There is a saying that goes back to Milo of Croton: lift a calf everyday and when you grow up, you can lift a cow. The story goes that Milo, a famous wrestler in ancient Greece, gained his immense strength by lifting a newborn calf one day when he was a boy, and then lifting it every day as it grew. In a few years, he was able to lift the grown cow. The calf grew into a cow at about the rate that Milo grew into a man. A rather freakish man apparently, since grown cows can weigh over 1000 lb. The point is, the calf grew old along with the boy.
I have been pondering this story for a couple of years, and it has led me to a very fertile idea about product design and entrepreneurship.
I call it the Milo Criterion: products must mature no faster than the rate at which users can adapt. Call that ideal maximum rate the Milo rate.
It seems like a simple and almost tautological thought, but it leads to some subversive consequences, which is one reason I have been reluctant to talk about it. The most subversive effect is that it has led me to abandon lean startup theory, which is now orthodoxy in the startup world.
As a consequence, I have mostly abandoned notions like product-market-fit, minimum viable product, pivots and the core value of “lean.” I only use the terms to communicate with people who think in those terms. And I can’t communicate very much within that vocabulary.
Physical Products and Services
Commercial airline travel is an example of a service product that satisfied the Milo Criterion during its evolution. In the early days, the user experience was not very different from riding a train or bus. Airport designers modeled their early efforts on railroad stations leading to familiar experience for early air travelers.
But modern air travel, which has evolved over nearly a century, is a very different complex of behaviors that has drifted far from bus and train travel. Just look at the enormous number of complex behaviors we’ve learned:
- Checking in (online and off)
- Security checks and rules about carrying liquids
- Gates and air-bridges that look nothing like railroad stations
- Checked baggage and hand baggage rules
- Seat belts and rules about staying seated at certain times
- Baggage carousels for retrieving luggage
- Dealing with layovers
- Online bidding for cheap ticket deals
- Airport parking and car rental options
- Duty-free shopping
- Visas, passports, immigration, customs
- Rules about when you can use electronic devices
We’ve been able to get this far successfully because we took our time. By a happy coincidence, the physical constraints of the technology limited the rate at which airline travel could evolve.
Another example is driving, which is estimated to involve close to 1500 separate sub-skills. It took us about a century to get to modern driving, GPS, zipcars and all, starting with horse-drawn carriages.
This sort of long evolution trajectory is generally the case for physical products and services. They are naturally rate-limited by a variety of factors, so they tend to evolve and mature in ways that naturally satisfy the Milo Criterion.
Web Products
Thanks to the lack of physical constraints, Web products can go from paper napkin to fully realized vision in months rather than decades. They can evolve at rates that far exceed the Milo rate.
It takes decades to build out airline infrastructure in a country. Even the Chinese government cannot move arbitrarily fast.
For Web products though, there appear to be no real limits, other than typing speed, to how fast you can build things. And thanks to certain pathological externalities, they perversely go as fast as they can. In fact going faster and faster has become the motto of the sector.
Successful Web products do seem to satisfy the Milo criterion though. I tried applying the criterion to a whole bunch of products, and it turns out to be a pretty reliable way to sort the two classes. Google Wave fails the criterion. Google+ satisfies it.
The criterion seems to be descriptive. Is it prescriptive?
Consider modern software development. A set of behaviors have emerged in the last decade that appear to increase the success rate of Web products:
- Starting small and simple
- Building, testing and iterating rapidly
- Testing with active users as frequently as possible, starting as early as possible
It is important to note that these principles were discovered bottom-up by the practitioners, rather than prescribed top-down by the theorists. Theoretical codification followed rather than led. So it is possible to criticize the theories while accepting the empirically validated practices.
I have come to believe that much of the theorizing about why these methods work is mostly noise. These theories — including lean startup theory — are mostly a set of just-so explanations that serve to motivate practically effective behaviors, the way religions motivate moral behavior.
So even though the theories lead to the diffusion of useful behaviors, the flaws limit their potential.
I won’t attempt a full critique here but offer a basic axiom for an alternative theory:
The primary reason these behaviors are effective is that they slow down the process of software development and maintain the optimal behavior modification rate for humans.
In other words, the Milo Criterion is not just descriptive. It is prescriptive. It is the dominant dynamic for successful products.
It leads to alternative explanations for why the effective practices work. It leads to building blocks that are different from the ones recommended by lean startup theory.
In fact it is a pretty fertile starting point for a whole different approach to thinking about entrepreneurship and product development. I’ve been developing these ideas, mostly in private, and applying them to my own business decisions.
Slow Marketing
I don’t like being cryptic, but in this case, I am not going to elaborate further (at least not right now) because the very thought of the tedious and potentially acrimonious arguments that might result is enough to turn me off. I don’t have enough skin in the game to make it worthwhile. Perhaps I am getting old and conflict-averse.
So I am not going to share my explanations or alternative building blocks. In fact, I deleted a couple of much longer draft posts, something I rarely do, since I hate wasting writing effort.
