Maybe 68%, 95%, or 99.7% is Better than 100%

I really want to believe that work is like Chess. I love chess’ simplicity, complexity, and strategy. Its rules are defined, and even the most basic computer opponent can beat me without much effort. What is beautiful about chess is the allure and perception of getting very close to 100% confident in the right next move. My work isn’t like chess, and sometimes it feels like my weekend are spent refreshing that perception only to be dashed by neverending waves of Monday, Tuesday, Wednesday, Thursday, and Friday. Why do I hold onto the perception of perfection? Because I want to be 100% right. This doesn’t work at home with my wife or kids. Why would work be any different? The short answer is it doesn’t.

Work (and life) is more like Poker than Chess. It is more like placing bets and managing probabilities. It is about prediction and forecasting. It is managing your and others’ insecurities, irrationalities, and inadequacies. Therefore, shattering the 100% perception forces us to manage life’s probabilities.


This is where the Empirical Rule (or 68–95–99.7 rule) provides more sanity as you meander through life. What is the Empirical Rule? Given a normal distribution (or Bell Curve), 68% will fall within one standard deviation (1 sigma or σ) of the average, 95% will fall within two standard deviations (2 sigmas or 2σ), and 99.7% will fall with three sigmas or 3σ. Said another way, if an event occurs on average daily, an event will happen once every about twice a week 68% of the time, within three weeks 95% of the time, and within a year 99.7% of the time. For those mathematicians out there screaming that life isn’t a normal distribution, you are correct, but that misses the point (See Chebyshev’s inequality for nonnormal distributions). The point is that getting to 100% is really, really hard. Actually, going from 95% to 99.7% is over 10x more difficult than going from 68% to 95%.

So why are we so confident we can achieve 100%. I blame modern manufacturing. Over the past three decades, the efficiencies we have seen have been nothing short of astounding. For example, companies like Motorola have popularized six sigma, allowing only 3.4 defects per million, and they have reduced their defects rate by more than that.

Such a low defect rate is only possible with machines. The process must be standardized and the end product must be a commodity. If you ever have a chance to go to a manufacturing facility, you will notice something missing, people. About a decade ago, I got to go through a facility that made french fries, and I remember seeing 1 to 2 people and a bunch of machines and conveyor belts. The workers I saw were responsible for maintaining the machines. They were babysitting the perfected multi-million dollar mechanical process.

On the other hand, in knowledge work, we toil with an overabundance of imperfectly wonderful people. People doing their best to do a good job and move their company and career forward. And all the work they are doing can’t be done by machine. Knowledge work is creative and complex. It is not a commodity.

Within every business is neverending change. Step back and you will and you will quickly realize that 100% isn’t even a possibility. How long does it take a department or process to break down or have a defect as a knowledge worker? For the sake of argument, let’s assume that a breakdown happens daily on average. I know it wouldn’t take me long to make a mistake in a week. Conservatively, I bet I have a “defect” about two times a week. This is when I have to clarify an email, fill in a missed gap, or provide additional clarity. In other words, if my conservative estimate were accurate, my internal error rate would fall within one sigma of the daily average defect or within a 68% range. The chart below shows the daily error rate in 1/2 sigma ranges for a daily event.

Wikipedia Daily Event Frequency

If you analyze most knowledge work breakdowns, they are due to communication issues and not the systems or processes. Let’s say we clear up the communication problems. We might be able to make it three weeks or two sigmas, which might be doable. However, it would not be advisable to shoot for a three-sigma defect rate because that would go an entire year without error. This rigidity might put you out of business. Business moves too fast to be 99.7% accurate.

Businesses can still use the empirical rule, but it needs to be tweaked slightly. First, one sigma communication breakdowns should be cleared up by the front-line employees as quickly as possible, with the goal being to catch them quickly. This is why a daily scrum can be effective. Secondly, two sigma errors should not make it to your boss because of the rigor of the front line employees. This correlates to running effective two-week sprints. Lastly, anything that has a duration of greater than two sigmas should be a process and system that the employees work in. The processes should be general and flexible enough to bend without creating defects.

In short, clarify in the short term and systematize the long term. The creation process is messy and full of twists and turns, and the system that the creation process lives in can be flexible, steady, and stable. Let’s go be empirically better.




I like to make products and processes elegant

Love podcasts or audiobooks? Learn on the go with our new app.

Recommended from Medium

A practical guide to diversity in business

8 Reasons Why I Love Remote Work

The Essence of a Great Corporate Culture: Love and Learning

Journey from a Service based company to Microsoft!

Why you need to be a Digital Mentee in 2021

Credit for the image to

Negotiation — How to Ask for What You Want (And Actually Get It)

A view on a Hyper Island journey from anxious eyes

Why you need interns and how to find them | Sourcr

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Chris J Terrell

Chris J Terrell

I like to make products and processes elegant

More from Medium

Most Useful Distributions to Remember

5 insights of Seattle’s Airbnb open data for better travel decisions

Journey from Correlation to Causation

Feedback series part 2: Build a habit of asking for feedback