours spent in bed to monitor health and eventually help identify disease. He says the company is researching body scanning and hopes to detect cancer in the future. These are research ambitions, not evidence that the existing product can detect cancer or replace medical care.

Asked whether he would start the company knowing how hard it would be, Franceschetti says probably not. One early challenge was manufacturing: after Y Combinator and a seed round, he flew to China and stayed for months to resolve problems the inexperienced team could not fix remotely.

The seed round benefited from enthusiasm for hardware and thousands of preorders, he says, but subsequent fundraising became more difficult. Hardware demands capital and presents risks in manufacturing, returns and replacements. He thinks investors’ caution is often understandable, even if a few companies can build substantial businesses.

His first test for a hardware business is whether its unit economics work. He recommends calculating the economics of an individual sale, including product costs, acquisition, card fees, returns and replacements. In his view, a company that loses money on each unit can make its situation worse by growing faster.

Franceschetti suggests gross margins above 50%, ideally 55–60%, and customer acquisition cost below 20% of price as targets for a strong company. These are his rules of thumb, not universal benchmarks. He says Eight Sleep had positive margin on day one and had been profitable for two quarters as of the interview.

At an earlier point, he says, gross margin was half its current level, acquisition cost twice as high, and air freight made each large unit expensive to ship. Sales were growing, but every additional sale consumed cash. He describes that as growth at any cost rather than a sound business.

The team raised prices and set a hard ceiling on blended customer acquisition cost. Growth slowed while the team found channels that could work within the constraint. Franceschetti says he calculated the ceiling from comparable marketing spending and the contribution margin the company needed.

Some executives accustomed to spending freely resisted the new constraint, he recalls. He believes forcing a lower cost encouraged the team to discover efficient channels. The lesson, in his view, is to hire executives whose prior experience matches the financial conditions in which the company must operate.

Eight Sleep increased the pace of testing advertising images, words and messages. Franceschetti estimates that its Facebook and Google acquisition cost had fallen about 75% over two years and blended acquisition cost about 50%. These are company-reported comparisons, not independently verified figures.

The company also uses podcasts, influencers and brand campaigns, including Mercedes F1 sponsorship. Franceschetti describes a seven-touchpoint rule: because the approximately $2,000 product is a considered purchase, people may need to encounter the brand repeatedly before buying.

Attribution is imperfect, so he looks both at individual channels and at total marketing spend relative to sales. He says word of mouth accounts for roughly 30–40% of revenue and podcasts work for the company, while recognizing that podcasts will not work equally well for every brand.

Franceschetti resists advice to reduce prices automatically because acquisition costs may not decline proportionally. He says a price increase in 2021 had little effect on demand and favors adding value or offering products at several price points. His pricing conclusions reflect Eight Sleep’s experience.

Stebbings questions the economics of a mattress bought only every several years. Franceschetti says Eight Sleep had moved back from selling mattresses to selling a technology-equipped cover that works with a customer’s existing mattress. In his account, the cover reaches a broader installed base and some customers purchase each new generation.

Early returns often came down to comfort: changing how a mattress feels can make one or both partners unhappy. Franceschetti says the product has improved. He also describes moving from the first cover to a mattress and back again as the company tested its model.

The hardest fundraising was the Series A. A prospective lead investor wanted a licensing agreement with a large mattress manufacturer; when that agreement fell through, the investor withdrew. Franceschetti says the condition had been clear and remembers calculating that the company had only weeks to solve the problem.

Another investor ultimately helped, including with manufacturing in China. Later, Keith Rabois and Khosla Ventures backed the company. Franceschetti says that period brought a rebrand, a premium position, the “sleep fitness” concept and the Pod, while previous products were discontinued.

He credits well-known investors with helping attract some hires and supporting bold decisions. His advice to founders is to identify the individual partner who understands the vision, rather than only pursuing a firm’s name. He says many investors passed because they did not share his long-term view of sleep technology.

Stebbings notes that a compelling vision still has to meet the realities of unit economics. Franceschetti believes operational problems can be fixed by a strong team, but says an investor must also believe in the underlying market. His vision of a very large sleep-technology business remains a prediction.

For clarity of thought, Franceschetti insists on writing. Putting an argument on paper exposes missing evidence and undefined outcomes, he says. His “grandmother rule” is to express a complicated idea simply enough for someone outside the team to understand it.

He wants premium, aspirational imagery paired with a simple value proposition. For Eight Sleep, he says, the central message is temperature control on each side of the bed. A benefit figure he mentions is a company claim, and he does not present supporting evidence in this discussion.

When a team suggests testing a new marketing channel, he asks why, whether existing channels can still grow, and what the data show. After debate, he wants the reasoning written down. Stebbings adds that teams need to define success before an experiment so they can judge its result.

