Breaking The Wheel

A picture of Dwight Eisenhower, a man who understood the paradox of planning all too well

The Paradox of Planning

One of my favorite quotes revolves around planning.  It came from Helmuth Von Moltke, a 19th century German Field Marshall: “No plan survives first contact with the enemy.”1. The implication here is simple enough: the plan that makes total sense on paper quickly falls apart when confronting the entropy of reality. And yet planning is essential for getting a team moving in the right direction. As Dwight Eisenhower said, “I have found that plans are useless but planning is everything.” And thus we arrive at what I like to call the paradox of planning: planning is the act of creating something that is simultaneously infinitely valuable and completely worthless.


By Reading This Post, You Will Learn:

  • Why planning, in the waterfall sense of a sequence of events, it highly problematic
  • What tool to use instead for long-term project management
  • Why models are better than plans
  • The pitfalls of using models
  • The contexts in which planning is still useful

The Problem with Planning

Plans have their uses – aligning a team towards an objective, or determining if a course of action has some semblance of practicability. But they have three major downfalls.

First off, plans are expensive. Planning a major effort takes a lot of time: mapping out all of the activities, identifying dependencies, and taking a swing (or SWAG) at estimating schedules. Second, plans are labor-intensive to update. Changing the outcome of any element has the knock-on effect of jumbling up everything else, like trying to move a brick in a wall. Third, plans tend not to have much tolerance for failure. As Michael Douglas’ character, Remington, said in The Ghost and the Darkness “We have an expression in prize fighting: everyone has a plan until they’ve been hit.”

Plans, by their nature, tend to be rooted in optimism: “We need to accomplish this thing. How can we accomplish the thing? Here’s how we accomplish the thing. Great!” And when portions of that optimism prove to be unfounded…then what?

Further, as I mentioned in “Game Planning With Science“, there is a distinct difference between precision and accuracy. Precision is a measurement of your margin of error, while accuracy is an assessment of whether the actual outcome fell within that margin. And plans are a tool of precision – you are trying to nail down a specific date/time by which an effort will be completed or, at least, a specific sequence in which things will happen. And experience (and a study of history) have brought me to an inescapable conclusion: precise plans tend to fail in catastrophic ways. We’ve all experienced this. Whether it’s a travel schedule, an agenda for the the day, or Operation Market Garden. Once the first domino falls, the rest tend to tumble after it.

So I’ll take accuracy over precision any day, for one simple reason: I would rather be correct than specific.

What I Prefer To Do

Rather than going to the time and expense of planning the waterfall sequence of events known as a plan, I’d rather take that time and invest it in building a model.

What do I mean by a model? Simply put, I want a data-driven approximation of reality. Something that accounts for the the work we think we need to do based on our most current understanding of the game (scope) and takes the rate at which we tend to consume work (velocity) and spits out some time frame in which all of that work will be consumed.

Or, in simple arithmetic: Scope/Velocity = Time

Models are less precise than plans, for sure – as stated above, models are an approximation of reality. An approximation accompanied by a margin of error. But that’s a strength, not a weakness. The margin of error inherent in a model makes it better able to absorb unexpected events without collapsing. Plans tend to have zero margin of error, meaning that any deviation tends to send the whole construct into dissarray.

Obviously, I’m drastically over-simplifying. How does one estimate the scope to be done? Or measure the velocity? That is (tongue planted firmly in cheek) out of scope for this post but (wouldn’t you know it?) some nice man spent weeks of his life detailing those very topics here.

The final analysis boils down to this: your plan will (almost) inevitably be wrong. And, if you know your plan is going to be wrong, it doesn’t make a lot of sense to invest a lot of time on assembling it or maintaining it.


Resources That Informed And Influenced This Post

If you have ad blockers turned on, you may not see the above links. Breaking The Wheel is part of Amazon’s Affiliate Program. If you purchase products using the above links, the blog will get a portion of the sale (at no cost to you).

Why Models Are Better Than Plans

Models carry significant advantages to plans:

Models Are Faster To Update

If you’ve taken the time to build a proper model, then updating it in the face of new information simply requires entering new data. Scope increase? Add the data. Adding people to the project? Add the data. Velocity lower than expected? Add the data.

Further, because they are so much easier to update, it’s also easier to spot deviations from the forecast. If your forecast called for churning through 40 points of scope a week and you’re only chewing through 32, it’s much easier to a) identify the shortfall sooner, b) understand the impact faster, c) observe if mitigation tactics are actually having the desired effect.

Models Are More Robust To Failure

If the end results don’t match the desired or foretasted results, you don’t need to scrape the plan and spend days or weeks re-rigging everything. Just update the data in the model.

Models Are Easier To Communicate

Rather than walking someone through a tedious and confusing chain of events, you can just say “Based on or current data, here’s what we think will happen.” You can also, as the saying goes, show your work. Rather than having to rhetorically argue about what can and can’t be done, you can simply say, “Here’s the data.”

Models Are Quanitifiable

If you are forecasting development based on actual data, you can quantify your forecast. Instead of saying “I am highly confident this plan will work”, you can say “We feel 95% confident that the target date falls in this window of time.” where neither the confidence level nor the window of time are guesses.