I wrote this post primarily as a way of saying hello to others who might already be thinking along the same lines I am. If you are, chances are the Milo Criterion will spark some productive thinking for you. If not, at least you learned the story of Milo of Croton, for use at cocktail parties.
I will share one more clue: I’ve started calling my developing theory slow marketing. Read into that what you will.
About Venkatesh Rao
The ideas here are appealing to me and certainly mesh well with some of the stuff I’ve been exploring on my own blog. I would ask though, to what degree would those 3 simple ideas spread without the fancy packaging? How did the ideas that they are displacing spread in the first place?
Ideas without packaging basically don’t spread beyond the local idea economy. They are like farmers’ market produce.
Maybe that’s a good thing, and we should leave some good ideas unpackaged. The moment you coin a term and offer a definition, you’ve created a falsehood. It’s the ultimate necessary evil in searching for knowledge I guess.
I’m just now learning about lean startup theory, but one thing that has been in the back of my mind is that if you start with a fairly radical vision of your ideal goal, pivoting according to market discipline could lead you to abandon that vision.
Sometimes that’s good, if you truly discover that the vision is ‘wrong’, but maybe the vision is still good, just very hard to convey to the market, which learns slowly due to its conservative nature.
If you continue to believe in your vision even when the market seems not to be responding to your early efforts, you could bisect instead of pivot: give them something that’s halfway between where they are and where you are, to help them learn a little bit at a time.
If that’s anything like what you’re thinking, then, yeah, it seems like that’s something that might be missing from lean startup theory, but on the other hand, it seems like it might not be too hard to synthesize the two views.
Pivots are among the most problematic concepts in lean startups. They are not really about letting CEOs steer better. They are about letting VCs steer better (this does not apply to angels though). As a caveat, this is a description of some, not all VCs.
The root cause, IMO is traffic/revenue seeking at a certain YOY rate that meets set growth expectations set by those who provide the capital (limited partners). Pivots are a visible and controllable gatekeeping mechanism for VCs to steer a company towards either/both within a fixed time frame during which returns are expected. It’s a case of you-control-what-you-measure. Traffic and cashflow both provide apparently objective and incontrovertible empirical evidence on which to base a valuation (and therefore a quick flip, allowing early investors to cash out before the piper needs to be paid). “Apparently” because the first Internet bust proved that traffic isn’t a good measure of the value of a business, and the current boom, via companies like Groupon, is proving that cashflow isn’t a good measure of fundamental value either. In fact, the only measure of fundamental value is careful thought. Metrics are inputs for careful thinkers, crutches for lazy ones.
Pivots encourage traffic/revenue seeking over building fundamentals by letting the VCs strongly influence direction when the CEO is politically at his/her weakest and least able to resist. CEOs typically want to build fundamental value since that’s the scenario that has the highest payoff for them. For a VC, numbers-and-pivots driven stewardship of investments provides a paper-trail that allows plausible deniability against the charge of bad stewardship, AND a basis for flipping companies based on inflated numbers-based valuations. The poor success rate of acquired properties testifies to that (partly… there are other things going on post-acquisition). It’s a win-win situation for VCs in particular, since, like hedge fund managers, they get both a management fee and an upside share. If they can steer a company to an inflated value exit, they win. If they fail but can convince their limited partners (via the paper trail) that it was a case of bad luck rather than bad stewardship, they win again. To the extent that VCs have a monopoly on providing exposure to entrepreneurial risk in the portfolios of other investors, and all of them do equally badly, they can blame the government, earthquakes or anything else. Since broader-portfolio investors will always want exposure to entrepreneurial risk, it’s a seller’s market. The VCs have all the tools they need to manage expectations.
In a way, this situation is the exactly analog of shareholder-value-maximization dynamics for big companies, where CEOs steer towards (managed) Wall Street share-price expectations rather than business value. Except that here, it is the VCs who put on the expectations-management theater. I blogged about this a couple of weeks ago in my review of Fixing the Game.
If you look at the history of investment in entrepreneurship before the professionalization of the activity via the VC sector, you’ll see a consistent theme: the visionary investors backed their bets through thick and thin, for as long as they had the resources. They looked at the numbers and steered by them, but not towards them. That’s because it was their money and they were taking true win-lose risks, not win-win risks with other people’s money. They enforced cost discipline as a matter of principle, not to maximize exit multiples.
There ARE good VCs today who do (or say they do) operate like this, but I think they are rare, as the returns in the last decade show. Lots of flipping and money made via exits, but bad actual return rates.
As an aside, this critique is mostly informed by my evolving understanding of OODA. A good deal of lean startup theory is sort of derived from OODA. Unfortunately, it mostly throws out the baby and keeps the bathwater.
OODA-theorists make a distinction between “driving up tempo” and “getting inside the tempo of the enemy (or market).” They offer subtle arguments about why the former is misguided and the latter the smart thing to do. Pivots are basically the “driving up tempo” error.
Oh well, I said I wouldn’t post my actual critique, but looks like I’ve gotten sucked into doing so in the comments.
