The Lean Startup

Miles: A team spends six months building something. They launch it. Nobody arrives. Not a disappointing number. Nobody.

Miles: So they assume it isn't good enough yet. They fix the bugs, they add features, they get so quick at releasing that new versions go out many times a day. The number stays where it was.

Miles: Eventually somebody does the thing none of them had done, which is find actual teenagers and ask them. Two answers come back in an afternoon, and both say the six months were unnecessary.

Miles: Today: The Lean Startup, by Eric Ries. In two sentences: a new venture isn't a small business, it's a standing question about whether anybody wants the thing, so progress gets counted in questions answered rather than work completed. Two days running we've watched books move work faster. This one asks whether it's moving anywhere.

Miles: Sixty Books. Presented by Stoar Consulting and Coaching.

Miles: Book thirty three, and the reading room's open. Sixty Books, Miles talking. Terms as ever: sixty books, one to an episode, the best a hundred years of management writing has on offer, each put to the same two questions. What's here for a person who leads, or gets led by, a mind that runs on other rails. And how much survives a working day with AI sitting in the middle of it. Nora keeps the evidence, and she had this one sourced before a line got written.

Miles: Nora, we closed yesterday on what neither of those books will say out loud. Both hand you a plant that hums. Neither asks what it ought to be making.

Nora: Ries has a word for the gap, which is waste. A perfectly efficient line producing something nobody wanted is still producing nothing, and worse than nothing, because it looks like effort and costs like effort and nobody inside can tell.

Miles: So flow isn't the question.

Nora: Flow is the second question. The first is whether the thing at the end has a buyer, and that's answerable long before the line is finished. The cold open is his own company, in the book because he got it wrong. Six months went into sparing users the bother of installing a second chat program. Then they asked the teenagers, who turned out to be running three of them already.

Miles: So the problem had been solved before they started.

Nora: It had never been a problem. Which is a particular kind of expensive, and this book's method exists to make that discovery cost an afternoon rather than half a year.

Miles: Tell it properly, because it's the spine of the book.

Nora: Two thousand four. Ries co founds IMVU with Will Harvey, building three dimensional avatars you chat through. The strategy is careful, and it's the kind that sounds wise in a room. Don't build a chat network from scratch, because the existing ones already hold everybody's friends. Sit on top of them. Be an add on.

Miles: So the engineering goes into interoperating.

Nora: Months of it, against every existing network. Then they ship, and the download count sits at roughly nothing. They read that as quality, so they fix and polish and release faster than almost anybody at the time, and the count doesn't move.

Miles: And then the interviews.

Nora: Two things. First, a girl who won't invite her friends, because she doesn't know yet whether it's any good and she isn't going to find out in front of people whose opinion she cares about. The whole strategy depended on her doing exactly that. Second, the barrier they'd built the strategy to remove wasn't there in the first place.

Miles: And the fix.

Nora: Small, and it's the point of the story. They add a way to be matched with a stranger, so somebody can try it alone before anyone they know sees them. That works, and the company stops being an add on and becomes its own place. What the revenue did afterwards is in the book, and I'll leave it there, because one company's outcome proves nothing about a method. Which matters more here than in most books.

Miles: Give me the author.

Nora: Eric Ries, born nineteen seventy eight, computer science at Yale. As an undergraduate he co founded a company meant to put student profiles in front of employers, which went under in the dot com collapse. Then an engineering job at a firm building a three dimensional world online, which also folded.

Miles: Two failures before the book's own company.

Nora: Two, and the method is visibly assembled out of them. IMVU starts in two thousand four. Steve Blank comes in as an investor, and Blank's customer development frame is one of this book's two parents. The other is Toyota's production system, which is where the word lean comes from.

Miles: And he leaves.

Nora: Steps down as chief technology officer in two thousand eight and starts writing the material up on a blog called Startup Lessons Learned that same year. The book arrives in twenty eleven from Crown Business and enters a New York Times list at number two. Over a million copies and more than thirty languages, which is the publisher's figure rather than an audited one.

Miles: And after.

Nora: A second book in twenty seventeen, on doing this inside large companies. Then an alternative American stock exchange, built to reward longer horizons, which the Securities and Exchange Commission approved in May of twenty nineteen.

Miles: Take me through it, in his order.

Nora: Three parts. Vision, Steer, Accelerate. Vision sets the definition, which is the quietly radical sentence in the book. A startup is an institution built to create something new under conditions of extreme uncertainty. Nothing in there about size, sector, age or a garage. If you're making something new and genuinely cannot know whether it works, you're inside it, four people in a company of forty thousand included.

