We had one of our many (non-stop) email conversations among the USV crew last week about a situation in our portfolio where nobody could agree on something. I lamented that “VCs are such optimizers.” It takes one to know one you see.

Nick Grossman replied that he prefers satisficing to optimizing. I had never heard that term. So Nick sent me to Wikipedia which says:

Satisficing is a decision-making strategy or cognitive heuristic that entails searching through the available alternatives until an acceptability threshold is met.[1] This is contrasted with optimal decision making, an approach that specifically attempts to find the best alternative available. The term satisficing, a portmanteau of satisfy and suffice,[2] was introduced by Herbert A. Simon in 1956,[3] although the concept “was first posited in Administrative Behavior, published in 1947.

I love the concept of satificing instead of optimizing. It is something I have been trying to adopt (changing behavior is hard) for close to twenty years now with a good measure of success. But I never had a word for it. I do now. Thanks Nick!

#life lessons#VC & Technology

Comments (Archived):

  1. William Mougayar

    Sounds like a good consensus strategy. For added context, what outcome were you trying to optimize, and how did you communicate it to the startup (in case you did)?

    1. fredwilson

      can’t go there

      1. William Mougayar

        Ah, a satisficied answer!

        1. LE

          Beating around the bush like a politician probably would have been a satisficed answer. Fred answered it directly saying he wasn’t going to answer the question because he can’t. That’s like taking the fifth at least you know where you stand.

          1. William Mougayar

            Did you know you’ve already mentioned that word in the comments last year, here:

          2. LE

            How did you find that? I’m not seeing an easy way to do that (since gawker is gone). Google search doesn’t do that. Is that an mod perk?

          3. William Mougayar

            Nope. I searched my Gmail, since I get email notifications on all comments. If you searched yours, it’s probably there.

  2. christianbusch

    Barry Schwartz wrote about this in depth in The Paradox of Choice (… – great book that highlights the dilemma and options around it. I’ve been trying to become a satisficer for a while as well – no major success so far though.. Highly recommend the book though.

    1. fredwilson

      thanks. i will add it to Kindle.

      1. bsoist

        Excellent book! cc @christianbusch

    2. SubstrateUndertow

      His Ted Talk on the book with some very good mobile phone humour !It is a very funny talk !



    4. sachmo

      I didn’t like it at all. It was boring and repetitive. It was written by a satisficer : )

  3. Avi Deitcher

    Like, very much. Especially useful for over-analysers. We could have used this in B-school; it took a long time for our team to get past “paralysis by analysis” (and specifically the least hyper-intellectual pushing us to get it). Actually, he had another term for it, not appropriate for public fora…

    1. William Mougayar

      Yep. Analysis has many levels. You could agree on some, but not all of them.

      1. Avi Deitcher

        I am not completely sure. Could we spend the next hour analysing it? 🙂

        1. bsoist

          I know I could 🙂

  4. JimHirshfield

    Good enough?

    1. JimHirshfield

      Great is the enemy of good?

    2. fredwilson

      pretty much

    3. jason wright

      exactly. the word contradicts its principle.

      1. JimHirshfield

        I’m not sure there’s a contradiction.

        1. jason wright

          edit.”good enough” is satisficing. “satisficing” is optimizing.

    4. leapy

      Example: the updated Disqus Privacy Policy page – good enough but it would have been more useful to highlight what had changed since the last policy.

      1. JimHirshfield

        I believe we have been very transparant about the changes, with notifications sent to every registered Disqus user. And a comprehensive article about the changes:

        1. leapy

          I agree about the notification. That’s how I found the page.The article states what is included but does not summarise changes vs previous policy.

          1. LE

            The article states what is included but does not summarise changes vs previous policy.Let me ask you this. How many people do you think really care about that? My gut feeling is practically nobody.

          2. JimHirshfield

            A fair criticism. My understanding is that the sections we point to in that article, are the sections that were changed. We’re not trying to “hide the ball” – these changes relate to our ads business and related user data/tracking. And there’s a link there to opt-out.I really respect those services that bullet point list the things they’ve changed. Most that I’ve seen, however, are still fluffy and the only way to know for certain is to do an MS Word file compare. And I haven’t seen any service go to that extent. But it would be nice to see the “market” move in that direction. (my personal opinion).

  5. JimHirshfield

    Also sounds like = satisfy + sacrifice:-/

    1. LE

      I tend to think of sitcom humor as satisficing. It’s really the low hanging fruit of comedy to me. Kind of like most of the “jokes” that Carson used to tell at the desk such as sitting across from Zsa Zsa holding a cat and saying “may I pet your pussy”.

      1. JimHirshfield

        I think that “joke” of his triggered a law suit by her…or at least a huge uproar. So…low hanging, but heavy implications.

        1. LE

          Actually, this is what happened it wasn’t even a good example:…Go figure that on something I didn’t check on before opening my mouth I’d be wrong about.

    2. SubstrateUndertow

      Yes, a good portmanteau like a good joke should not need to be explained, its blended meaning should be perceptually obvious.As a portmanteau meant to blend the meaning of an “available+sufficient” solution, “satisficing” is truly self-referential. It is an available portmanteau solution but so sub-optima your left pondering or looking up its implied meaning.Or as you’ve done here derived an alternate, and possibly a more immediately interruptive, implication for that portmanteau’s meaning.Fred’s implied context here cries out for an effective portmanteau to describe a decision style that bends “available+sufficient+comprise”.A decision style that blends “available+sufficient+comprise” is certainly prominent enough in its real world significance to warrant its own word/portmanteau but “satisficing” offers barely “satisficing” perceptual functionality.In short ! (shocking I know)I love that portmanteau’s blending intent but hate its barely “satisficing” blending solution and to add executional insult to perceptual injury it doesn’t exactly roll off the tongue !

    3. ShanaC

      that is something that is happening when you satifice.

  6. LIAD

    We rarely achieve our optimum outcome.===> Life is a (sub-conscious) lesson in satisficing. (aka compromise)========> Maturity is the ability to accept it. With grace.——————-we have many other phrases for this.good enough is good enoughdone is better than perfect.satisficing perhaps works better as it gives a positive spin and alludes to achievement and success rather than a lack or loss.—————-Q. Net net, do we lead richer more fulfilled and impactful lives by ardently optimizing or satisficing?

    1. sachmo

      The perfect is the enemy of the good – Voltaire.

  7. jason wright

    a common sense human behaviour so not in need of being labelled with a word like that.

    1. LE

      so not in need of being labelled with a word like that.Maybe not the case. Words and analogies (which I use very frequently) are great shortcuts for communicating concepts and ideas in shorthand. They are time savers. That’s why it’s easier to have conversations with people that you know and/or have common interests. You don’t have to spend time explaining and ramping up to make your point.Think of Kid Mercury saying “pulled a Suster”. If you know what that means (because you’ve heard it here) you automatically understand the concept and what the Kid [1] is trying to convey. Especially with analog ideas which are loosely defined. Like pulling a Suster. (There is no measurement for “to long” but like Porn you know it when you see it”). [2][1] Where has he gone btw?[2] Which is another shorthand helpful concept courtesy of Scotus.

  8. awaldstein

    Behavior feels like me all over. The word is brand new though.Not an analyzer by nature and honestly if you work in startups for long enough and have closed the door or taken a walk to come out or back with a decision on action based on collective guts, you satisfice all the time.

  9. jason wright

    so how was the T-Mobile retail experience?

    1. fredwilson

      sucked. their store systems were down. i went back to my office and ordered it on t-mobile’s website. i will get it in two-three weeks

      1. jason wright

        oh, well as an online purchase you can now return it. that’s how it works here.

      2. jason wright

        did you see this? so good it came out twice.japanamericaaustraliaukfrancegermanychinaitalyspain

      3. William Mougayar

        Whaaat!! How about from the Apple Store, then go to T-Mobile to switch the SIM?

      4. Joe Biden

        VZW is solid; I’ve been a customer for over 12 years. Just dealt w/ CEO’s office on an FCC E911 issue that was traced and resolved very fast (a particular LEO dept. had a software issue).

    2. Donna Brewington White

      probably not very satisficing is my guess

      1. jason wright


  10. Tom Labus

    Good one. Great job @nickgrossman

  11. Donna Brewington White

    I think it would be easier to adopt the concept than it would be to adopt the word.