I’m not thoroughly versed in Lean Startup literature yet, but ‘pivots’ imply a sudden or drastic change in direction, which at first pass strongly reminded me of the buttonhook maneuvers that clued John Boyd (fighter pilot period) into the beginnings of his E-M theory. The analogy that originally came to mind is that money in the bank is akin to fuel in a fighter plane, giving a startup the ability to agilely execute sudden maneuvers as required, as in the phrases “pivoting to meet a more promising product-market fit” -> “executing a buttonhook move to get on an enemy pilot’s six.”
However, after reading this post and the comments, it seems that (some? many? most?) VC-funded startups may be losing the initiative to execute radical maneuvers of their own volition as bootstrapped organizations in exchange for a seemingly forced ‘pivot’ in the eternal search for traffic/revenue. In short: investors exert too much undue and detrimental influence on product-development than is necessary.
Maybe I’m taking the metaphors too literally, but Lean Startup pivots are seen to be a “driving up tempo” error only when viewed within the shadow of undue investor influence, as they also tend to denote a positively _reactive_ (i.e. “surprised at the great findings/feedback”) action. One could split hairs all day long attempting to define whether a tight OODA-loop should be termed “reactive” or “proactive” in the context of a near-optimal Milo rate. As you develop your line of thinking in further posts, my hope is that the term ‘pivot’ might be broadened to denote the even more positive association of pivoting on the balls of one’s feet in _anticipation_ of the enemy (or market) tempo.
I think investor influence is part of the reason, but the theory is also set up with just the kind of over-simplification that it encourages a faster-and-faster error. I’ve heard one of the key thought leaders speak and actually use the “iterate faster” language. My interpretation of OODA’s approach to fast transients is somewhat different… the distinction between energy and usable energy (energy adjusted for entropy) actually makes a difference. When you make moves using fuel in the tank, that’s a brute force move. When you manage entropy, PE and KE well, that gets you the OODA type model.
I don’t think Boyd ever connected up E-M theory to OODA and Patterns of Conflict in a rigorous way, because it is hard to define notions like energy, velocity, momentum and altitude for an abstract entity like a startup, where you are using dollars instead of fuel, but I think it can be done in limited ways.
This is admittedly tricky to carefully develop. I think of the “potential energy” of a product as roughly the directions it can easily go, a priori, based on the imagination in your visioning. Engineers vaguely recognize this and use the crude platform vs. killer app distinction to talk about potential vs. kinetic energy, but that’s too crude and static.
Pivots are insufficient conceptually because they don’t capture PE/KE type notions and also because they rely on the iteration process to provide the turning agility. This assumes that the reasons and decisions to change course will come in via (say) sprint meetings, shortly after getting fresh data from outside. I’ve concluded that the iterative structure of the effective s/w development process is not about turning agility in the market but about building reliable, tested, bug-free stuff that is periodically brought back to a usable state. It’s mostly a QA thing, not a marketing steering mechanism. To burden it with that role is to force it to go faster, because as an agility variable, it ends up being a “market sampling” process and you are forced to drive the sampling rate up. In the process, as product manager, you also drive your development team crazy. You also leave less room for refactoring, and trade robustness for steering ability.
It IS useful to speed up the tempo sometimes, for example, in the build up to a major release. This is like the wide receiver giving up on maneuvering and starting to run really fast once he senses a potential straight-line dash to a touchdown. That’s about building momentum in a fixed direction, after a significant feature and direction freeze. User feedback here serves QA and UX refinement purposes; it is rarely a market signal that can tell you how to turn.
The actual turning agility is generally required much further upstream. There, the constraint is not the speed of the iterative process, but PE/KE management. Often at this stage, what you are reacting to is not user feedback but broader environmental feedback from the industry, competitive environment, fronts opening/closing for disruptive attacks etc. At this stage, you are mostly positioning your main pieces.
If you think of pivots as something you do after getting a major round of customer feedback (or more likely, failing to get it), it’s probably already too late.
Most of the turning and maneuvering happens before you fail, when you still have PE left to burn, not after. When you truly fail, you need to build up PE before you can do any more maneuvering. Color.com is in the position of having to do that now.
PE in this sense is not really about money so much as it is about defensible market position in terms of accumulated PR/marketing value. To over-simplify a bit. Your PE is the portion of this value that you can retain if you decide to make a specific maneuver. You can retain a lot or a little of this value depending on how good you are at managing PE/KE. This in turn depends on whether your product has gotten ‘inside’ the tempo of the market, which in turn depends on the interplay of situation awareness and richness of mental models built up through visioning.
Non-bootstrapped external money in a startup is in fact something of a liability, because it buys you time at the expense of demanding expectations.
I am being very conceptual here because I want to avoid talking about specific products, including ones I’ve been involved with. But I’ll add one more note in this general analogy to OODA and E-M: the biggest source of difficulty in startups as opposed to fighter combat is that there is no defined vector defining “down” (unlike gravity in the air). “Down” depends on how you frame things, which is why your PE/KE situation is limited more by the complexity of your mental model than by the variable many people wrongly assume is the equivalent of gravity (money).
