Lean Startup is one of those books that if it comes up in a conversation, and you haven’t read it, the conversation is then about ‘how have you not read it’. So when I say I got non-zero value from reading it, part of that value is not having to talk about it anymore. (Writing this post is erasing some of that value).
One-line summary of the book: product-first companies should be managed as products, not as ‘traditional businesses’. And therefore you shouldn’t do normal forward accounting until your product has gone through a release cycle. As best I can tell ‘traditional business’ is code for a McDonald’s franchise, i.e. a template model entering a new market.
I can’t tell if it’s targeted to big companies who he’s telling to be ‘more startup like’, or to me, a solopreneur who knows zilch about marketing but still trying to launch + sell my thing.
I’m generally interested in ‘the information revolution means that all of our beliefs about management are wrong’ claims. This book has traces of that (‘company as portfolio of products’) but it’s not the main idea.
Also I can’t tell if it’s a product design book or a management book telling you to learn more product design. If the former, it needs more product content. If the latter, it could be an index card; and certainly it seems like the author was seriously influenced by Steve Blank’s product book (see ‘books it draws from’ section).
Also the case studies are bonkers (see ‘bond villains’ section).
If you came here for LARPing content and made it this far, you have my deepest apologies. It’s just a funny title.
- Core ideas
- Case studies mostly Bond villains
- Metrics and growth
- Company as portfolio of products
- Manufacturing and process control
- Books it draws from
The book had a few ideas that I liked, which were novel or at least current in 2011, and that if you’re not doing you should.
A company that exists to launch a product should organize itself for product development. (Yes obvious but people still somehow don’t do it).
Cross-functional teams with ‘launch authority’ are in many cases the right model for developing good products.
Deep discussion of MVPs. Release novel ideas ASAP with quality compromises so you can see how the market responds to them. Listen to user reactions but don’t always take their explanations at face value; don’t always trust them about ‘what we want instead’. (‘Product vision’ exists at this friction point).
Product management as a job is about learning from releases. Like the terminator, an organization or a team is a learning computer, in that it succeeds by iterating over feedback.
A leader’s job in a product-driven company is to develop a system for enabling many people’s innovations to ship and be tested, not to have the ideas and not even to manually cultivate innovation.
A lot of the case studies make me feel like this is a book to use the appeal of startup ROI to sell the startup model to big companies. (‘Big’ relative to me, of course, call it 100+).
The insights from manufacturing also feel contrived; he writes at one point:
I began to study other industries, especially manufacturing, from which most modern theories of management derive.
I question the word ‘modern’ in this sentence.
In a lot of places Ries tries to use process control arguments from manufacturing to advocate for product development practices. I think this practice is ‘skeuomorphic’ in that it uses a convenient metaphor from our racial memory. That doesn’t make it an argument.
He also, in places, disses non-startup businesses as wasteful approaches in a high-uncertainty environment. He also disses business plans as not suitable for startups because they’re rigid and rely on information that the founders don’t have. I don’t know if I agree with this, but at minimum, I believe that one should plan + estimate to the extent that’s possible, while avoiding the trap of overpromising or overselling.
This ‘no plans’ argument feels at times like an argument aimed at reluctant investors. Like this:
(talking about non-startups). its success depends only on execution – so much so that this success can be modeled with high accuracy.
This is why so many small businesses can be financed with simple bank loans
Because startups often accidentally build something nobody wants, it doesn’t matter much if they do it on time and on budget.
Having seen what happens to a company at the end of its burn, I’ll quote a different management textbook to rebut: When the fall is all that’s left, it matters a great deal.
Case studies mostly Bond villains
The book was written in 2011 so grain of salt here, but a the companies he talks about are in retrospect a rogues’ gallery of failures and scams.
Like Intuit. I’ve used their product. Not only is it shit, but Intuit, in order to keep existing, lobbies to keep the tax code shitty, which in turn makes their product complex and impossible to fix. Like Intuit is such a bad place that they’ve solved Hanlon’s razor: they’re incompetent because they’re evil.
There’s a chance that when we look back on the ruins of our civilization, Intuit’s aggressive lobbying to reduce state capacity is the root cause of the jenga tower collapsing.
Intuit is his crowning example of a good company; it gets more screen time than the company Ries founded, which made 3D avatars or chat software or something. Perhaps Intuit was innovating in the 80s when it started or the 00s when this book was released but at present it’s something between a lice and a mosquito.