Miles: So what counts as progress.

Nora: Validated learning. Not features finished, not milestones hit. What you now know and have demonstrated. First up are his two leap of faith assumptions: the value hypothesis, whether the thing gives the person using it anything, and the growth hypothesis, how new people find it.

Miles: Right.

Miles: Then the middle section.

Nora: Steer, and it's the engine. Build, measure, learn, run as a loop, and the instruction everybody skips is that you plan it backwards. Decide what you need to learn, work out what would measure it, then work out the smallest thing you could build to produce that measurement. Most teams start at build, because build feels like the job.

Miles: And the minimum viable product.

Nora: Is the most misread term in modern management, and it isn't hard to read right. It's the version that starts the learning loop for the least effort, and it needn't be a product. He describes concierge versions delivered by hand to a handful of people, and versions where the customer sees something automatic with a person behind the curtain. Smallest experiment, not smallest product.

Miles: And measurement.

Nora: Innovation accounting, three steps. Baseline, with something real in front of real users. Tune the engine, pushing those numbers toward what the plan needed. Then pivot or persevere, on evidence rather than on who spoke last.

Miles: And the metrics he won't have.

Nora: Vanity metrics. Totals that only ever rise. Cumulative users, cumulative revenue, cumulative anything. His alternative is cohorts, taking each group who arrived in a given week and asking what that group did, plus split tests. And three properties he wants from a measure. Actionable, so it shows cause and effect. Accessible, so the people it's about can read it. Auditable, so you can check it against a human being.

Miles: And then the word everybody borrowed.

Nora: The pivot, defined narrowly: a change of strategy without a change of vision. You keep what you were trying to do in the world and you change the route. He names ten kinds, from zoom in, where one feature becomes the whole product, out to a technology pivot, same problem by other means. Ten of them, and out in the world the word has collapsed into one meaning, which is that the last idea didn't work and we'd like to keep the lease.

Miles: Okay.

Miles: And Accelerate.

Nora: Small batches, where yesterday's book and this one shake hands. Three engines of growth, run one at a time: sticky, viral, paid. And the five whys out of Toyota, with a rule attached he calls proportional investment. Fix a fault at a scale matching the damage.

Miles: And where does it run out?

Nora: With the problem sitting over the whole category. Every case in a book like this is picked afterwards, from among the survivors. If a method's evidence is its examples and the examples were chosen because they worked, the evidence is the choosing.

Miles: So the failure numbers, because the genre rests on them.

Nora: They're shakier than people think. The figure everybody repeats is ninety percent, and I went looking for the study under it. There isn't one. It circulates with no citation, and the trail ends.

Miles: So what's actually counted.

Nora: Two things, neither of them what the figure claims. The United States Bureau of Labor Statistics tracks every private sector establishment from the day it opens. Of those born in twenty thirteen, seventy nine point six percent were open a year later, fifty point six percent at five years, thirty four point seven at ten. A census, so the count is solid. What it isn't is startups. A business that shut because its owner retired sits there beside one that ran out of money.

Miles: And the second thing.

Nora: Venture backed firms, and this part belongs in a statistics class. A Harvard Business School lecturer went through data on more than two thousand companies that raised venture money between two thousand four and twenty ten. One data set, three answers. Failure as assets liquidated at a loss to investors, thirty to forty percent. Failure as investors not getting their capital back, about seventy five. Failure as the company missing projections it set itself, over ninety five.

Miles: So the number is the definition.

Nora: The number is entirely the definition, and it reached the public through a newspaper rather than a journal, so the working isn't there to check.

Miles: All right. Has the method itself been tested?

Nora: Twice, and it's the strongest evidence any book in these three episodes has behind it. Twenty twenty, Management Science, by Camuffo, Cordova, Gambardella and Spina. A randomized controlled trial, a hundred and sixteen Italian startups, fifty nine treated and fifty seven control, sixteen rounds of data over about a year. Everybody got the same ten training sessions. The treated group was also taught to state falsifiable hypotheses, design tests for them, and fix in advance what would count as a fail.

Miles: Results.

Nora: Treated firms earned more revenue, pivoted more, and were no likelier to quit. Then twenty twenty four, Strategic Management Journal, a replication at scale. Seven hundred and fifty nine firms across four randomized trials, two in Milan, one in Turin, one in London. Treated firms were nine point eight percentage points likelier to terminate the idea, and terminated about two point seven weeks sooner. Likelier to pivot once, less likely to pivot more than twice, and about seven thousand euros ahead on revenue.