Enjoying the interplay of fighter maneuvering w/ startup theory. Every pilot training grad has a simplified version of Boyd’s EM theory drilled into his head by instructors while operating in the 3-d block of airspace called the military operating area (MOA). The basic idea is the constant balancing act between Potential and Kinetic Energy as a function of airspeed, altitude and g-forces. The pivot maneuver, as you rightly point out, requires a large amount of potential energy as it will rapidly be bled off in the form of g-forces. If one attempted a rapid 9-g pull at a low PE state (low alt, airspeed) they would stall and crash. Before going into a pivot (high-g) maneuver one must maneuver into a high PE state either by climbing in altitude or increasing acceleration (burn gas/cash). An expert pilot can make use of altitude and airspeed in such a way as to never be caught in the deadly, unmaneuverable state of low and slow, i.e low PE, low ability to pivot. He does this via planning one or two moves ahead and using the jet’s own momentum to efficiently maneuver in space without having to constantly push up the throttle.
Of course, all of the above doesn’t even get to the issues that come into play once an enemy enters the airspace. Manging one’s PE/KE in isolation is hard enough, but throw an opponent into the pictures and now those skills must already be tacit in order to spend the limited brain bytes orienting to the opponent.
In EricRies’ new book, he defines a Pivot as a change in Strategy, within an unchanged overarching Vision. A Pivot is something you hope not to do, but recognize that the odds are significant – it doesn’t seem tied much to tempo/OODA.
But in a post a couple years back, he used the term as “pivot from one vision to the next”.
Great post. Does your slow marketing theory bare any resemblance to the approach you advocate in this Quora answer? http://www.quora.com/Some-startups-never-get-the-publicity-they-deserve-What-can-be-done-to-obtain-more-exposure-especially-in-the-critical-start-phase
Those were some early exploratory thoughts. I think I’ve gotten more radical now :)
And I thought the idea of starting your marketing two to three years before writing a line of code was already pretty radical!
Been following the blog for a while; first time I have felt moved to comment.
LEAN production ideals are useful in the setting of manufacturing, where process analytics and continuous quality improvement are helpful for decreasing manufacturing defects. The whole ‘avoid waste’ idea is useful, but comes at a cost – resilience and robustness. This is relevant when you consider the overapplication of LEAN and fetishization of statistics to prove things that should never be even remotely considered as applicable to LEAN tasks.
Here’s what the real danger is: over-optimization, or curve-fitting. Traders have long known about this – try applying a fourier transform to the daily close of the S&P500 and then trade off of it. When you blow up (lose all your cash), you’ll realize that the system failed because you assumed periodicity when none was present, and in your zeal to most closely approximate the price activity, you curve-fit so perfectly that small errors in the real world translate to large errors in your model. (Example – the 20 sigma events we saw in 2008 and are seeing again, just a few years later).
So consider your startup model – you are relying on what is (likely) a small customer base and optimizing your product based upon their feedback. However, as you gain more customers and optimize further, ultimately you reach a point where you think you are 100% in sync with your customers expectations and go live. Unfortunately, there has been a shift in sentiment by that point, and revenues are not as expected. Over-optimization.
Compare that with apple’s model – we’re going to give you what you need, even though you don’t know you need it YET – but when you see it, you’ll know you do! A through the moon hit. Not replicatable by LEAN at inception. Thereafter, yes, you can apply it (and they seem to be doing so).
LEAN is a tool. Not a panacea.
It’ll take me a while to process this connection to curve fitting. I think I get what you’re saying, but I’ll have to chew it over some more.
Think about it this way. Each running process has a certain degree of variation that is statistically analyzable. Make 1000 widgets, and 50 are defective. So 95% of widgets are OK and you are running a 2 sigma process (2 standard deviations = 95% of the normal curve, defined by statistics). This is on your standard widget machine which is a non-specific machine that can be retooled, etc… but is slow with multiple non-automated processes. So bring in the LEAN guys and have them go over your process, make changes, and bring you to four sigma (99.99% defect free) by installing very specific, expensive, and single-function machines to make the widgets, automating everything, which means that you also have to re-arrange everything at high cost. So now you have your defect free plant producting widgets at an optimum rate with minimal human intervention and minimal defects. Great.
Except that 6 months later, your primary customer decides he wants fidgets instead of widgets. And you can’t retool, unlike the first (slower) process. So, you’re pretty much out of business. You curve-fit too much; six-sigmaed out your resilience and robustness, and by excessive specialization, made yourself extinct. By the way, those sigmas after the first or second don’t come cheaply ($), either in terms of human effort or sunk costs.
In a world where profusion increases asymptotically, over-specialization, while clearly the way to optimal profit (c.f. above posts with your hedge fund buddy), increases the risk of premature extinction. Resilience and robustness, while initially less profitable, increase odds of survival.
Have I made it clear?
Ah I see. I tend to think of this as Knuth’s “premature optimization is the root of all evil” effect.