    1. William Mougayar

      Try saying it 10 times in a row & with a straight face 😉

      1. Donna Brewington White

        Or even one time … with a straight face. 😉

  12. pointsnfigures

    as long as it isn’t a massive decision in the short run it’s probably optimal.

  13. John Pepper

    Watch out Words with Friends foes. I have a new word.

  14. Raghu Hariharan

    In the class I took with him, Adam Grant in my first year of b-schools, discusses the differences between maximizers and satisficers. He also discusses the same in his Wharton ’11 remarks:

  15. Kasi Viswanathan Agilandam

    Stapled words … they sound so nonvincing (not so convincing).

  16. Kasi Viswanathan Agilandam

    I still remember my 6-grade episode … one of my class met could not recall the word “Optimized”. Though we were all studying in English medium school …English is still a foreign language for us.He wrote “maximumly minimized and minimamy maximized” …our teacher appreciated his thinking and awarded full mark for that question and encouraged students to always think of alternative way of expression.

  17. William Mougayar

    I have a vague recollection that we have already discussed that word here before. Where is Comments Search when we need it.

    1. JimHirshfield

      #sadface #comingsoon

      1. William Mougayar

        Pretty please with a cherry on top 😉

  18. Joe Biden

    It’s easier, psychologically, for many people to bikeshed on process and irrelevant details than tackle the complex reality of steering a course through the abundant grey with hard & smart work.

  19. Maxwell Wessel

    Interesting take – So the real issue is getting to alignment on what is a definition of adequate. Forces you to be more thoughtful up front than most.

  20. LE

    I do this all the time (and refer to it as satisficing) but I think the wikipedia definition overly complicates the concept.Satisficing in practice might be as simple as deciding to book a hotel in NYC and just going with a Marriott or Four Seasons because you think they will be good enough rather than taking the time to read all the tripadvisor reviews to confirm that you are correct. It might be sticking with a particular brand because in the past that brand has hit the mark most if not all of the time. [1] Or maybe returning to rather than seeing if there are any other alternatives. Or going with the last guy who did tree clearing for you because he did a decent job and you don’t want to “re-invent the wheel”. [2]And I don’t think that it implies a compromise either in any way. I think it’s a practical solution to not having enough time and potential diminishing returns.The definition of the word may be one thing but in practice the concept is much broader. By that I mean that “searching through the available alternatives until an acceptability threshold is met” doesn’t always mean you are actually searching a list of alternatives.[1] Key point (I say) and why brands are so powerful.[2] I’d love to hear the paragraph that the academics use to describe that concept.

  21. Richard

    Interesting post. Optimizing is often not the right solution and particularly, when done without a processor optimizing is highly inefficient.Take the case of purchasing x. People often spend hours trying to optimize ten event bases on minimizing the purchase price.A much more efficient approach is to set a price below which you will make the purchase (period).

  22. Simnett

    General Patten used to say that (paraphrased) a reasonable plan, executed effectively today, is much better than a perfect plan some time in the future.

    1. John Draghi

      I love it. It feels like far too often the pursuit of the optimal ultimately ends in inaction.

  23. jkmacr

    As a maximizer in need of help, I found this post quite helpful.…It drives home the point that we need to make the best of our choices, not strive for the best choices. This helps, but I still struggle with big choices.

  24. Zooko

    Coincidentally, I just used this concept, word, and wikipedia article to help close some multi-party negotiation about partnering to launch a new business. Which, by the way, I’m eager to tell you about, Fred. 🙂

    1. fredwilson

      pls send me an email Zooko

  25. hypermark

    Another variation of this concept is the notion of a BATNA – best alternative to no agreement.This is a construct whereby all parties identify not just their walk away point, but what the “best alternative” is in the walk away case. The classic case is buyer and seller, where each party has the supposed best and final offer, but really hasn’t done the honest exercise of determining what the real alternative is in such a case.BATNA is a great mechanism for getting beyond the ideal and optimal, and getting anchored on the real.

  26. JLM

    The ability to make good decisions is the critical skill of any business executive. Most decisions are a challenge to properly frame and evaluate the alternatives. I am appalled at how poorly many even successful executives are at employing an orderly decision making methodology.The quality of the information and the depth of analysis are often the critical elements in a good decision making process.Dwight Eisenhower is a perfect example of a man who was not gifted with great intellectual powers but had an extraordinarily effective decision making process. He worked for Marshal, MacArthur and Fox Conner–arguably the smartest General officer in the history of the Army.When he was running WWII in Europe he was able to balance the fight, the Allies and Churchill/Roosevelt while planning the most complex military undertaking in the history of warfare (Bradley was the chief planner. He never gets enough credit for planning the landings in N Africa, Sicily, Italy and Normandy.)As President, Ike was a fabulous decision maker. He kept us out of war–extraordinary because it was his wheel house. He balanced eight straight budgets. He built the American nuclear arsenal. He envisioned and began the construction of the interstate highway system (based on the German autobahn which allowed the Germans to move troops long distances quickly). Oh, he never really had a Republican Congress and had to negotiate every single policy initiative he ever undertook. He could get stuff done.He was notorious for demanding “complete staff work” which became the mantra in the military for providing the commander with all the alternatives, a complete discussion of the pros/cons and a recommendation with logic for the recommendation.When his staff was hitting on all cylinders, all Ike had to do was to nod and say: “Make it so.”It was the quality of the staff work and the completed recommendations that made his system work. Today many decisions are not completely framed and what should be a “decision meeting” turns into a roundtable discussion on the alternatives themselves. This is incomplete staff work and would have won one an ass chewing from Ike.The guy was a 5-star General, ran WWII in Europe, President of Columbia, NATO commander and President in no small part because he understood how to make decisions.I would say the greatest weakness of Pres Obama is his inability to frame and make decisions. The revelation today that he attends only 40% of the PDBs and supposedly reads the intel shows how flawed his decision making truly is. What is more important than the PDB? Golf?His lack of executive experience is fatal and shows up in such things as telling us he didn’t know that ISIS was a growing threat. Hello? They began to drive on Fallujah in early Dec 2013. They captured it on 4 Jan 2014. And he didn’t know ISIS was on the move?JLM.

    1. LE

      perfect example of a man who was not gifted with great intellectual powersWhat’s the definition in your mind of “great intellectual powers” then?How can someone become a five star general (in the US ARMY no less) and President of the United states and not possess “great intellectual powers”?I think you are equating intellectual powers with “doing well in the traditional school system”. By that measure Obama could be judged in a positive light. [1][1] No matter what my feelings are for Obama (even given the luck of the Oprah blessing that put him in office) I think anyone who is able to rise to that level gets the credit they deserve as having great intellectual powers even if they make mistakes as everyone does.