Miles: Say more, because terminating more sounds like losing.

Nora: It's the finding I'd frame and hang. The training made nobody more determined or less. It made them quicker to find out, in both directions: the weak idea died earlier and the thrashing stopped. What it tests, though, is teaching a scientific approach to founders in an accelerator. It is not a test of this book, and I'd rather say so than let a good result stand in for one.

Miles: And the argument from theory.

Nora: Felin, Gambardella, Stern and Zenger, twenty twenty, Long Range Planning. A hypothesis is not a guess, they say, and rushing to test skips the part where somebody works out what's worth testing. Customer feedback reaches only what customers can already picture, so cheap experiments suit improving what exists and suit badly what nobody has seen yet.

Miles: Then the tier statement.

Nora: Plainly, then. A practitioner's book, reporting no study and claiming none, arguing from cases its author picked, which is the weakest evidence there is and the currency this shelf trades in. What's unusual is that researchers tested the central instruction afterwards, at a size that makes the answer worth having. The teachable core survives that. The certainty around it doesn't, and the terms he coined have since sold a great many things he never said.

Miles: Our half. And you flagged this one before we started.

Nora: Because the loop rewards a particular person, and it isn't the one your review cycle is built to find. Iteration makes a virtue of the rough first attempt. A performance year does the opposite.

Miles: Start with the first half.

Nora: With training research, because it's been measured. Keith and Frese, two thousand eight, Journal of Applied Psychology. A meta analysis of twenty four studies and two thousand one hundred and eighty three people, comparing training that encourages learners to make errors and work out what happened against training that walks them through the correct steps.

Miles: And the encouraged errors won.

Nora: Mean effect of point four four, which is moderate. Then the number that matters. On tasks structurally unlike the training, meaning something genuinely new, the effect was point eight zero. Laboratory and classroom work with trained tasks, so don't stretch it to a career. The direction runs against instinct though: the further the real task sits from what you practised, the more it helped to have been allowed to get it wrong.

Miles: Now our population.

Nora: And here I've got better than an analogy, because somebody pooled it. Tran, Wiklund, Antshel, Jhawar and Montgomery, in Entrepreneurship Theory and Practice, online since late twenty twenty five. Two hundred and ninety eight effect sizes out of forty seven studies on ADHD and entrepreneurship.

Miles: And it comes out mixed.

Nora: It splits, and the split is the lesson. The hyperactive and impulsive side is positively associated with entrepreneurial attitudes and with entrepreneurial behaviour, meaning actually starting things, and shows no significant relationship with how the venture did afterwards. The inattentive side is negatively associated with those later outcomes. Total symptoms and a diagnosis run the same way: positive on the starting, negative on the outcomes.

Miles: Which flatters nobody and insults nobody.

Nora: It describes a mismatch, and it's this book's argument turned round. The trait that gets you into the loop isn't the trait that carries you through it. The measures underneath are mostly symptom questionnaires rather than clinical assessment, the studies observational, and a correlation across forty seven samples says nothing about any individual. Nobody listening should hear a portrait of themselves.

Miles: And there's a cost side.

Nora: There is, because a loop is built out of not knowing yet, and for some people that's the expensive part. Jenkinson, Milne and Thompson, twenty twenty, in the journal Autism. A review and meta analysis of twelve studies, all self report questionnaires. Of the ten reporting usable statistics, nine found a significant link between intolerance of uncertainty and anxiety, at about the strength the same measures show in people without an autism diagnosis.

Miles: So the method isn't free.

Nora: Nothing like free. Permanent uncertainty is a real cost and it falls unevenly. The answer isn't less iteration. It's saying out loud what's being tested, what would count as a result and what happens either way, before anybody starts. Which is the discipline the trial taught, and the part offices leave out.

Miles: And the review cycle.

Nora: Is the part with no study behind it, and I'll say so before I make it rather than after. A performance year measures twelve finished months against goals set at the start of them. A column for delivered, a column for missed, and none for started six things roughly and closed four with something worth knowing, which is a good year inside this book and a thin one on a form. So a company runs a build, measure, learn loop on its products and a nineteen fifties scorecard on its people, and the person who's best at the loop pays for it in March.

Miles: And the second caution.

Nora: The loop can turn into a permission slip. Constant change is not iteration, and the replication found that too: trained firms pivoted once more often and three or more times less often. If everything's a pivot, nothing gets built on top of anything, and the people who lose most are the ones who need the ground to stay put.