I think I am saying something stronger: in this case, I’d say ALL optimization is premature and evil. Optimization involves legible and impoverished finite models, even when it doesn’t succumb to overfitting.
But good point even if you don’t go all the way to my more extreme position.
If the purpose of product development is the creation of a customer, and “a customer is a novel and stable pattern of human behavior” (https://www.ribbonfarm.com/2009/06/15/marketing-innovation-and-the-creation-of-customers/), it certainly makes sense that the rate product development should be tied to the “optimal behavior modification rate for humans.”
I still think you’re overcomplicating things. You just need to make something useful. If it’s useful even just for you, you’ve won.
I disagree, I think The Milo Criterion is the just-so story. One way to examine this is look at completely new users into the system. A recent immigrant from OutwardBackistan is going to initially be confused with airports but it takes, what, 10, 20 trips top before they’re a pro. Similarly, new users are encountering baroque pieces of software like Photoshop all the time and while there’s an initial pain, people eventually get over it.
In reality, it appears that engineering constraints seem to be the primary handbrake on moving too quickly. I’ve heard from people even at Google or Facebook who complain about how lack of engineering resources are hampering them from pursing obvious innovations.
Maybe, but I wonder if you’re failing to distinguish between individual behavior and herd behavior. An individual OutwardBackistanian will adapt to modern air travel pretty quickly due to a combination of peer pressure, social proof, and a lack of alternatives. But a world composed solely of OutwardBackistanians, if presented with modern air travel, might reject it wholesale.
Milo is certainly not a theory of everything, but IMO it’s more than a just-so story.
The acquisition of marginal users is a significantly different problem than acquiring initial users. Marginal users are not questioning the system; they learn the Morris-dance and shuffle along, since everybody else is doing it. Initial users can be attracted to bizarre and baroque behaviour-modifying systems like bicycles, Photoshop and religious dietary restrictions only if they offer an advantage so unique and powerful as to make the negotiation of the steep learning curve and/or discomfort worthwhile, viz. cheap horseless transport which beat the hell out of walking, grafting celeb heads on non-celeb bodies, eternal salvation.
The vast middle ground, where the app has neither a user-base, nor does it offer life beyond death, but still demands significant behaviour modification, is the land of Milo. G Wave is an example of too-fast, and Facebook may be an example of just-right.
http://www.smbc-comics.com/index.php?db=comics&id=2366#comic
Patrick and tubelite have said what I’d have said. I originally had a remark about distinguishing first-wave learners of new features from imitative learners. After all, we manage to compress a usable 18 year learning curve into the K-12 scheme what it took humanity about 5000 years of first-wave learners to figure out.
The limiting factor is the learning rate of the first-wavers.
Complaining about lack of engineering resources to pursue “obvious” innovations isn’t an indicator of anything. Whether it is a prematurely worked out product or one that’s keeping pace with user learning, there are always local optimization holes that scream for more resources. I experienced this personally on Trailmeme (and did the screaming for more resources). Trailmeme was initially going at a faster-than-Milo clip, but fortunately we recognized it early before getting too far ahead of users, and retrofitted Milo-rate features that allowed users to adopt a lot more easily. Still, it was an expensive mistake. If I had to do Trailmeme again, I’d have taken a somewhat different and cheaper route…
Hehe, I had just invented the Croesus Criterion: innovation is limited by the ability of the capitalist to understand innovation.
The (quite understandable) position of venture capitalists faced with a complex idea is usually: we can’t efficiently compute whether this is going to be successful, so we’re going to take the easy way out. Either a. you find some other sucker to do the hard work of analyzing and predicting the trajectory of your product and valuing your company and investing in it, and we’ll be happy to tag along, or b. get your product to beta, and show us beta customer feedback.
Of course, if your product is not capital intensive, then you fall directly into Milo.
You are being too charitable with your Croesus Criterion. Capitalists understand innovation just fine. They just subvert it to their advantage instead of prioritizing Schumpeterian wealth creation for the greater good.
See my response to Josh below for details. The hedge fund guy I mention is a mutual acquaintance.
While I haven’t yet taken the time to discern the areas of thought towhich this pattern may be valid, I enjoy this blog because you (Venkat) have a propensity to connect dots that are far enough apart to make it thought provoking (and interesting, of course) and for that I say “Thanks!”.
As to the individual “dots” in this discussion:
In the presence of competition in a stochastic world, when a product-developer does not develop the product at a rate faster than the AVERAGE customer can use, he/she/it risks (potentially permanently) losing the most profitable segment of the market to his/her/its competition – which is not good for generating the kind of revenues needed to generate/support competitive advantage (in the conventional sense). When working outside of or in front of the emergence of a competitive enviroment, Milo applies. In a highly competitive environment, probably not.
Lean in the conventional sense is the diametric opposite of agile. Lean in the Christensen sense – fail early and often (and thus most inexpensively) in order to get to the customer sweet-spot quickly and with enough money left over to make the product once you’ve gotten there – may well be the key to entrant survival.