Franceschetti prefers a written comparison of several options, including impact and effort, even when a team recommends one. Asynchronous documents let him approve or reject a proposal quickly; he also reserves time for spontaneous calls when a conversation is more useful.

He defines operational excellence as the quality and efficiency of output, combining speed with a high quality bar. Borrowing Rabois’s metaphor, he says “barrels” can carry a project from start to finish while others support them; the number of such leaders limits how many initiatives can advance.

His teams pair a goal with a counter-metric. Growth, for example, is measured alongside customer acquisition cost: maximizing growth alone could invite unlimited spending, while focusing only on cost might suppress experimentation. The two measures create productive tension, he argues.

He gives software quality as another example: spending months perfecting reliability can delay useful features. For important bugs he generally expects a fix within 48 hours or a strong explanation. His standard is to balance responsiveness with quality rather than treat either as an isolated goal.

Franceschetti prefers a good-enough version today to a somewhat better one much later, provided it clears the quality threshold. He says Y Combinator taught him to break large launches into smaller daily steps and learn from each one, then improve continuously.

The team calls fast execution “Eight Sleep speed.” Franceschetti recently asked executives to increase pace by 30%, a target he describes as a management goal rather than an independently established result. Some initially worried about quality, but he says the discussion produced concrete changes.

One proposed change was to avoid allowing rigid two-week plans to prevent teams from addressing urgent bugs. At an offsite, he challenged a proposed month-long validation to find what could be learned in three days with less than complete certainty. He says the teams developed a plan by the end of the week.

Hardware needs a different quality threshold from software because mistakes are harder to reverse, he acknowledges. The aim of speed is faster learning. To test the return to a cover, the team assembled a version in three days; he says it quickly drew 20–30% of sales despite being listed among accessories.

He also recalls launching a subscription and believes it could have been tested earlier and more thoroughly. Pricing can be changed, he argues, but companies must account for customers who saw an earlier offer and prepare support staff to explain differences such as warranty terms.

On hiring executives, Franceschetti warns against relying on prestigious company names. Someone who scaled a business when capital was cheap may struggle in a company focused on efficient acquisition. He says large-company habits may also conflict with the speed and risk tolerance he expects at a startup.

Eight Sleep gives every candidate a take-home project based on a real problem and asks them to present it to a panel, including senior executives. Franceschetti views it as a way to assess commitment, structured thinking and evidence. Candidates who do not want to complete it can opt out.

The company uses pass, hire and strong-hire ratings. He says it generally requires three strong-hire votes from roughly five or six interviewers, with the default being no hire. He would rather keep a role open than hire someone without strong support from the team.

After hiring rapidly in 2019, he believes many choices were mistakes. He tracks the number of candidates executives interview and the time between hiring stages. His rough company average is one hire per 100 people considered; that is an internal heuristic, not a general hiring rate.

Franceschetti says he personally passes on many candidates even after earlier rounds, but tries to document specific pros and cons rather than rely on a vague impression. He uses a consistent set of questions across previous jobs so that he can compare patterns across candidates.

For each role he asks how the candidate found it and what they were hired to do, their biggest achievement, their lowest point, what their manager would say in a reference check, and why they left. He looks for the candidate’s own contribution and supporting data, not only a team’s collective result.

Short tenures prompt him to ask why someone moved. A former manager offering an unusual opportunity or a layoff can explain a move, he says. Repeated layoffs cause him to examine the candidate’s judgment in selecting companies, although such patterns do not by themselves establish individual performance.

He sees excessive focus on title as a possible warning sign, while recognizing that titles can matter to people. At Eight Sleep, he says, an individual contributor must successfully lead squads before formally becoming a manager. He considers title changes after sustained strong performance.

The frameworks were not present on day one: Franceschetti says he developed them over years of studying other operators. On pay, he describes annual increases with different levels for top and solid performers and no increase for low performers; he says the company’s system still needs to become more sophisticated.

If performance is weak, he says managers should explain their concerns, understand the reasons and give the person a chance to improve. A demanding culture can be a poor fit for someone at a different life stage. If improvement does not happen, he favors a respectful separation and help finding a better fit.

Writing is central to his meeting process. He says he does not take a meeting without an agenda or memo in advance, ranging from a short page to several pages of growth data and analysis. Good documents are concise, clear, supported by data and organized around decisions.

A typical agenda contains an executive summary, points for discussion, progress against goals, what is working and what is not. Action items identify an owner and deadline. Franceschetti says he assumes same-day delivery if no deadline is specified, but admits a Sunday request caused confusion because he failed to state that expectation.

He thinks asking for Sunday work can be justified for a major Monday event, though not merely for his convenience. In meetings, attendees read the document beforehand and discuss comments rather than reciting every section. He says short, focused meetings become possible when the preparation is good.