Miles: Fair enough.

Miles: Sixty seconds, in the frame this show uses. The Lean Startup says that when you can't know whether a thing works, progress is a question answered rather than a task completed, and management's job is making the answer cheap. For anyone leading a brain wired to a different plan, or running one, here's Monday. Reward the rough first attempt out loud, because the best evidence going says training that permits errors beats training that prevents them, hardest of all on work nobody has done before. Write down before the test what you're testing and what counts as a result, because that sentence is the method and it's what makes uncertainty survivable for the people who find it expensive. Check the paperwork agrees with the loop: a form with room only for delivered and missed is punishing the person you told to experiment. And keep the pivot rare and named. Once, on evidence, is the method. Every quarter, on mood, is weather.

Miles: Now the machines.

Nora: The standing warnings about this class of tool sit after the music now, in a wording nobody moves, so I'll gesture at them and spend the minutes on what this book alone gives us.

Miles: Go on.

Nora: Three steps in the loop, and these tools have made exactly one cheap. Build has collapsed in cost. Measure and learn haven't moved. So the predictable failure is a team making nine versions in the time one used to take, with the same single channel for finding out whether anybody wanted any of them.

Miles: And the hit rate on what you do try.

Nora: Has been published, by the people running experiments at industrial scale. On one large commercial testing platform, roughly a third of the ideas tried improved the measure they were built to improve. A third did nothing, and a third made things worse.

Miles: So the supply of ideas was never the limit.

Nora: It never was. Valid tests were the limit, and we've spent two episodes learning what to call that. More candidates aimed at the slowest station raises the queue, not the output, and producing candidates is the thing these tools are best at.

Miles: And accuracy.

Nora: Is what belongs in this episode, and there's a field experiment on it. Twenty twenty three, seven hundred and fifty eight consultants at a large firm, pre registered and randomized three ways: no tool, the tool, the tool plus training in prompting it. On tasks the tool suited, the assisted groups finished twelve point two percent more, twenty five point one percent faster, and rated forty percent higher on quality, with the biggest gains to people who'd scored in the bottom half beforehand.

Miles: And the other task.

Nora: Was built to sit outside what the tool does well. There the control group was right eighty four and a half percent of the time, and the assisted groups came in nineteen percentage points behind. Same people, same afternoon, same tool.

Miles: So it's confidently wrong in one place.

Nora: Yeah. Confidently wrong in one place and genuinely excellent in another, with no signpost between them. And a confident wrong answer is a vanity metric with a better vocabulary. It looks like progress, it can't be argued with, and it isn't auditable, which was the third thing he asked of a measure.

Miles: Last word.

Nora: His own test, which costs nothing and is harder than it sounds. At the end of each loop, ask what you know now that you couldn't have written down last week, and whether somebody's behaviour taught it to you. These tools will hand you a hundred things to try before lunch. Not one is an answer until somebody outside the building acts.

Miles: That's The Lean Startup.

Miles: Next time: The Checklist Manifesto, by Atul Gawande. Three episodes we've spent making the work go faster, and fast breaks in a particular way. It doesn't drop the job. It drops one item out of the middle, the fourth thing on a list of nine, and the fourth thing turns out to be what everything after it was standing on.

Nora: And the evidence on that one has a twist in it that most people quoting the book have never heard.

Miles: That's a lot, Nora. Thank you.

Nora: That's the book.

Miles: Back tomorrow with the next one.

Miles: Sixty Books comes from Stoar Consulting and Coaching. Logan Williamson is a Licensed Professional Counselor in Texas and Louisiana. What you got was a book explained, for people who run teams and people who work in them. Not coaching. Not counseling. And not one word of it was a reading of your situation.

Nora: Citations for all of it are with the transcript on the episode page. Book thirty four lands tomorrow.

Miles: Two things on the record. Both voices are generated. There's no Miles and no Nora, neither holds a license, and none of what they said was directed at the person listening. Every script is drafted with AI help, then edited and approved by Logan Williamson, a Licensed Professional Counselor in Texas and Louisiana, at Stoar Consulting and Coaching.

Nora: And one note about the AI segment. The evidence that these tools help neurodivergent people at work is early. There are controlled trials now, but they measure things like conversation skill rather than how a job actually goes, and for work itself most of what exists is individual providers reporting success. So take it as promising and not proven. The tools also get rebuilt every few months, so whatever you set up today is built on something that moves.