Marketing was the only productive element that Drucker ever called out as being critical. I’m pretty sure that the maxim “If you build it they will come” is pretty much BS. “Slow marketing” may actually be the key to the least expensive way to fail early and often, but care must be taken not to give away your intellectual capital (which in the instance of re-combinations of existing technology, may be emminently copy-able).
Thanks again for sharing thoughts which exercise the underused part of many minds,
Rick
http://www.linkedin.com/in/decisionscience
Excellent point. The existence of a competitive environment of above-Milo-rate players can force you to go hyper-Milo yourself. The end result is a race where society as a whole is worse off, but the losses don’t show up in any account books, because they are in the form of social costs. This is why ethics comes into the discussion as well.
I believe there are barriers to entry that slow marketing operations can build, but to be frank, not every operation will pull it off.
I am not sure lean and agile are opposed though. It simple depends on the level at which you do your cost accounting. The two share the same blind spots, so can be made consistent.
I have wondered if a product tends to carry a band of customers, a goldilocks generation, for whom the product has always been just right. Other people who started using the product too late find it arcane and if they have a choice will prefer the newer simpler alternative.
Someone once argued that as services become better designed they tend to evaporate customers, because as they focus more and more on being perfect for their target market, they will create a more and more specific niche for themselves.
In addition, not only will you build a slowly better model of “your market” which will do this, the more feedback is based on existing customers, the more it will suggest increased complexity. They will be able to handle it, having already picked up the basics. Others may not.
Possible solutions?
Multifactor marketing models that let different clusters appear, multiple differentiated sub-markets.
Basing feature improvements or additions on metaphors taken from existing functions, then making simple free products that operate purely according to these metaphors (teaching them “your way”).
Just make a neighbour product for the people you’re loosing, assuming you have the right feedback to know why they left.
Follow your particular customers through their changing life patterns with new products in order to maximise the value of your increasingly personal knowledge of them.
etc
It would be interesting to see if anybody has thought about applying the idea of demographic cohorts/waves to specific product or brand customer bases. The idea of hyper-segmentation and clustering based on absolute attributes (age, sex, postal code, salary range, etc) is well-known and commonsense, but one possible line of exploration of the Milo Criterion could be the idea of including the on-boarding timestamp as a means to determine customer cohorts.
For a complex product, this would extend the dimension to more closely/stringently result in efforts to classify where in the product evolution/learning curve particular customers lie. When it comes to profitability and repeat business, perhaps familiarity with the product/service is a more important determining and sorting factor to use for profitability and attention than other metrics.
“…I am not going to elaborate further (at least not right now)…”
Ha, I see you’re applying the Milo Criterion to this series of posts!
:)
I take it you don’t like the new facebook layout Venkat.
I’m not sure I buy your idea yet, but over the next 2-3 years I’m sure you’ll try to build up enough evidence to convince me otherwise.
In fact while your doing that you might share information with a few potential allies so that when you do try to promote the idea properly, you’ll be accompanied by a few different manifestations of the idea each focused on a different specific area, and all generally promoting the theme of slow marketing.
I take it from the name that you’re considering spinning it off an idea in an area not likely to end up in direct competition with slow marketing; that of slow food. You know, to reduce unfamiliarity, and let people know how to approach it.
Maybe it’ll work, but you’ll need some magic to match lean’s pithy statements and obvious applicability.
Yes, I am referencing slow food. But there are some very unrealistic elements of that. I am headed in a much more pragmatic direction I think.
But the main reason I am not too interested in promoting this idea is that it is destined to remain a tiny-minority model until there is is significant reform in the capital markets and Wall Street.
A lot of what goes wrong with the lean startup world is due to its strong coupling with Wall Street (both directly, and indirectly, via the VC firms and the acquisition market dominated by Wall-Street-listed companies, as the two bookends of the startup pipeline).
In fact, my Aha! moment came a week ago, talking to a friend of mine, a veteran hedge fund guy who is also involved in a startup and is familiar with lean startups. His remark hit the nail on the head: lean startups are great for investors, not for entrepreneurs. The ideas enhance legibility for investors at the expense of survivability for the business. Investors can increase their expected returns even while lowering survival probability. The theory rationalizes VC interest as business interest. VCs win overall, entrepreneurs are worse off. There is a sort of Gervais-Principle-for-startups at work here (with investors as sociopaths, entrepreneurs as the clueless). Okay, I may have said too much :)
So in a way, slow marketing requires slow money (i.e. not traditional capital markets money). Slow money is currently joined at the hip to slow food, and I don’t buy the idea in its current form. Currently, the only source of slow money is retained earnings, which means only bootstrapped companies can attempt slow marketing.
If slow money gets generalized more, slow marketing will be more broadly applicable.
The majority of VCs have always engaged in founder-hostile behavior. It’s worth alerting VC-dependent founders to the risk (to their job and vision) of adopting Lean.
(Though at least the Lean framing gives a *legible/shared* set of game-rules, giving founders better warning signs for when their job/vision is at-risk.)