    2. sigmaalgebra

      My view on Obama and “decision making”:(1) Please base of supporters, remembering that those people don’t pay much attention to issues, want a more generous welfare state, have no use for foreign policy, and are easily satisfied by just headline level remarks.(2) To please the supporters, toss out a lot of headlines, photo ops, positions, and support of pie in the sky programs proposed by others where Obama knows in advance that all of this will flop soon or shortly after he leaves office.(3) Avoid being blamed. So, “vote present” and otherwise do as little as possible. Wait until nearly everyone else is ready for some action, and then have others mostly be responsible. In case of any blame or criticism, have a ready made list of others to blame and change the headlines.(4) For decision making for the good of the US, that’s too difficult to do well and mostly just a way to be open to a lot of criticism, so f’get about it and, instead, concentrate on jump shot and golf game.Obama is giving a lesson on how to be president without really trying.It’s what a lot of the people who elected him like.For Obama’s foreign policy, I have to believe that it is based on total junk thinking. If we really need some competent foreign policy, then we will be in a lot of trouble.E.g., I’ve recently watched Tora, Tora, Tora, essentially a documentary on the attack on Pearl Harbor. So, (1) a country A did some invasion of country B for economic and other reasons, and we didn’t like the invasion. (2) We replied with economic sanctions on country A. (3) Due to the economic sanctions, country A made some efforts to negotiate. (4) We didn’t much respond to the negotiations. (5) Country A felt seriously threatened and concluded that they had only a military solution.First Example: Country A = Japan. Country B = ChinaSecond, Maybe an Example to Be: Country A = Russia. Country B = East Ukraine.But on ISIS, incompetent or not, just playing with headlines or not, as much as I don’t like Obama, and I don’t, I find his policy on ISIS so far okay:Why? My explanation is that the Sunnis in Syria and Iraq were getting beaten up by the Shiites. So, there were a lot of totally pissed off Sunnis. So, there was fertile ground for ISIS. And supposedly ISIS got some funding from the Sunni Saudis. And they have gotten a surprisingly large flow of funds from oil in the Sunni areas and various criminal activities.ISIS nasty people? Definitely. Are they about to have a constitutional government, free and honest elections, freedom of religion, speech, assembly, and the press, a strong, modern legal system, a government of laws and not men, free public education, equality of women, freedom to learn about Darwin, e.g., move forward a few hundred years and catch up with the US and Europe as of, say, 1800 or 1900? Nope. Are they a shitpit? Yup. Contemptible, disgusting, inhumane, a hellish scourge on earth, to be hated? Yup. Would they like to be a threat to the US? Yup. Are they much of a threat to the US? Nope. If they try to harm the US, should we be able to detect and stop them? Yup. Am I disgusted? Yup. Worried? Nope.But so far ISIS has made progress essentially only in Sunni areas. How? If you were a Shiite soldier in a Sunni area and supporting the government in Syria or Iraq, how long would you want to stay in a Sunni area? You’d take the first, or even no, opportunity to get back to Shiite ground or just anywhere else. And they did — right away. No big surprise.Is ISIS a threat outside of Sunni areas? Maybe not or not much. With good intel, the US should be able to track the situation, and I am confident, no matter what Obama said, that they are.Should we bomb some of ISIS? To keep them in just Sunni areas, maybe.Can we defeat ISIS? Likely that would be about the same as defeating all the Sunnis or occupying and governing all the Sunni areas, and those are not very good solutions.Do we really need to defeat ISIS? If we can keep them in Sunni areas, mostly not.What will come of ISIS? I suspect that ISIS will calm down, and the Sunni areas will revert to tribal culture.Obama’s foreign policy on ISIS? No doubt brain dead, irresponsible, and driven by headlines, etc. Otherwise, motivations aside, for the actual policies, so far okay with me and much better than what I saw from W/Cheney.Or, for a mere $3 T of US money and several thousand US lives, W/Cheney successfully replaced a Sunni secular dictatorship in Baghdad that beat up on the Shiites and Kurds and fought with Iran with a Shiite dictatorship in Baghdad that beats up on the Sunnis and Kurds, created fertile ground for ISIS, and is allied with Iran. And even for our $3 T, we still didn’t get the oil.To me, so far Obama’s Mideast foreign policy is about $3 T and several thousand US lives ahead of what W/Cheney did and with so far no significantly worse results.

    3. sachmo

      I was reading the work of Alex Osterwalder recently and came across an interesting thought.Often in business school – people are taught that decision-making is the critical skill that can make or break an outcome. That the hard thing to do is make a choice between Alternatives A, B, and C – or so the logic goes.But in Design school – what they teach you is to not accept Alternatives A, B, and C. They teach you instead to make NEW alternatives E, F G, H and I. And once you are done making those new alternatives and realizing they suck, throwing them out and making alternatives J, K, L, M, N, O and P. And then refining those concepts down to Q, R, T, U, V. And then refining those again to X, Y, Z and THEN you make the decision. In design school they argue that making a decision is the easy part. The hard part is creating new alternatives.To have the mindset of not accepting the status quo, of being willing to expend the effort to make new alternatives, I think you have to lean towards being an optimizer and not a satisficer – to link the thread back to Fred’s original post.By the way – regarding situation in Iraq – I do happen to agree with you on one thing. It’s amazing that Fallujah was not an alarm bell. US intelligence should have been on the ball long before Mosul fell. I think Obama is a step behind here, but I don’t think that makes him a bad leader. There really are no good options right now, and it’s a struggle to create new ones.I don’t think it makes sense for americans to put troops on the ground. Air strikes alone aren’t enough to deal with ISIS. There’s not a single ‘ally’ in the region without their own ulterior motives, or who aren’t secretly supporting some faction against another one of our allies. Really we should have simply never have gone into Iraq for a 2nd time. It was a horrible mistake 10 years ago and we’re still paying the price.