The only pleasure that Intuit has ever brought me is the steady trickle of Propublica articles that embarrass Intuit, the IRS and the congress alike. But if I lied to the IRS my ass would be in jail and Intuit somehow gets to stay free, so on net, I hate them.
Intuit had an internal product called SnapTax that Ries champions as an example of the ‘in house startup practice’ that he wants to sell to other midrange SaaS companies. Ries footnotes a ‘private interview’ as revealing to him that it was senior management that enabled this, not smart leaders managing up.
Did they face constant meddling from senior management, which is the bane of innovation teams in many companies? No, their executive sponsors created an “island of freedom”
What allowed the SnapTax team to innovate was not their genes, destiny, or astrological signs but a process deliberately facilitated by Intuit’s senior management.
More information about how this worked would make a great case study. As things are he’s just showcasing a company that was maybe his client? And turned out to be pret-ty scummy 9 years on.
Another source of anecdotes is Toyota, a classic case study for every B school program, and which I’m guessing he knows from books and not firsthand. A lot of the Toyota stuff is to justify his manufacturing and cycle-time metaphors. Meh.
Another case study is a company funded by “Facebook’s initial investor Peter Thiel” that can “process voter identity” for 47 states to deliver what I think may be targeted ads.
Groupon is one of the fastest-growing companies of all time.
Kodak’s history is bound up with cameras and film, but today it also operates a substantial online business called Kodak Gallery.
Despite its $500 million budget and high-profile origins, the CFPB is really a startup.
He obsesses over the Toyota ‘five whys’ root cause analysis process, and a lot of the examples are bonkers. At one rails shop, he concludes a five whys postmortem with the claim that:
Without the Five Whys, we would have never discovered all of the information we did here.
The resolution in this case was to use bundle for modern Ruby packaging and to invoke it in CI. I mean, fuck. There are a lot of other commandments that get you here, like ‘don’t use shitty tools’ or ‘if you use Ruby hire at least one person who knows Ruby’.
Metrics and growth
Innovation accounting will not work if a startup is being misled by these kinds of vanity metrics: gross number of customers and so on.
‘Innovation accounting’ is I think something he made up and either is never defined or I had a brief optical migraine and missed it on the page.
His ‘vanity metrics’ claim is more interesting though I kind of hate the framing. It’s saying that topline metrics like revenue or user count are bullshit (that part is bogus) because they don’t directly measure the quality of your product development process. (Half-bogus – they may not directly measure process qualtiy but that doesn’t make them BS).
He thinks growth that’s ‘not from improvements driven by product development’ is silly, which I in turn think is silly.
The good part of his metrics argument is that it’s worth analyzing your funnel and drawing different conclusions about improvements at the top from improvements at the bottom. And that if the bottom of your funnel isn’t healthy, there’s no point in investing in the top. (I think the reverse is true too).
Another good point here is that ‘viral, sticky, and paid growth models’ are different, and that growth which happens ‘as a side effect of product usage’ is often more sustainable. VC-$$$-driven growth (especially combined with inexperienced teams) can create truly shitty companies who create surprisingly little value before the training wheels come off, at which point we sometimes discover the bike didn’t have a rear axle and the ‘stabilisers’ were load-bearing.
Company as portfolio of products
You know there’s that end part of a book where the editor was like ‘nobody’s really going to get here probably, you can write what you want, I’m sure as fuck not reading it, this is where you put that stuff I forced you to cut from the intro wink wink.’
I’m particularly glad that I read the ‘epilogue’ because it has this tidbit:
established companies struggle to invest consistently in innovation … [because of] intense pressure from public markets
this is a consequence of the accounting methods we have developed for evaluating managers, which focus on gross “vanity” metrics
(note ‘gross’ here means ‘total’, not ‘disgusting’, probably)
What is needed is a new kind of stock exchange … I propose we create a Long-Term Stock Exchange (LTSE)
Before he fully decamps to ‘let’s reinvent stocks around a metric I invented’, he talks about treating a company as a ‘portfolio’. I think he means this from a management perspective (have a lot of balls in the air) rather than an investing perspective (allocate your money to hopefully independent assets based on some sort of time-series analysis).
He doesn’t get to the guts of how this works or present a math model.
I’m interested in the question of how to invest in things that may compete with each other and merge them under one roof. There are arguments that Blockbuster would have killed Netflix if they bought it, and that they’d have been right to do so to postpone the date at which their high margins went away. If your ‘portfolio’ model can’t thread the needle here and explain how to manage the timing of low-margin disruptive competition, your portfolio theory of in-house startups is incomplete.