Personally, I’m more interested in applying Lean/CustomerDevelopment tactics in the OneManShow environment. I believe there are a lot of people doing that same thing. The lack of OPM and need for food/shelter demands a shorter time-horizon of feedback/progress (at least for 1 of your bets). I have one project I’m getting underway that will take a Lean approach, and another that will probably be Slower.
I agree that Lean-constrained startups are less likely to create big WorldChanging things. But the odds of that outcome for *any* venture are so low that I’m not sure that’s a relevant criterion for anything…
I’m with you Venkat. We need to slow down. The tortoise wins the race, right?
I just found your website today through Quora, and I’ll be spending more time here after some of the interesting things I’ve read.
Anyway, I wanted to comment on how the Milo Criterion reminds me of Clayton Christensen’s disruption and particularly the graphs he uses to illustrate the reason it happens. He graphs product improvement against the customer’s ability to accept the new improvements (in this case, the Milo criterion). The product improvement always moves upwards faster than the customer’s ability, and at some point, the new improvements not only become irrelevant to all but the most discerning of customers and even hinder the utility of some of the other customers. He hypothesizes that it is because of this oversight that “inferior” products are able to move in and slowly disrupt the established company.
One of your comments indicated to me that the current capital markets and VC structures help promote this situation, and I was wondering that maybe the basic corporate structure undermines the pursuit of the Milo Criterion on many of the larger companies as well.
Very interesting ideas. I look forward to hearing more.
Short answer: yes. But the problem is broadly recognized in big companies. The startup sector has this self perception of immunity from such “evil big company” dynamics.
Startup culture has distilled some of these sentiments into a pithy saying:
“Early is the same as wrong.”
I’m probably not that smart, and the language barrier dumbers me further, but I couldn’t get your idea for some time. Now I have some model of it in my mind, but I’m really not sure about it.
Are you saying that:
1. if you’re going too fast, you’ll lost the customer base (consisting of averages)?
2. The lean methodology leads us into temptation of slowing down and taking tactical, maximizing revenue or some other number approach
3. But lean methodology, being introduced as the apple of knowledge, leads only to momentary knowledge, thus leading to falling into stupidity (in the sense of avoiding further truth-seeking)
4. So, the idea is – work slowly, but don’t use the advantage of slow speed to turn too much?
It seems that Apple doing something like this
“The primary reason these behaviors are effective is that they slow down the process of software development and maintain the optimal behavior modification rate for humans.”
I disagree – I would say these behaviours are effective because they provide a quick and effective diagnosis of that common entrepreneurial disease, *building something nobody wants*. That, to me, is the core idea of lean startup theory, and I think it’s essentially valid. I think more products fail because they are fundamentally *meh* than because they are over-complex. It seems to me like you are trying to reduce everything to a problem of user interface design, though I’m sure later posts will enlighten us more :)
I know quite a few people who have been working away on some secret web project for a few years, and I often find myself trying to persuade them to just launch the damn thing already. They usually counter by saying they don’t want to drive everyone away by releasing a crappy product. Probably most readers of this blog have been involved in similar conversations. I actually don’t think the people I argue with are *wrong* – I think this dispute represents a fundamental personality split amongst entrepreneurial types – “opportunists” vs “perfectionists”. Opportunists are unlikely to waste years developing useless products. Perfectionists are unlikely to fall into Fordian “faster horse” traps or waste years flitting from project to project. Opportunist strategies work well when the cost of failure is low, and perfectionist strategies work well when it is high.
Other thoughts:
– exactly how latecomers to the market become educated to the use of a complex product (like the OutwardsBackistani from above visiting the airport for the first time) would be an interesting avenue of thought. You’ve already identified the role motivation plays in getting early adopters signed on. With latecomers, there may be a combination of increased motivation from seeing everyone else using the cool new thing, and reduced complexity from the presence of friendly guides.
– interesting use of Google Wave as an example. Google Wave was a fairly spectacular flop because it was made by Google. Had it come from an unknown startup, it would likely have been an unnoticed flop.
– I saw an interesting picture in a magazine today that might be a good illustration of the Milo curve – it was for John Lewis (a British home supplies retailer) and showed how their mixing utensils had evolved over time from a simple whisk to a complicated blender with multiple programmable settings. It reminded me of the observation that customers often like to buy more complexity than is actually useful.
“These theories … are mostly a set of just-so explanations that serve to motivate practically effective behaviors, the way religions motivate moral behavior.”
I was on a mailing list for practitioners of lean methodology at some point in the past. What caused me to unsubscribe was a thread in which the OP described his successful business — already profitable and doing $4M in revenue — and asked if his business could be described as lean. The response from one of the most frequent commenters was that unless X, Y or Z criteria were met, the OP was better off folding the business (a profitable venture already!) and start anew.
That’s a critique of the community more than the method, but somebody once gave advice about judging a tree by its fruit…
This sounds like link-bait to me. Vague denunciation of the dominant paradigm with vague promises of salvation.
There is a nugget of truth about developing a core cadre of users who both validate the ideas of a new product and serve as educators/examples for the second and third waves of adopters.