  27. sigmaalgebra

    Yes, at times optimizing can be from okay up to good or much better, e.g., when actually can optimize and the benefits are significant, but this is not nearly all cases.Roughly since Simon, there has been a lot of work, especially academic research, on how to optimize. E.g., in investing we have the work of Markowitz, Sharpe, Black-Scholes, and, if look carefully, e.g., Markov processes and potential theory (one way to look at exotic options), more.But some of what happened was not good: People used optimal as a stick to beat on others in some dysfunctional organizational competition toward some poorly considered gains. At times, not being optimal was regarded as a failing, transgression, something like a sin no one wanted to be accused of.So, it became important, for saving face, avoiding criticism, winning in organizational competition with people “down the hall”, to be able to claim that had made/saved the last tiny fraction of the last penny, even if saving the last penny was some horrendously difficult, time-consuming, and expensive work, which for plenty of practical problems it easily can be.This mistake has become a big thing: E.g., at one time Bell Labs wanted to know how to design communications networks at least cost. Not necessarily a bad goal: Later as the Internet was growing quickly. T. Magnanti, a Dean at MIT, gave a Goldman lecture at Johns Hopkins on the problem.Soon enough a group at Bell Labs discovered that the design problem was a case of combinatorial optimization and, in particular, linear programming where insist on all the numerical results being integers (whole numbers — no fractions permitted). Why? Can’t deploy just 38.473% of a piece of equipment or long haul data link!So, from that group at Bell Labs was the now famous, seminal, bookMichael R. Garey and David S. Johnson, ‘Computers and Intractability: A Guide to the Theory of NP-Completeness’, ISBN 0-7167-1045-5, W. H. Freeman, San Francisco, 1979.There they argued that, for an optimal solution to the network design problem, about the best approach we know is just to evaluate all possible designs and pick the one with the cheapest evaluation. Then, sure, the failing here is that the number of possible designs is from a lot of combinations, a combinatorial explosion, and easily enough in practice can be absurdly large. E.g., once I did a little back of the envelope calculation that the number of designs could be larger than all the subatomic particles in the visible universe even if that volume were filled with matter of the density of neutron stars. So, can’t hope to evaluate all the designs one at a time, right, not even on a 5.0 GHz Intel processor with 8 cores and 16 threads soon!Some people take combinatorial optimization quite seriously; these problems are one of the main motivations for efforts in building a quantum computer which might have a chance of actually evaluating all the combinations for all the designs.From that book by Garey and Johnson and a lot of related work, especially from Stephen Cook, there got to be the question in math and computer science P versus NP. Or, for a problem like evaluating the costs, etc. of a network design it’s easy enough to write computer software that will evaluate a design in computer time proportional to some polynomial (right, like in high school) in the size (with nearly any reasonable definition, say, just counting bits) of the input data, i.e., specifying some one design.Then for all such problems with such a means of polynomial evaluation, is there necessarily also a way to find the cheapest design from software with running time also, including for worst case problems, just some polynomial in the size of the input data? If so, then P = NP, and for various reasons, some seemingly quite far from optimization, that will be a biggie, a big thing, for math, computer science, the economy, and civilization. E.g., there is a standard argument that P = NP would essentially trivialize all of math and its applications.So, we ran into the question of P versus NP considering network design. Okay, but Garey and Johnson showed that there is a gigantic range of problems, in optimization and much more, e.g., computer security, that are essentially equivalent to the problem of network design, that is, either there is a polynomial algorithm to solve all of them or no polynomial algorithm to solve any of them.Now the question of P versus NP is taken as one of the most important problems in research, and Clay Mathematics in Boston will award a prize of $1 million for the first correct solution. All things considered, for a solution, that prize may be chump change.The question P versus NP is close to a philosophical question about the universe. It’s a biggie question.But, in a significant sense, especially in practice, Garey and Johnson were badly wrong: In their book they have a cartoon with a mathematician standing in front of the desk of a business executive and confessing that they (the mathematician) is not able to solve the executive’s (optimization) problem but pointing to a long line of other mathematicians and computer scientists that can’t solve the problem either.Well, so far, usually in practice, except for an especially bright business executive, his problem has a solution, at least one he will like. Maybe the executive, e.g., as at Bell, just wants a network design that will save money and will be happy saving, say, all but the last few percent of the savings of an optimal solution. That is, if he has a design costing $10 million and there is an optimal design he can’t find that would cost $8.5 million, then he might be plenty happy with a design that costs $8.5 million plus a few hundred, thousand, or tens of thousands of dollars. So, he may be happy with a nearly optimal solution.In this way, is the business executive easy enough to please? Usually in practice, yes. Why? Because the applied mathematicians have come up with one heck of a collection of methods, and associated software, that make some approximations that in real problems save all but the last few percent of the available savings. Indeed, part of the challenge of the question of P versus NP is insisting on polynomial performance also on worst case problems. That is, Garey and Johnson were looking for an algorithm that delivered much more than the business executive really needed immediately or near term. Sure, in the long term, if someone shows that P = NP, then the changes in civilization will likely also affect the business executive.The role of worst case problems is too commonly lost even on people with good academic introductions to P versus NP: Once I was at a startup in Plano, TX that was trying to do network design for the growth of the Internet and mentioned one of my successes in combinatorial optimization. The reaction was essentially that I was lying because somehow the listeners had come to believe that all problems in combinatorial optimization are too difficult to attack successfully in any reasonable sense. This belief is wildly false; optimal solutions, that is, down to the very last tiny fraction of the last penny of savings, on worst case problems of significant size, will often be too difficult to attack. Very nearly optimal solutions for real problems can be quite reasonable to find.Net, usually in practice, the really difficult part is saving the last tiny fraction of a penny.So, at times, optimization was set up as a goal that, on more careful examination, was, for the immediate business need, usually a bit silly and at times even harmful. Or, the anxiety ridden, obsessive mathematician was terrified of being accused of some sinful transgression and not being optimal and, thus, wasting money, even the last $1000 on a $10 million or $8.5 million project.So, for too many years, the too difficult goal of an optimal solution stood in the way of many reasonably available significantly better solutions. Bummer.Sure, much of the reason was the different outlooks! The applied mathematician may want a problem he can work on that is really difficult to solve, or that he can be proud to have solved, e.g., find an optimal solution, while the business owner wants to make money and, for some expenses, not waste too much time, money, or effort. Sometimes not the same things!Yes, these issues of P versus NP all come up with full force in airline fleet scheduling (yup, I wrote the first software to schedule the fleet at FedEx and encountered these issues), airline crew scheduling, lots of problems in logistics, lots more problems in cases of resource allocation, a huge range of problems in scheduling work on computers, etc.For much of what to do in practice, there is, sure,George L. Nemhauser and Laurence A. Wolsey, ‘Integer and Combinatorial Optimization’, ISBN 0-471-35943-2, John Wiley & Sons, Inc., New York, 1999.Right: Nearly everything in the book encounters the question P versus NP.For software, there is, sure, CPLEX, now owned by IBM, and developed by R. Bixby, long a prof at Rice (supposedly he has a relatively nice house in Houston).Once some guys thought that they saw a nice business opportunity from a change in banking laws that would permit a merger and, then, cross selling. So, in the short term, the main problem was allocation of existing marketing resources, and they formulated the problem as linear programming with 40,000+ linear equations with 600,000+ variables and asked that all the variables take on whole number values. They wrote some software using simulated annealing, using a random number generator to wander around among the 600,000 variables, ran on a computer for some days, took the best solution so far, and gave up. Gads.I got their linear programming formulation, did some derivations in non-linear duality theory (it turns out that necessarily there is a nice case of convexity to exploit), wrote some software, and in 905 seconds on a 90 MHz single core processor found a solution guaranteed to be within 0.025% of optimal. Optimal? Maybe not. Showed that P = NP? Nope. “Close enough for government work?” Yup. Quite doable and helpful in practice? Yup. Really difficult for me to do? Nope. Where’d I learn about non-linear duality? Grad school and some work for the US DoD. Where’d I learn to do math derivations? Decent background in pure math.Results: The guys who had tried simulated annealing were assuming I’d fail. When I didn’t they didn’t like me anymore! Thus, they didn’t want to continue working with me and, in particular, didn’t want to take my solution to their target customers.Lesson: Before doing such work, have a written agreement on compensation, say, based on getting paid a fraction of the cost savings produced! Since the people I was working with had little faith that there could be anything better than their simulated annealing (maybe they’d read about it in, say, Wired), up front they might have signed a good deal for me!Ah, once again, “get what you negotiate, not what you deserve” (@JLM). Or, “never give a sucker an even break”. Or, “there’s a sucker born every minute”. Or, how the heck is the common man in the street going to be able to evaluate where the available techniques of combinatorial optimization do and do not work? Even more for the rest of pure/applied math? Right: This situation is an obstacle the flip side of which can be an entrepreneurial opportunity.Ah, did another one: There has long been the problem of just what pharmaceutical salesmen should do, that is, which physicians to visit, what products to explain, what free samples to leave. In this case there is enough data to attempt some optimization, right, combinatorial optimization. A little company had a little, first-cut computer program based on some simple heuristics.I was able to formulate the problem as least cost network flows on a network where the arcs all have capacities that are integers. Presto! Bingo! Grand slam! Such a problem is next to trivial to solve, especially in practice. So, just use the simplex algorithm of linear programming. Can, if wish, slightly modify the algorithm to take advantage of the special structure of the network flows. In either case, if start with an integer (but maybe not yet optimal) solution and then apply the simplex iterations. In such a problem, simplex will maintain integer solutions and will find an optimal solution which is also integer. There is some good, relevant work from W. Cunningham, one of my profs, and later Chair of the important Department of Combinatorics and Optimization at Waterloo.Result: I submitted a nicely done math paper showing the network formulation and was well along on writing the network optimization software (the company wanted their own software and didn’t want to resell existing, sufficient commercial software), and this progress with the prospect of software for clean, optimal solutions caused the guy with the heuristic to play politics and cut my support. Again, I should have had a contract!So, sometimes in practice, combinatorial optimization can be quite doable. Alas, there is nearly no one in US business middle management who understands enough about such applied math to bet part of their career on such a project; standard problem in business and, indeed, one of the opportunities for entrepreneurship and startups!But in practice, and maybe closer to what Fred has in mind, in business can struggle where need to make some decisions, have maybe not nearly enough data or maybe too much to analyze, are tempted to find decisions that are best possible in some sense, with this temptation can grind teeth, furrow brow, lose sleep, make Starbucks richer, etc., all for, maybe, no very good reason.Indeed, for such nasty problems there are some mathematical approaches that at times can work: Do some probabilistic optimization. But, too commonly in practice, this approach is what to do if had a lot of detailed data on probability distributions that mostly in practice will not have.My Ph.D. dissertation was on probabilistic optimization: I found a nice mathematical solution and wrote corresponding software. In this case, it was reasonably easy to get the needed probability distributions — to get such a doable problem, usually have to pick carefully!Ah, that’s enough of Optimization Theory and Applications 101!

  28. Joe Marchese

    An even more valuable insight from Herb Simon: “Whatinformation consumes is rather obvious: It consumes the attention of itsrecipients. Hence a wealth of information creates a poverty of attention, and aneed to allocate that attention efficiently among the overabundance ofinformation sources that might consume it.” Remarkable in that he said this in 1971. Happy to give my (occasional) attention to AVC.

  29. Stephen Bradley

    Sort of like achieving consensus vs unanimous support. Nice to put a name on it.

  30. panterosa,

    Can’t Get No Satisfaction? Satisfice!Almost sounds like a TM or ® exercise program, like Prancercise®.

  31. johndodds

    Wharton should hangs its head.

