Reinventing Business
Discovering Your Best Organizational Structure

Tuesday, June 12, 2012

Organizational Design Principles

Naturally I've been hoping that the perfect organizational structure will fall into my lap, much the same way Open Spaces Conferences appeared when I was attempting to create a new kind of conference. I've seen some inspiring experiments and variations on themes, but so far the structure that magically changes everything hasn't shown up (yet).

In the meantime, I have started discovering problem-solving tools that seem to be very useful when faced with organizational design challenges. These are not answers themselves, but different ways of thinking about the problem; often the shift in perspective can effortlessly produce the solution, or suggest a path to a solution.

This list is by no means comprehensive; I'm hoping to add to it as I discover more. Also some principles are more tested than others, and I'm happy to add things to the list that are just speculation -- when you're struggling with a problem you never know what might trigger an insight.

  1. Transparency

Instead of keeping things hidden, the principle of transparency opens everything up. When people understand the real issues the company faces, it enables them to make better choices. You'll be surprised at how many problems transparency will eliminate.

Since a hierarchy is primarily about communication, one of its main jobs becomes not just to communicate, but to control information. Transparency reduces or eliminates this job.

Transparency reduces repetition, and you don't have to worry about what you told one person in case they shouldn't know as much as another. You don't have to decide what people should know, which removes a significant burden.

People can be particularly uncomfortable about financial transparency, but often this is where the greatest opportunities lie. For example, suppose you run a bookstore and coffee shop in a resort town. During the off seasons you just break even or possibly lose a little money to stay open. During the high seasons you're so busy you have to hire people to work in the store, and for these people their only perception is that you seem to be making lots of money. By being transparent about the yearly financial curve, you can show employees how you might be just getting by, and thus how important it is to be careful about income and costs during the high seasons. By showing the numbers, you don't leave anything to the imagination which makes a much more convincing argument.

Here's an example of transparency (he also suggests "no surprises" as another maxim, although it seems to me that transparency implies no surprises).

  1. Experiment

Our brains are powerful speculation machines, so they can do imagination experiments without cost. However, without real-world experiments to check results and provide new factual input, they easily get biased and go off-track, producing invalid results.

Don't "know the answer." Discover the answer by trying things out. Create small, survivable experiments. Although it's great to follow perfect experimental procedure with a double-blind control group, sometimes you just don't know enough -- the problem is too vague to determine what to control. You'll still learn more by trying something than by guessing or deciding the problem is too hard to solve. For me, one example of the "experiment" principle is to move from just reading about business to visiting companies. The direct experience of talking to people who do things has a big impact on me. It's not the same as doing it myself, but it's much closer to the metal.

An excellent example of experimentation is the measurement of velocity in agile programming practices. Instead of deciding how long a project will take based on guesswork factors which are virtually never right, the team actually measures its development velocity during iterations, so you find out the truth -- even if it's less precise/perfect/complete than you want -- rather than making it up because someone demands a number.

When thinking about experiments, prefer "cheap, fast, survivable" to "expensive, long, risky."

  1. Alignment 

Much of economics is ... questionable. But at least two concepts seem to hold true: (1) People respond to incentives and (2) People won't usually respond in the way you expect them to (The Law of Unintended Consequences). Alignment asks the question, "Is my culture incentivizing the desired result?" I've mentioned before the way Costco gains all its profits through membership fees, and this aligns everyone's motivation to "customer delight."

For an especially bad example of alignment, consider stock brokers and financial advisers who take a cut of transactions -- this incentivizes them to simply get you to buy or sell stocks when they need/want money (it even has a name: churning), regardless of the actual effect it will have on your portfolio. You're much better off with someone who gets a salary. Even better would be a financial manager who only takes a percentage of your profits and loses a percentage of your losses, because their interests would then be perfectly aligned with yours, but I've never heard of such a thing.

Any kind of pay incentive can backfire, as documented extensively in Drive. Typically the only way you can really know whether you are aligning in the right direction is to experiment.

  1. Invert/Distribute Control; Bottom-Up Control

Ask the question, "What happens if I were to let the person at the other end of this process make the decisions?" By switching things around (even if it doesn't seem to make sense, or if you resent giving up some control) you might produce far more effective results. The basic idea is to find some variable and see what happens if you turn it upside down.

One of the most important ways of inverting control is to push the decisions as close as possible to their point of impact. Zappos allows representatives to make decisions about how to solve a customer problem (this often entails giving something extra to the customer) rather than the usual approach of requiring such decisions to go up and down the command chain. The command chain has more interest in short-term savings, whereas the representative has a strong interest in making the customer happy (which results in longer-term profits through happy customers).