But that seems entirely orthogonal to lean startup techniques used to direct and prioritize product development.
With the examples of Twitter and Google Wave, both were strange new technology ideas but Twitter had an active early user base demonstrating the value and usage. Google wave provided a pallet of tools for things no one understood or needed, because that initial user base and community was missing or not publicized.
I like the example you give of google Wave vs Twitter.
Wave lacked clearly defined the benefits in fact it was a third party who wrote and published a book on the benefits.
Twitter uses emerged with time from their simple system (such as real-time event coverage) and they adapted their service to showcase those benefits.
As such wave couldn’t educate their users, Twitter did.
Google seems to have learned from it as the public rollout of Google+ had a lot more focus on hangouts.
There’s a reason Isaac Newton said “If I have seen further it is only by standing on the shoulders of giants.”. Maybe there is one assumption that is wrong — when Venkat talked about user adoption he didn’t specify how, in fact it seems that users can help be adopt “artificially” by educating them.
And that kind of “artifical” adoption can be seen at work everywhere where people try to change behaviors (see positive deviance theory)
I like the framing “slow marketing”; but I don’t agree that this is a new insight in this Venkat; merely putting a name something that is already known (you are in fact making this more visible).
Matthew E. May in “The Elegant Solution” (about the Toyota Production System) said something quite similar — that a successful complex system always start as simple system that has evolved overtime.
In particular I see this everyday in the Medical Tech domain (think eHR etc…) that vast majority of those system started as basic tools and evolved overtime to become complex.
In the end this all comes back to System Theory, start simple to satisfy the minimal need of your customer and enhance your product/service over time and it will succeed.
Now with all that said I had a talk about “gamification” recently (again this is merely renaming a phenomena used over and over again) .
The interesting element here was the effort put into customizing the communication (both active via email/phone and passive with UI/process changes) with the user to make the user successful. They used “gamification” to educate their users (and educated users are more likely to remain loyal and trust you as they put a value for your expertise).
Not fully onboard yet, but intrigued. Examples abound of products that were ahead of their time and failed. Everett Rogers’ work on diffusion is inline with this idea. His argument is that products that diffuse faster are ones who do well on: complexity, compatibility, triability, observability, and relative advantage.
This is why, I suspect, your book struggles against your blog. In order to adequately explore concepts in Tempo, it assumes a slow build-up of a whole set of core concepts (embodied minds, legibility, systems, feedback loops, etc) that have been covered on this blog for sometime. The response I’ve seen from people not sufficiently primed for Tempo has been being simply overwhelmed.
Have you thought about pumping out another book which is more of a “best of” compilation of the blog which introduces different ideas in a progressive order such that it is more legible? Might be good for existing readers but great as giving as a gift.
Hell, I’m open to curating it.
Working on something sort of like that.
Actually the book seems to work unexpectedly well stand-alone, at least for a significant subset. There appear to be a few other approaches to accessing it besides my blog (such as prior exposure to military thinking or literary narrative theory). There are other ways people are primed for Tempo in other words.
I am surprised fairly frequently by new book readers who get stuff in apparently “inner” layers that I’d put in with blog readers in mind.
After building my prototype and after using it to approach customers, I found that they fell more on a discrete spectrum rather than bunched up w.r.t their needs.
All of them are in the same domain: Adventure Sports travel. All of them effectively offer the same product but their needs are different because of the state at which they were comfortable with technology. Some just required a website (prospects with crappy websites), some required a CRM and some wanted a JS widget they could use on top of their own very well built solution. The customer with a crappy website would feel any features above the ability to create a new website were actually negative to the overall product.
The whole Lean Startup movement seems to be an extreme response to the “product waterfall” model that used to be common place entrepreneur thinking. I dont see it as a theory that will benefit all entrepreneur … even just web entrepreneurs. It has one big benefit – it makes entrepreneurs interact with customers instead of sitting away in a corner building something that may not be needed by anyone.
I would love to hear more about the theories that you’ve been working on. IS your “Milo Criterion” specific to limiting the solution to a level graspable by the customer or is it applicable to marketing, prospecting clients & other parts of starting a company as well?
Just made a connection between the ideas here and a book I read several years ago. In Christopher Alexander’s book “Notes on the Synthesis of Form” he spends quite a bit of time making the same point as you–that the truly successful designs are those that are created piecemeal instead of all at once. The form and the context slowly dancing together.
Fascinating book and very relevant to this discussion.
Copycats may be as effective as they are in part because they share customer development.
I believe the principle underlying lean startup methodologies is the best way so far to make certain that your “calf” is the right weight, at the right time, for the right calf lifter (customer). Said otherwise, I believe the lean models are the best way to regulate your development so that you don’t get ahead of your customer — so that you can maintain pace with their development. In other words, lean models are the best way to deliver upon your “milo” concept.
If you pivot to early, or lose your vision otherwise, that’s not a fault of the model, but rather a poor decision on the part of the entrepreneur. Really, it’s a question of measuring and responding to the right things.