  32. Twain Twain

    @fredwilson:disqus – In some cases, portfolio constituents follow the “optimizing” model. In some cases, they suit “satisficing” strategies more. An example of the first would be if the startup is pure quant (e.g. like SecondMarket). An example of the second would be where the startup has subjective extraneous factors that influence its success (e.g. like Twitter where it was a user rather than the founding team that invented the usage of @ and #).Mathematically, optimizing is from an era where there is a single discrete end-point of success and everyone is on the same linear path. Satisficing is from John Nash’s 1950 construct for bargaining theory where there are continuous trade-offs and bargaining dynamics and multiple favorable outcomes. This is the basis of today’s gamefication, btw.I tend to aim towards COHERENCY rather than optimizing or satisficing. This factors in that decision-making is more than a purely objective exercise. Every founder worth their salt would present 3 outcome cases for their model projections (optimistic — optimizing; pessimistic — geared; and realistic — satisficing).However, subjectivity is always at play; this hasn’t been factored into any of the mathematic models for optimization, investor “pattern recognition” or heuristics (even those by Daniel Kahneman).The subjectivity happens both in the form of personal chemistry between the investor partners, between the investor(s)-founder(s), between the startup and the market.So coherency is whereby the objective metrics that are geared towards optimization are integrated with the subjective factors that are geared towards satisficing in such a way that it makes sense for all the parties involved.Decision-making intelligence is of great interest.

    1. SubstrateUndertow

      COHERENCY rather than optimizing or satisficingCOHERENCY seems like a very organic, bio-mimicking solution approach that in essence defines the core concept of non-linear optimization ?

      1. Twain Twain

        Non-linear optimization assumes there’s some type of convergence around multiple saddle points (either convex or concave) on a topographic basis.Coherency is closer to what happens in biochemical synthesis.In chip design there’s emergent examples of bio-mimicking in neuromophic structures:*…Coherency is also different in this respect: ALL mathematics to-date has been about tools to measure rational, objective, optimal and repeatable (and therefore predictable) phenomena. Those tools have stemmed either from Deterministic Logic or Probabilistic “Fuzzy Logic” frameworks.No mathematics has been developed or able to measure our subjectivity and contextual biases with any degree of coherency. Conjunct analysis doesn’t cut it either.Fred’s lament about Optimization vs Satisficing makes me think of Einstein: “Not everything that counts can be counted, and not everything that can be counted counts” and “The intuitive mind is a sacred gift and the rational mind is a faithful servant. We have created a society that honors the servant and has forgotten the gift.”The rational mind probably tells USV partners to optimize but that would only be seeing 50% of the picture of the portfolio constituent.

        1. SubstrateUndertow

          All very very good points indeed !”The intuitive mind is a sacred gift and the rational mind is a faithful servant. We have created a society that honors the servant and has forgotten the gift.”Beautiful quote !Still cells don’t utilize any formal”Deterministic Logic or Probabilistic “Fuzzy Logic” frameworks”Cells and other living-systems utilize that very same universal subterranean sacred gift, distributed recursively-adaptive iteration, to seek out collective best fit survival strategies.( the massively trial/error version of Probabilistic “Fuzzy Logic”)Just saying:Biological COHERENCY solutions seem reliant on organically recursive/iterative processes rather than on negotiated upfront fixed solutions.That implies the organic possibility of substituting negotiated upfront fixed contractual solutions with algorithmic App-based contractual agreements which define/enforce specific recursive distributively-adaptive feedback/ratio-driven contractual behaviours on all parties.Distributively adaptive network Apps, acting as contractual social nervous-systems, are our opportunity to run experiments in mimicking that sacred subterranean organic-mojo gift at the human organizational level.(where blockchain is just the low hanging fruit)The network-effects involved here could even potentially help stomp-out/police the poisoner’s-dilemma post-contractual defectors?Just some “pie in the sky” food for thought :-)http://img1.southernliving….

          1. Twain Twain

            There’s an even better quote from John Von Neumann: “When we talk mathematics, we may be discussing a secondary language built on the primary language of the nervous system.”What does this mean? Well, if maths is a secondary language then Probability is too.What then precedes Probability? It’s not Determinism.So let’s apply that to a Distributively Adaptive Network scenario. If the Blockchain is low hanging fruit (the probability factors), what’s the fruit up top?

          2. SubstrateUndertow

            “When we talk mathematics, we may be discussing a secondary language built on the primary language of the nervous system.”That is a fabulous quote, how the hell did I manage to miss that one. I’m not really a math guy but still I can’t believe I missed that one in my travels. Thanks – that is a real keeper for me !What does this mean? Well, if maths is a secondary language then Probability is too.What then precedes Probability? It’s not Determinism.Isn’t that question more about the limits of both natural and mathematical language ?Just for fun I like to ask this question when the Probability vs Determinism fur starts to fly.Is the universe determinately-indeterminate or just -indeterminately-indeterminate ?Computaional-space is a subset of perceptual-space and both are subsets of representational-space beyond that all forms of language stall into endless self-referencial meaninglessness.So let’s apply that to a Distributively Adaptive Network scenario. If the Blockchain is low hanging fruit (the probability factors), what’s the fruit up top?That is the $64,000 “Social Networking Question” but it probably involves us using biology as nature’s visionary cheat-sheet !The problem at hand is to renormalize as much as possible all that “computaional-space + perceptual-space” into an inclusive, meaning optimally-sharable and integrative “sacred gift + the rational mind” dual servant space accessible to the rest of us.By renormalizing I mean taking all that abstraction and subjective situational relativism and translating it into a falsely-objectied yet more socially workable linearized set of widely accessible new narratives/metaphors/lexiconal-memes.Simplified/renormalized narratives/metaphors/lexiconal-memes that cater to the endemic nature of linear human perceptual ergonomics.Getting everyone on board with a simplified/renormalized set of narratives/metaphors/lexiconal-memes that catalyze our collective visualization/participation in a palette of practical/reusable patterns of distributively adaptive behaviour is a potential game changer.An accessible mass-culture of simplified “Organic-Process-Literacy” sets the experimental power inherent in the collectively recombinant human imagination loose on all the distributavely-adaptive social-potential made possible in an emerging world of unlimited networkable synchronicities.The spread of the Phoenician alphabet with its simplified isomorphic transcription of phonemic units made printing and print-literacy simple enough to become accessible to mass-culture via the catalyst of the printing press thus accelerating the evolution of human culture on all fronts.Simplifying the narratives/metaphors/lexiconal-memes associated with the organic social-potential inherent in distributively-adaptive social-networking synchronicities is an analogue of similar importance to human cultural evolution!

          3. Twain Twain

            @SubstrateUndertow:disqus — You wrote: The problem at hand is to renormalize as much as possible all that “computaional-space + perceptual-space” into an inclusive, meaning optimally-sharable and integrative “sacred gift + the rational mind” dual servant space accessible to the rest of us.Getting everyone on board with a simplified/renormalized set of narratives/metaphors/lexiconal-memes that catalyze our collective visualization/participation in a palette of practical/reusable patterns of distributively adaptive behaviour is a potential game changer.”@sigmaalgebra:disqus — You wrote: “Since taste, appearance, mouth feel, etc. don’t get handled well by mathematical optimization.”Ok, so the potential game changer would need a tool other than Probability since we know it works for linear and non-linear optimization (optimizing and satisficing) but taste, appearance, mouth feel etc. are all subjective factors that Probability can’t deal with.The question is what’s the complementary counterpart to Probability and logic? What’s the tool that could measure this counterpart (which has some factor of intuition and taste in it)?Ah, well……at some time-point I had a crazy idea that neither optimizing nor satisficing make sense. The first fits with Gaussian and Poisson probability distributions and the second fits with Nash’s bargaining theory and matrix theory. And then Deep Learning with applied Gradient Descent is just a a variation of probability.But what if John Von Neumann, Albert Einstein and Daniel Kahneman are right? Our brains are not built for Maths and Probability first or as a default, and there’s a neurosystem process that PRECEDES whatever probability and logic calculations we experience.A neurosystem process that permeates in a network-effect way how we make decisions?How can we measure this neurosystem process?@fredwilson:disqus — This is what’s great about AVC bar. We go down rabbit holes, follow random strands of string and somehow things segue and intertwine.