Here's an example I heard about in Poland: The employee handbook is edited and improved by the most recently-hired employee. This is an excellent idea, because the new employee is not only the most interested and affected party, they are also reading the handbook for the first time and errors are much more likely to jump out at them. Knowing they are responsible for editing and adding to the material forces them to focus more. Also, they know exactly what is missing from the handbook; it's whatever they need to know and cannot find. When filling in the missing piece, the new employee doesn't assume the reader knows anything, whereas a more experienced employee might unknowingly leave out some "common" knowledge. Note that the employee's first task (editing the handbook) aligns with the company's goal that employees understand what's inside the handbook.

Consider Netflix' travel expense policy: "Act in Netflix's best interest." Instead of creating a booklet of expense policies and a group of bookkeepers and accountants to enforce those policies, Netflix just says "act in the company's best interest," which inverts control. The employee feels trusted to make appropriate decisions at the point of impact. The company saves all that money because it doesn't have the extra accounting group for expenses (that group might even cost more than the expenses themselves! Nobody does the calculation because we "must" have oversight on travel expenses). And what happens if someone seems to consistently abuse their expense account? That person is worth closer inspection -- perhaps they produce exceptional results for the company. Do you want to hobble them to conform to a standardized policy, and reduce their results so you can say that everyone toes the line? Or perhaps the employee is simply being abusive, and upon closer inspection you find that this employee is abusive in other ways and is not a good fit for the company. By giving free reign in a less important arena (travel expenses) you allow the employee to reveal important things about themselves and perhaps you can head off larger problems because of this (This might even be the seed of another principle, drawn from the idea of cheap, low-risk experimentation: "Find low-risk ways for an employee to reveal their true character.")

Zappos has two control inversion practices that stood out for me. First, customer service reps have (as far as I understand it) complete freedom to do whatever is necessary to make the customer happy. Zappos does not decide an "optimal call time" for reps; they can talk to a customer for as long as is necessary (a story they like to tell during the tour is a rep who talked to a customer for 8 hours. The customer turned out to be a competitive runner and no doubt told all his friends and other unmeasurable benefits). Reps will often add perks for a customer. Zappos also has a policy of shipping you replacements before they get the return shoes back, to produce a better customer experience.

I've heard a similar story about Zingerman's deli in Ann Arbor, MI (which has a group to train other companies in customer service). Something went wrong with a special order, and the rep got on a plane with the replacement to make sure it got there. A short-term significant extra expense created a dedicated customer for life (plus a good story).

Zappos' second control inversion policy has to do with product experiments. The ability to try something new is pushed as far down the relatively-shallow hierarchy as possible. So apparently a group manager can try an experiment. I believe I heard that one group tried belts (made of leather like many shoes) and that didn't work out. Someone else tried glasses frames (which initially don't sound like as good a fit as belts, until you start thinking "general fashion accessories") and apparently that has been a big success. If all such decisions had to be made at the top, neither of these experiments might have happened.

  1. Un-Control

Sometimes the answer is not to invert control, but to remove it altogether. Why do you think something needs to be controlled? Is it because someone else convinced you with an argument amounting to "We/Everyone has always done it this way?"
  • Consider experiments that show you what happens when control is removed without producing disastrous impacts.
  • If the human under control acts in a negative way during an experiment, is that behavior something that indicates fitness for the position? Perhaps it's better not to control behavior and to let the person's nature reveal itself through action, instead of controlling behavior in order to force inappropriate people to fit the job.
Removing control may not only be cheaper for the company, it could also produce better results by allowing people more freedom to experiment.

  1. Prefer Goals to Controls
As described in this post.
  1. Improve the Feedback Loop

Feedback is our tool for knowing how well we are doing, and adjusting our system accordingly (if you don't make adjustments, it's not a loop -- it's a suggestion box over a trash can). Just as the company is a system, the feedback loop is also a system, but a secondary one: Your company can be operating OK even with a broken feedback loop. Because feedback is a secondary system, it's easy to pretend that it doesn't affect the bottom line (in the short term, anyway), but I consider the feedback system to be one of the most essential things that keeps a company on track and healthy.

Feedback is one of the trickiest things to create and maintain, because we are fighting our own psychology (see The Mum Effect and The Dunning-Kruger Effect, for starters). We usually don't like hearing negative feedback, and sometimes we're even suspicious of positive feedback.