People learn to write these agendas by practice and feedback. Managers ask for evidence, alternatives, second-order effects and a clear explanation of effort and impact. If someone is busy with an urgent incident, he says a few useful bullet points may be sufficient.

Franceschetti argues that preparing a board deck or meeting memo creates value for the writer by forcing them to examine what happened and why. Reading in advance lets others request more data and arrive ready to decide, rather than expecting an instant answer in the meeting.

At weekly executive meetings he shares lessons from hiring and operating mistakes, including his own. The team also writes a “pre-parade” describing what would make the coming quarter successful, then compares that expectation with a postmortem afterward. He says the documents are shared across the company, apart from compensation information.

He encourages respectful disagreement but says the directly responsible individual still makes the decision. He sometimes chooses to disagree and commit. For delegation, he weighs impact and his own knowledge: a company-threatening decision warrants more involvement than a minor issue outside his expertise.

He recalls a period of rapid hiring when the company became slower and quality fell as headcount grew. He believes he should have managed that culture better. Subsequent changes around the pandemic helped the company adjust, in his account.

Franceschetti describes his preferred leadership style as demanding and supportive: set a high bar, allow thoughtful risks and stand beside employees if a well-considered experiment fails. He argues that a leader’s job is to help people reach more of their potential, while treating them professionally and respectfully.

Stebbings challenges the idea that every employee wants to be pushed as hard as an elite athlete. Franceschetti acknowledges people contribute differently but says those who do not want this culture may not fit the company. These are his preferences for his organization, not a prescription for every workplace.

On dismissals, he says managers must consider the wider organization while treating the individual with respect. Employees should know how their manager views their performance and have the opportunity to correct course. He prefers regular feedback to waiting for an annual review.

Franceschetti admits that frustration can affect his delivery. He tells colleagues that he is direct and invites them to challenge his view with evidence. He also recognizes that he must control how he communicates, even while valuing prompt feedback.

The discussion turns to sleep. Franceschetti says duration and a consistent schedule matter, and offers seven to nine hours as a general rule of thumb, with individual variation. He is the CEO of a sleep-product company, and his health comments here are personal views rather than individualized medical advice.

He says warming the body with a bath or sauna before bed may help some people sleep, and describes body temperature changing across the night. His company sells temperature-controlled bedding, so he has a commercial interest in this topic. The interview does not establish that a particular routine or device will work for everyone.

When Stebbings asks about parents who cannot sleep nine hours, Franceschetti acknowledges that he does not have children and cannot speak from personal experience. He describes an employee with a child who scores well in an internal sleep competition, while recognizing that very young children can interrupt sleep.

Franceschetti says alcohol can disrupt sleep and offers observations about carbohydrates and caffeine, while emphasizing that responses differ between people. His numerical estimate of alcohol’s effect is not substantiated in the interview. He says evidence about sex and sleep is limited and does not provide a firm effect size.

He personally avoids late exercise because he finds it hard to fall asleep after evening tennis or padel. He recommends allowing time for heart rate and temperature to return toward baseline; the effect of exercise timing can vary by individual and intensity.

Franceschetti describes his own routine: around nine hours of sleep, frequent exercise and tracking heart-rate recovery with a watch. He says he follows a ketogenic diet, often eats once a day or uses a 16:8 fasting schedule, and has pizza with his wife on Fridays. He presents these as his habits, not a validated plan for everyone.

He admits that constantly trying to optimize can affect his mood. His emphasis is on consistency over months and years rather than perfection every day. He says he stopped drinking in 2019 because alcohol made him feel bad.

His wife and co-founder Alexandra Zatarain leads growth and marketing. He says they learned to set boundaries around work discussions, sometimes using Slack for business and WhatsApp for personal conversations. Early in the company, they had to adjust to the pressure of growth targets and his tendency to push too hard.

In the rapid-fire round, Franceschetti predicts that beds could become preventive health platforms. He recounts a customer who noticed unusual biometrics and later received emergency treatment, but gives no clinical details; an anecdote does not establish that the product detects or prevents disease.

He predicts sleep-time body scanning within roughly three years or less and says consumers should own their data. He argues that a plugged-in bed offers room for sensors and long-term measurement. These are forward-looking claims and the interview supplies no evidence that the envisioned scans are clinically validated.

He says he would like to lead a Mercedes or Ferrari Formula One team for a day, and admires Tesla’s customer experience and Nike’s treatment of athletes. The advice that stuck with him is to prepare carefully because a first impression cannot be made twice.

Franceschetti describes experimenting with recovery on vacation. He finds a few days of stillness useful, although he prefers remaining reachable for urgent work. He treats this as his personal approach to recovery rather than a rule for others.

Looking toward 2033, he hopes Eight Sleep will be public, have improved millions of lives and saved a million lives. These are the CEO’s aspirations, not current results. He says he wants to build an enduring global brand around sleep and preventive health.