      2. sigmaalgebra

        For numbers a, b, and c, the algebraic expressionax + by + czis linear in variables x, y, z.Why? Because we think ofax + by + czas a function, say, f from the set of all triples (x, y, z) to the set of real numbers R orf: { (x, y, z) | x, y, z in R } –> RWell, think of the triple (x, y, z) as a vector u. Similarly for vector v.For real number p, we definep (x, y, z) = pu = (px, py, pz)Then f has the property that for any real numbers p, qp f(u) + q f(v) = f( p u + q v )So, f is said to be linear.Why linear? Because the set of all points x, y plotted on an ordinary X-Y corrdinate system where for real numbers a, b, c we haveax + by = c is a straight line, and similarly in dimensions higher than 2.Linearity is nice stuff: In G. Simmons, Introduction to Topology and Modern Analysis, “the two pillars of analysis are linearity and continuity.” Yup. Continuity? Sure, e.g., it ensures that the Riemann integral of freshman calculus exists.Linearity permits doing a lot of good work just by pushing symbols around instead of a lot of more difficult things. Linearity is a great thing to have and, with generalizations, a big thing in mathematics.For positive integers m, n, an m x n matrix is a rectangular array of numbers with m rows and n columns.Then for some m x n matrix A, and 1 x n c, m x 1 b, and n x 1 x we consider the linear program L1 z = c xAx = bx >= 0 where x >= 0 means that all n components of x are >= 0.Definition: We need to define the matrix product cx: Well since c is 1 x n, write it as the row vector[c<sub>1</sub>, c<sub>2</sub>, …</sub>, c<sub>n]and since x is n x 1 write it as a column vector (we use round things instead of square ones)(x<sub>1</sub>, x<sub>2</sub>, …</sub>, x<sub>n)Then the matrix product cx is the sum c<sub>1</sub>x<sub>1</sub> + c<sub>2</sub>x<sub>2</sub> + … + c<sub>n</sub>x<sub>n</sub>Then for Ax = b for each of the m rows of A we multiply that row by x as we did in cx and we let the result of that matrix multiplication be the component of the corresponding row of b, that is, row 1 of A times x results in row 1 of b, and the same for each of the rows of A and b. So, that’s how matrix product works.Why does matrix product work that way? Because (1) that is how the matrix equation Ax = b is really the same as we saw in high school with m linear equations in n unknowns (the components of x ) and (2) makes matrix A work like a linear operator from R<sup>n</sup> to R<sup>m</sup> which is powerful and, thus, important. For much of the power, of course, there is the now classic, right, heavily from NY,Nelson Dunford and Jacob T. Schwartz, Linear Operators Part I: General Theory.End of definition!So, in L1, we seek x to make z as large as possible.This subject is linear programming and was heavily the work of G. Dantzig from his work at Rand in the late 1940s and from his experience in WWII US logistics.Or, more generally for linear functions f: R^n –> Rg: R^n –> R^m and in z = f(x)g(x) = b we seek x to make z as large as possible.If the functions f, g are not linear, then we have the subject of non-linear programmiing. Linear programming where we also insist that the components of x be integers is integer linear programming, strictly part of non-linear programming, and now a very big deal by itself due if only to the importance of the question of P versus NP.Generally linear programming is easier and, indeed, is often a good way to attack non-linear problems iteratively.Both linear and nonlinear programming have been important in oil refining in Houston.The Markowitz work on portfolio optimization is a relatively simple example of non-linear programming.One of the pillars of non-linear programming is the Kuhn-Tucker conditions. Here is an intuitive view: Suppose are in a cave with vertical walls and an uneven floor and are looking for the lowest point on the floor. Well, the Kuhn-Tucker necessary conditions say that if put a marble on the floor and it starts to roll, then are not at the bottom. What is tricky about the Kuhn-Tucker conditions is how they handle the case of the marble being next to one or more (in higher dimensions, might be next to several) walls.For these conditions to hold, the constraints, that is, the walls, must be reasonably well behaved, that is, not really pathological. So, there are some constraint qualifications that, if they hold, where put down the marble, ensure that the necessary conditions hold.Some of the constraint qualifications are easy to verify but apply less commonly; some of the qualifications are more difficult to verify but apply more commonly; some of the qualifications imply some of the others. Two of the more difficult qualifications are the one due to Kuhn and Tucker and another one due to Zangwill. Does one imply the other? Actually no; they are independent. While I was in grad school, I worked this out and later published it in JOTA. The work was a bit tricky, which is why the result was new.At one time mathematical economics was highly interested in linear and non-linear optimization. There is a famous paper by Arrow, Hurwicz, and Uzawa. My little work on the constraint qualifications also answered a question posed in their paper but not answered there.With some convexity assumptions, can also get the Kuhn-Tucker sufficient conditions for optimality.In practice, one approach is to use, say, linear programming iteratively to find a point satisfying Kuhn-Tucker necessity and argue separately that that is the only such point. If can do that, then can find the optimal solution.End of the overview lecture in Mathematical Programming 101.

        1. SubstrateUndertow

          I meant “non-linear optimization” as in a distributively-adaptive social network that homes in on best fit volitional coherence between stakeholders via recursive iterations of adaptive response made possible by a mix of synchronous/asynchronous nesting of network effects.Maybe “non-linear optimization” was a poor choice of words ?

          1. sigmaalgebra

            “Non-linear optimization” may have several meanings. One of the oldest and well defined meanings is in applied mathematics, e.g., as I have outlined in other posts in this thread.It may be that the meaning you have in mind could be connected with the applied math meaning: So, take a situation such as you described, provide a corresponding quantitative description, and do some math to show the convergence you suggested would follow from adaptive iterations. The applied math convergence might be from optimization of some criterion from using some iterative technique from the applied math of non-linear programming.The first guess at such math would be steepest descent from finding a gradient and moving in the downhill direction given by the gradient.If some constraints were involved, then might also use gradients of the constraints and get a linear program and solve that for one iteration; then find the gradients again and repeat.

    2. sigmaalgebra

      In mathematical optimization, something like satisficing was called goal programming.Another area of optimization was called constraint programming (or constraint logic programming) where just want a solution that meets some requirements or constraints and otherwise don’t care. But in fact constraint programming is essentially equivalent to optimization in all respects.So, for an application, suppose that yesterday you killed 5000 hogs and hung the results in a cooler overnight. Then, today, you are busy cutting the pork and putting the pieces in boxes and have some 18 wheel refrigerated trucks arriving at your loading dock to take on the pork, ballpark 40,000 pounds per truck.You know what boxes you will have available for loading when and what boxes each truck needs to have loaded and in what order and you want to know the order in which to bring the trucks to the loading dock to get all the trucks loaded and on their way ASAP after the last box is filled. Otherwise you don’t care. Real problem.The company iLog in France was into constraint programming, and as I recall at one time SAP bought them. In the end for their technology and software they worked with R. Bixby, at Rice, and creator of CPLEX, now owned by IBM.

      1. Twain Twain

        There’s also the Diet Problem, the Air Control Problem and the Seed Planting as examples of constraint programming in OR.We can’t assume in the Diet Problem that we’re trying to optimise weight loss. The person may satisfice for losing less weight and upping their vitamin intake and absorption.

        1. sigmaalgebra

          The usual objective of the diet problem is to minimize cost while meeting nutritional requirements. Since taste, appearance, mouth feel, etc. don’t get handled well by mathematical optimization, the more appropriate cases of the diet problem are for feeding livestock and there likely there are savings to be made.E.g., my father in law typically had 40,000+ chickens on his farm. Do the right things with chicken nutrition and can get the chickens to market weight and healthy quickly and make good money; otherwise can go broke buying chicken feed.My understanding is that Ralston-Purina when mixing animal feed, say, to meet the nutritional content on the package labels, ran many linear programs a day. Why linear? Because no doubt quite accurately the relationships are linear, i.e., combine the corn, fish meal, vitamins, etc. and each variety of each of the fats, amino acids, carbohydrates, vitamins, etc. don’t interact and, instead, just combine linearly.

          1. Twain Twain

            You wrote: “Since taste, appearance, mouth feel, etc. don’t get handled well by mathematical optimization…”This indicates you don’t believe Probabilistic optimization works for such subjective factors?What then would?