Models Aren’t Without Their Vulnerabilities

Models Encourage Finagling: A Morality Tale

Everybody likes to bag on accountants as being the squarest of the squares. But I’ll let you in on a little secret: accountants run the world. How? A little technique know as “earnings management.” You see, businesses are complex, amorphous entities, so it would be impossible to create hard-coded set of accounting rules that could apply to everyone. Thus accountants tend to operate under what are called “Generally Accepted Accounting Practices,” or GAAP. Basically, there’s a large umbrella that the field consisders legal and ethical, and, within that umbrella, accountants have wide latitude to do their work.

Here’s how this can play out: Corporation X needs to maintain a debt/equity ratio of 25%. But the books currently say they’re running a D/E of 45%. Bill, the CEO, stats sweating – he needs to hit that target or the board will fire him. So he goes to the head of the accounting department and says, “Steve, help me out here.” Steve goes and checks the company’s finacial statemens and notices that the company’s pension plan currently estimates that the average retiree will live to the ripe old of age of 90.

Buuuuuuuuuuuuut….if the pension plan instead estimated a life expextancy of 80, that would take $73MM off of Company X’s liabilities. Now, the D/E ratio is down to 23%. Bill gives Steve a big clap on the back and huge year-end bonus. Great work, Steve!

The Moral of the Story

The scenario I’ve described above is 100% legal. Ethically dubious, to be sure. But legal. And accountants do this kind of crap all the time. Recatorgarize an expense here, adjust the cost of goods sold there. If large corporations are the dominant force in a capitalist society, then accountants are the worlocks who ensure those corporations reap the largest windfalls possible.

And this one-act play above presents the moral hazard of runnning with a model. It’s REALLY tempting to fudge the numbers, especially when you need to report them to a boss, publisher, or investor.

Garbage In, Garbage Out

“GIGO” is a saying in the various fields of decision science: garbage in, garbage out. You can have the most robust, refined model in the world. But if you give it garbage data, it will give you garbage result. Examples of garbage data:

  • Measurements made in an inconsistent fashion (Some weeks you measured scope using story points, other times you measured it using estimated hours)
  • A mix of actual measurments and guesses/filler (you forgot to collect data the third week of July, so you just slap some numbers on that bad boy)
  • Data reflective of a situation that no longer exists (example: you take the data of Team A and just apply it to Team B with no forethought as to the differences between the teams)

“Discipline equals freedom”

There’s my favorite Jocko Willink quote again. In order for the model approach to work you have to be disciplined about:

  1. Tracking the work you want to do
  2. Estimating the scope of work in a consistent fashion
  3. Tracking when you close individual units of work

There Is Still A Place For Planning

The take away from this post is not “never plan.” That’s as dumb as planning everything in detail. The point is that you should invest your time in the areas that provide the most value. Plans tend to provide zero predictive value. But they provide massive strategic value. Plans align teams to a strategy and a goal. So I plan at a strategic level – what are the major goals of the project? What are the most important questions we need to answer? How can we answer them as efficiently as possible?

For instance, I’m currently starting a new game in my own company. Rather than plan out the whole project in detail (again: pointless), I am, instead, focusing on the general time frames I want to spend in the various phrases of production (3 months in incubation, 6 in pre-production, etc) and the critical questions I need to answer during those phases. Is the core mechanic fun? What is the right number of players? What’s the ideal session length? Once I know the question I need to answer, I can list out the tasks that I think will answer that question, and then use a model to forecast timelines. This sort of strategic planning has the added benefit of being very easy to communicate: “Here is the current goal.”

Plans are also useful in the short-term. What do we need to get done this week? In what order should we handle a group of tasks to best account for dependencies? But there should be an inverse relationship between the precision of the plan and the length of time it encapsulates. You can plan a day in detail, but longer-term plans should be abstracted to a strategic/project goal level.


Further Reading If You Enjoyed This Post

Video Game Statistics: A Primer – Game Planning With Science! Part 3

Planning Games Using The Central Limit Theorem – Game Planning With Science! Part 4

Scheduling Video Games Scientifically! – Game Planning With Science! Part 7


Summary

If all you have is a plan then everything looks like a waterfall. So add the concept of models to your production aresenal. As you manage a project over the long term, think carefully about where it makes sense to dig into the weeds, and where it’s more appropriate to leave a margin of error. And also keep a thoughtful watch on how much time you are spending on any aspect of management versus how much value that activity provides. If you know a plan is going to fail, then it makes zero sense to put all of your managerial efforts into maintaining it.


Key Takeaways

  • Plans are valuable, but they are expensive to create, labor-intensive to update, and not robust in the face of failure
  • There is a direct correlation between the precision of a plan and the likelihood that it will fail catastrophically
  • Models are more valuable for long-term planning, because they are easier to update, easier to communicate, and more robust against failure
  • Models do come with their own set of piffalls which must be carefully mitigated
  • Planning has a use, but the amount of precision should be inverse to the length of time the plan covers
  • A plan for a day or week can be detailed to a high degree of precision
  • But a long term plans should be abstracted to strategic goals

  1. Okay, the actual quote was “No plan of operations extends with any certainty beyond the first contact with the main hostile force.” But the abridged version is the more popular version.

If You Enjoyed This Post, Please Share It!

Creative Commons License
“The Paradox of Planning” by Justin Fischer is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

Leave a Reply