Positive feedback is easier to hear, and the results are often automatically positive. Happy customers recommend you to other people and you get more business. Unhappy customers typically don't go to the trouble of telling you they are unhappy, or why, because it requires effort and involves negative emotion. Instead, they just stop coming. You eventually get some kind of feedback but it's not useful or actionable -- your business goes down and maybe you go out of business, but you don't really know why. Inefficient feedback produces many sticky points in the so-called "free market," which is supposedly driven by the feedback of customer choice (but it's far easier to change coffee shops than to change banks, so the market is much freer in one case than another).

The essence of agile programming is improving the feedback loop -- onsite customer, daily standups, velocity calculations, Kanban boards -- all are improved forms of feedback.

When considering a new programming language, I read a lot of books. Most of these books are poorly written, and I sometimes consider writing a negative review, but fall back to "if you can say something nice, don't say anything." Eventually such books will be filtered out by the readers; unfortunately in the meantime a lot of people will struggle through them and often come away with a bad impression of the programming language because of a badly written book (I'm looking at you, Scala). It's also possible that new book authors will look at the existing books and believe that they represent a good model, then imitate it and produce yet another poor book. These authors do not get valuable feedback -- although I'm fairly certain most would ignore the feedback anyway (another reason not to give it) -- and their writing doesn't improve.

To solve this problem you must first break free of the loop of editors and reviewers. It's not that those people don't add value, but they easily become as insulated from reality as does the writer (Publishers are so distanced that they consider the buyer to be the employee at the bookstore who orders the books). Your ultimate goal is to satisfy the reader who wants to buy your book, so you need to find a way to get feedback from that reader. I unknowingly began to do this when writing Thinking in C++ and Thinking in Java because in both cases I started by creating a seminar, and the seminar evolved into a book. In the process, I road-tested the material dozens of times and each time I could see what did or didn't work, modify it, and test it again. Only in hindsight do I realize that I had solved the feedback problem by testing on customers. But I didn't have to set up any kind of formal feedback system; instead I could tell directly when something didn't work by seeing confused expressions on the faces of the audience (you can argue with editors and reviewers, but it's very difficult to deny those confused expressions). That very visceral feedback forced me to go back and rewrite.

To improve your feedback loop:
  1. Figure out who your actual consumer is -- try to eliminate all the intermediate distortions (editors and reviewers have a distorted view of reading, for example). The consumer isn't necessarily a customer in the traditional sense, but the person who is going to benefit the most from improvements in the feedback loop.
  2. Find a way to collect data that requires the minimum amount of effort for the consumer. In the best case (as during my seminars) the consumer isn't even aware that they are giving feedback. The least effective feedback systems require effort and emotional risks or investments on the part of the consumer, which discourages them from giving feedback.

  1. Proportional Investment

Don't hit a problem with a big hammer when a small one will do. Of course, you often don't know how big a hammer to use, so start small and work your way up using a technique like Five Whys. This principle can be considered a variation on low-cost, low-risk experiments.

  1. Autonomy, Mastery, Purpose

Apply the motivations from Drive not just to a person, but to the whole company. I'm really just guessing on this one, but it seems like things that motivates an individual might also motivate a group. It seems worth considering when trying to solve a problem.

  1. Pre-Project Checkin

Often it seems like the project or business venture goes for awhile -- perhaps even comes to fruition -- before you discover what the participants really wanted out of the experience. This discovery usually emerges when people don't get the thing they wanted. At this point you often make sense of some actions that seemed to be misaligned (at least, from your perspective).

What if we try to flush out misaligned goals before the project gets started? If this misalignment is diametric and unresolvable, it will likely prevent the success of the project or venture. Perhaps you'd be better off choosing another path. If this technique works, it could result in fewer failures, frustrations and wasted money and effort.

  1. Consistency

Humans are wired to be hyper-vigilant for fairness, and we are also constantly scanning to detect "the rules of the game," so we can play it better. If the CEO gets 400 times what the average worker is paid, no one believes him when he says, "we're all in this together." Instead, the message is, "those with more power get to take more advantage." In such a company, it should not be surprising when the point of being there becomes "fighting for power" rather than "making the best products and serving the customer the best way we can."

Although this point seems similar to Alignment, I would say instead that they are linked and depend on each other. Consistency is about general behavior whereas Alignment is about what the organizational structure incentivizes. Consistency means:
  • What we say and what we do are consistent
  • No trickery or lying is tolerated
  • We respect everyone: coworkers, customers, vendors, etc.
As a result, people learn to act in the best interest of the organization rather than only acting in their own interest. Put another way, people learn enlightened self-interest by participating in the organization, because they come to know that if the organization does better, everybody in the organization does better (not just those at the top).

A financial benefit to the organization is that people don't spend time either playing games or trying to figure out the games of others. More importantly, their creative energies are focused on making better products and services rather than figuring out the arcane unwritten secret rules of the company.