          2. sigmaalgebra

            Addendum: There is now a really cute article on the diet problem at… This indicates you don’t believe Probabilistic optimization works for such subjective factors? The situation is even simpler: We just do not have enough data to say, given a combination of foods, what the mouth feel, etc. would be. Moreover, the relationships involved might well be non-linear which could put us out of linear programming.E.g., in the dessert freezer compartment of my local grocery store is some small coconut cream pies. They look really good and taste pretty good.Then look at the ingredients list and see that the contents are not much like what Grandma might have made at home. Indeed, if took Grandma’s coconut cream pie and froze it, then likely the whipped cream would collapse and the custard filling would come out awful.Instead the food chemists making the frozen pies no doubt had to work hard to get something as good as they did once the pies were defrosted.How’d they do that? Of course, lots of attention to flavor and mouth feel. My rough guess is that they did a lot with various emulsifiers.The pies are from ConAgra, not exactly Grandma in her kitchen.Maybe they knew of and used some mathematical relationships, but I would doubt it. Maybe the relationships would let them have some cases of convexity they could exploit — maybe. Convexity? Maybe you asked? Okay: Intuitively it’s half of linearity.For a vector space V over the set of real numbers R, for an example of V, for a positive integer n the n-dimensional vector space R^n, suppose we have a function f: V –> R. Then, function f is convex provided for any vectors u, v in V and real t in the interval [0, 1] we have that f( t u + (1 – t) v ) <= t f(u) + (1 – t) f(v) So, the secant line on the right over estimates the corresponding function value on the left.Function f is concave if and only if (iff) function -f is convex.Convex functions are nicely common, and such convexity is powerful and often can be exploited with good results.It’s possible to go on and on with convexity, e.g., the now classic R. Rockafellar, Convex Analysis. It’s easy to show that a convex function is continuous, and as I recall Rockafellar also shows that such a function is differentiable almost everywhere with respect to Lebesgue measure. Cute.Even some EE profs get all excited about applications of convexity.

  33. ShanaC

    Satificing can lead you down some weird places – because you go to a random walk far away from where should be.Once you are there, it can be hard to go fix the issue.



  35. malieberman

    Satisfice is a human behavioral decion making process that we go with the first option that offers an acceptable payoff, i.e. good enough. This requires the bare minimum of thought to achieve the goal. This works well in every day simple decisions, which type of coffee to buy from my regular vendor. However, more complex decision making requires more thought and creativity. Satisfice sometimes isn’t “good enough.”

  36. John

    I just call it “Good Enough.”

  37. Pete Griffiths

    First encountered this term some 35 years ago. Hit a nerve.

  38. Matt Vose

    Best available solution versus mediocre, half-assed solution. Which is preferable?

  39. Sean Hull

    Yeah, that’s a great one. In early years of Oracle work, DBA’s used to talk about Compulsive Tuning Disorder, where they would continually try to tune databases, that were already running well. Identifying a threshold where things are “good enough” is surely important in a lot of areas of decision making.

  40. sachmo

    There’s a book called “The Paradox of Choice” by Barry Schwartz in which the author argues that ‘satisficers’ in general are happier than those that optimize decision making. The author supports this with many studies and such – however it was definitely not the most well written book and I doubt that it will be remembered in 50 years.Mark twain once said that the difference between the right words and the almost right words is the difference between lightning and a lightning bug. For books this is certainly true, and I think this is true for hardware and software products as well.I have a problem with satisficing. It just doesn’t feel right sometimes. If Barry Schwartz, the author of ‘Paradox of Choice’ had been an optimizer, his book would have been much better.I have a few thought experiments:You move to a new city. Do you do the apartment hunt for 2-3 weeks and satisfice? Or comb through different neighborhoods and spend months or even longer to find the just right apartment? (I spent a long time finding the ‘just right’ apartment, and every day when I come home it’s awesome.)Or what about with regards to a relationship with someone else? Go for satificing or optimizing? Or a job / career? (I’ve never satsificed in either of these categories)Or even better yet a product release? (Yeah right… see Mark Twain quote)So what type of decisions are worth satisficing on? I mean I don’t really care what type of shaving cream I use, but I doubt that what anyone spends their time optimizing this.What are typical decisions that many would over-optimize on, when they should just chill and satisfice? I’m curious to ask the satisficers out there.

  41. p7r

    There is a political theory that 90% or more of the population are “satisficers” and only the remaining 10% are “maximizers”. I can’t find the article where I read that, I’m sure somebody else will find a citation.The maximizers are the people who hunt around for the best interest rates on savings, credit cards, etc. and will move banks regularly. They will bargain hunt, load up on goods when they are discounted and stockpile, and they are keen proponents of the free market: they believe that any inefficiency when it comes to money lies with them and their laziness rather than the market.The satisficers are the people who are happy with what they have, because they don’t want to exert time constantly fearful they could be doing something better. As long as it is good enough and does not cause pain, that’s fine. They stay with the same bank for years despite awful interest rates on their savings. They pay too much in credit card interest. They go to a fixed interest mortgage rather than have the “worry” of a floating rate. They buy goods as they need them, at the shop they always go to. In a sense they are maximizing/optimising for lack of worry rather than for money. They are the reason why uncompetitive businesses still have custom in competitive markets (we’ll park Peter Thiel’s theory that competitiveness is the opposite of capitalism for another day).The politics comes in with the fact that whilst in the general population the split is 90/10 in favour of the satisficers, if you look at the halls of power you will see the ratio reversed. Maximizers are far more prevalent in the political classes and the business elite.This is important, because for the last 50 years we have seen policies and business lobbying structured by maximizers for the benefit of maximizers, and the general population is becoming increasingly restless.Globalisation is a maximisation solution, as is deregulation and privatisation. Most satisficers think they aren’t as hot an idea as maximizers do.In the UK at least (where I hail from), most people would probably agree that when they want a surgeon they want a competent NHS surgeon to be available in a reasonable time frame.Instead after 30 years of successive tinkering by the maximizers from Thatcher, Major, Blair, Brown and Cameron governments (all maximizers, all have been involved in the privatisation of what was once owned and operated by the state at some level), the population are being offered league tables of surgeons that rank them by competence, death rate, etc. and given a “choice”. “Choice” is something maximizers love, and satisficers resist, especially when it forced on them, because most satisficers accept that they do not know enough to competently choose a surgeon for an operation when there are multiple ways to rank them.Watch this space: I suspect over the next 20 years we are going to see a backlash against the maximizer class and the rise of the satisficer. That will involve more public ownership of assets that are essential for all (water supply and other utilities, public transport infrastructure), and increasing friction between the two groups.In case it is not obvious, the tech industry is very much the playground of the maximizer. Ayn Rand was certainly not a satisficer, and her politics runs through much of the tech industry as we’re all aware. The risk is that when people reject our work we default to calling them uneducated Luddites when in fact they’re just happy with the status quo as long as it is “good enough”.It is interesting then to hear that a leading VC is eschewing the maximizer label, and siding with the satisficer population. May you lead the charge, Fred. Viva la revolution!

  42. Josh Young

    I find Wikipedia’s other definition of satisficing more helpful: “it is optimization where all costs, including the cost of the optimization calculations themselves and the cost of getting information for use in those calculations, are considered.”

  43. William Mougayar

    Consensus in business is not good, maybe as an exception, its OK. Yet, in politics, itt’s mostly the norm, which explains poor outcomes.

  44. Elia Freedman

    How did you get to consensus from the definition above? It’s early here but I must be missing something. I didn’t see that at all.

  45. LE

    Be careful of the danger of consensus.I’ve noticed a pattern of companies sometimes achieving great things by having the one asshole in charge that will listen to everyone but in the end gets to be the decider and do what he wants.Part of the reason is it’s sometimes hard to quantify concepts that others don’t fully grasp because of their upbringing, qualifications and experience or biases. So if you get enough people together you get milk toast.



  47. Joe Biden

    Yup. Consensus results in “safe” moves rather than necessary calculated risks to shake up the status-quo.Ripping off Sir Francis Bacon: Talking with people individually helps to tease out honest opinions whereas people tend to be less heretical speaking as a group.

  48. LE

    Consensus in business is not goodCamel is a horse by committee concept, eh?

  49. Antony Evans

    It’s a different style. Japanese companies work well with consensus decision making. It takes longer to make a decision, but once the decision is made then execution happens more smoothly as everyone is already on board with the decision. In American companies you have more fighting and politics which makes execution harder. Both sides have their advantages (though I think startups probably do better with strong leadership).