I’d also want a little more gambling theory / financial modeling; even if the math doesn’t belong here, at least some of the high-level concepts like variance and how they apply to running a relatively undiversified portfolio with relatively low N.
The motivating problem for ‘company as portfolio’ is senescence:
when companies become larger, they inevitably lose the capacity for innovation, creativity, and growth.
I saw a description of a book called ‘the soviet innovation decision’ which IIRC says that the soviet economy had too many steps for approving any change, and that this created an anti-innovation culture, and didn’t just make the economy uncompetitive but also killed it. ‘Permissionless innovation’ is a phrase people throw around as well.
The problem of how to prevent companies from ending up here is a real one.
Manufacturing and process control
‘Knowledge work is actually just factory work’ is a thread in the book. It’s a daring metaphor, but ultimately a weak model; factory work is about optimizing repeatable processes with known resource constraints.
I don’t think anyone knows how to do this precisely for knowledge work, even for contexts where experts have sort of a template, like in agency work. And especially not for the form of high-uncertainty product development Ries seems to be focusing on.
(I learned from the book that kanban means ‘capacity constraint’, which is delightful. Then I learned from a different source that this isn’t the case; the second source was a japanese dictionary).
There’s a limited amount of liberty you get to take with explanations before you sound like Björk talking about her TV and I think Ries crossed the line.
There’s an example in one chapter about ‘single-piece flow’. We assume that when a process has steps (packing an envelope in his example), you can save time by doing 10 or 100 of each step for ‘economy of motion’ and then moving on to the next step. In practice, he says, this intuition is wrong because:
- you’re wasting effort moving around partial work, and
- if there’s something wrong with your process (the paper doesn’t fit in the envelope), it’s better to learn that early and retool.
This is the lead-in to a bunch of process control metaphors for product-based businesses. He talks about the ‘large-batch death spiral’, which is the case of queueing up a bunch of work before a feedback step and then trashing it when you release something bad.
I can’t see competent managers queueing up a bunch of work without quality control, regardless of how their teams are organized. I wonder how waterfall-era MSFT did user testing and redesigns.
I half believe his claims about non cross functional teams having to throw out design work when the designer is siloed:
These problems inevitably turn into interruptions for the designer, and now those interruptions are interfering with the next large batch the designer is supposed to be working on. If the drawings need to be redone, the engineers may become idle while they wait for the rework to be completed. If the designer is not available, the engineers may have to redo the designs themselves. This is why so few products are actually built the way they are designed.
I probably believe his conclusions about frequent validation and cross-functional teams, but the manufacturing and process control based arguments here aren’t speaking to me.
Books it draws from
Ries cites often:
- Steve Blank’s ‘Four Steps to Epiphany’ about product development
- The Innovator’s Dilemma, I think about why innovation is hard in institutional systems?
- Sources about Toyota, possibly some combo of the Ohno TPS book or Liker’s Toyota Way. He cites WE Deming as well, a management research person who lectured in postwar Japan and probably had contact with the Toyota people
Ries’ apotheosis of prewar industrial researcher Frederick Winslow Taylor is IMO a low point. ‘Taylor invented modern white-collar work’ is in there somewhere. I think Taylor actually invented a steel cutting technique, and maybe also 1940s blue collar work.
To the extent that Taylor still has an influence on management culture in an increasingly-services, moving-to-knowledge work economy, that’s a bad thing. His main management advice as I understand it is micromanagement. His ideas about standardization and remote control of workers by managers are totally unsuited to knowledge work.
If Taylor invented white-collar work it’s only to the extent that you define white-collar work as organizing manual labor and making it efficient. Process control specifically of people.
I’m not sure this matters as a criticism because I don’t think this book takes anything substantive from Taylor, but I hate the idea of Taylor and look forward to reading something he wrote some day so I can reinforce my bad first impression
Ries never cites, but mentions at the end, ‘Certain to Win’, the Chet Richards book about John Boyd. Boyd was a jet pilot who invented E-M theory and the OODA loop. Every time Ries wrote anything about cycle time in a competitive environment, I frustration-highlighted it and wrote ‘john boyd’. I’m glad he hat-tipped him at the end.
The actually right thought at the center of the book:
Having no system at all was not an option for IMVU and is not an option for you.
If you manage any group of people, you need some way to work together that’s predictable, fair and effective. It doesn’t have to be Ries’ but it doesn’t have to not.