  50. LE

    The weak, uninformed ones went along with whatever the dominant safe position was.What I have found is that you can typically overcome this by being the guy who everyone else watches out for. By that I also mean you are the most prepared with data and supporting information to back up what you are saying. Reams of it if you can. Enough to completely overwhelm anyone who is just using words and doesn’t want to do anymore than talk and offer a handout.Effort pays off. People are typically lazy and are unprepared.That said I’d rather have people that are “invested enough or and informed enough” for sure rather than having to resort to manipulations.

  51. JLM

    .In politics and non-profits, ten percent of the people do 99% of the work.The rest are there for the grand openings, the ground breakings, the parties.JLM.

  52. LE

    especially when working on a MVPFirst thing I thought of when I read the post was “MVP”. Lot’s of similarities between satisficing and MVP. An MVP really is a product that is a satisficed product. It’s just good enough and no more.

  53. Elia Freedman

    Got it. I thought you read that in the satisficing definition. Thanks for clarifying,

  54. awaldstein

    I’ve long stopped worrying about consensus. It’s enough to get all in, buy in from strong teams where they not only participate but own the process moving forward.

  55. LE

    the parties.The local do nothing small township “mayor” was at the local chinese restaurant last night. Mrs. Owner lady told me the day before when I went in to get my order and I noticed all the setup going on.After telling me why, she says (imagine this in broken english) “you show up then I get to charge for extra head, ok?”.

  56. Twain Twain

    It’s possible to achieve great things in startups without being an asshole/jerk? Paul Graham believes so:”It’s certainly possible to build a multi-billion dollar startup without being a jerk,” Graham says. “We’ve funded several, and the founders are all good people. In fact, based on what I’ve seen so far, the good people have the advantage over the jerks. Probably because to get really big, a company has to have a sense of mission, and the good people are more likely to have an authentic one, rather than just being motivated by money or power.”* http://www.businessinsider….Great leaders should be inspiring and inspired in their vision and intimidating to the competition in their execution.

  57. awaldstein

    the correlation between ‘good people’ and true vision’ is as far fetched a stretch of logic as I’ve seen.

  58. LE

    “Startups” as defined by Paul Graham (based on his short history in the business word as he sees the business world that is) are staffed typically by a group of individuals that is not the same that would work for an oil rig startup in Texas or a construction company in New Jersey. Young idealistic men and women using other people’s money many without families to support and being minted with good credentials to fall back on. Hardly dog eat dog. Not like running an airline or anything. [1]By the way, the definition of “good” almost certainly stems from the way they interface with people as opposed to the decisions that they make. In the end it’s always nice to fire someone and make them not feel like shit when you do so. In the end does it really matter if they are out of a job? If so, how much?the good people have the advantage over the jerksMy guess is that from his perch he can’t possibly differentiate between people being nice to him and sucking up to him because he is “PG” and what the same person is like to others (not saying it isn’t the same just that the possibility exists). When I sold my first business I had salesmen that called on me that I thought were the nicest guys. They wanted my business and the order. So they agreed with everything I said and never gave me a hard time. After I sold the business while they weren’t jerky to me about 1/2 of them (arbitrary to make a point) completely ignored me because I could no longer give them business.[1] In traditional business the idea of writing information that is helpful to your competition is close to ridiculous as can be. It happens but nowhere near as much as it happens on the Internet the way people have been raised and bred and the age group.

  59. Twain Twain

    Paul Graham’s definition of “good people and true (authentic) vision” would be different from Tim Cook’s definition of the same — Cook having worked with Steve Jobs who was reputed to have a fearsome temper and considered by some to be an asshole (not a good person). Yet equally it’s indisputable he had a true vision.In a way, it’s like when someone says, “Oh this wine is…complex / austere / unctuous / any other adjective.”Correlations and logic are non-starters in both examples: the Paul Graham one and the wine one.It’s all personal subjective opinion and instinct.

  60. awaldstein

    to a degree.being a complete jerk is often point–to make a correlation between jerkness and vision is like saying you need to be tall to be smart.counter intuitive if not incorrect

  61. Twain Twain

    Haha, got you.Did you read this Economist article on ‘The Look of a Leader’: “The typical chief executive is more than six feet tall, has a deep voice, a good posture, a touch of grey in his thick, lustrous hair and, for his age, a fit body.”*…Interestingly, Einstein was 5’9″. Marie Curie was 5′.Is it smarter to be a 6′ CEO whose tenure and power base is typically 2-5 years or to be any height and create something that advances global society and lasts for centuries beyond your own lifetime?Haha.

  62. sigmaalgebra

    Can get a lot of resentment andcreate a lot of office enemies showingup with a lot of good preparation. A better approach is to have animporant project, work on it quietly,and when it is done show it as done.If describe the project before it is done,then can get some attempts to sabotage the project.Even keeping the project relativelysecret until it’s done can still result in a lot of resentment andhostile politics.If have something good, then apparently the best approach isto be CEO of a business that gets revenue and earnings fromthe project.

  63. awaldstein

    Makes me smile.I can tell you that after reporting directly into some of the most visionary and successful individuals in the tech and entertainment biz, starting at my first job with Jack Tramiel, these generalities are the work of comedy.

  64. LE

    Interestingly, Einstein was 5’9″. Marie Curie was 5′.I don’t see the relevance of that as we have no evidence that Einstein or Curie would have been any good at managing a modern corporation anymore than we have evidence that they could run NYC or do well as a Supreme Court Justice.

  65. einstein59

    speaking of jerks….

  66. Twain Twain

    People who engage in those hostile politics obviously don’t appreciate Benjamin Franklin: “By failing to prepare, you are preparing to fail” or the proverb “Practice makes perfect” which has been around since 1550s.

  67. LE

    Can get a lot of resentment and create a lot of office enemies showing up with a lot of good preparation.Well for one thing I don’t work for anyone so that isn’t relevant in my particular situation.When I was a kid I had a friend that said “you have to be the crazy driver that others watch out for”. So if being nice doesn’t work as a first or even second try [1] it’s shock and awe time.A better approach is to have an imporant project, work on it quietly, and when it is done show it as done.Agree a good strategy in many cases and I’ve done that “here is the finished policy, any comments” (at the late meeting where everyone just wants to get home for dinner).If describe the project before it is done, then can get some attempts to sabotage the project.Agree and definitely always take timing into consideration. Similar in a way to making Friday announcements to cut down on comments and interest. By Monday it’s old news.Even keeping the project relatively secret until it’s done can still result in a lot of resentment and hostile politics.Agree. And it’s all nuance. File under journal of “things you can’t learn from a book or a blog post”[1] In other words instead of being appreciated you are seen as being weak (this probably happens to women more than men I’m guessing).



  69. taqroohi


  70. Twain Twain

    Einstein and Curie’s heights are given in relevance to Arnold’s point of “saying you need to be tall to be smart” — not to whether we have any empirical historical evidence that they returned 100X ROI or could change the Second Amendment on the right to arms.The smartness of a person in one arena may / may not translate into another. The skills sets, interpersonal qualities and discipline (work ethic) vary between arenas.

  71. Twain Twain

    Paul Graham giving his ‘Counterintuitive parts of startups’ talk at the new YCombinator Startup Series for Stanford University:* http://startupclass.samaltm…What’s measured as “good” does depend on:(1.) From whose perspective?(2.) Is it referring to a characteristic or a decision.(3.) How it’s defined and interpreted.The lazy low-hanging fruit way is just to say it’s the polar opposite of bad, and we have a 50-50% of being one or the other.

  72. GS

    > My guess is that from his perch he can’t possibly differentiate between people being nice to him and sucking up to him because he is “PG” and what the same person is like to others (not saying it isn’t the same just that the possibility exists).Like some people here might be sucking up to Fred because he is “AVC” and want to get funded?

  73. PhilipSugar

    Really what Japanese company did you work for??? I worked for Mitsubishi Corporation and could not disagree more

  74. William Mougayar

    I dunno if one can totally generalize about Japanese vs. American styles of decision-making, but there are some subtle differences.

  75. William Mougayar

    ha funny.

  76. William Mougayar


  77. Richard Houston

    Please watch video and share to all your friends thank youViral video of little girl being attacked.Dad Sprayed with Mace..So Funny: watch and share