Your non-linear problem of 90% utilization



Does it feel like everyone is working very hard, all the time, and yet not accomplishing as much as everyone would like?

Maybe this is why.

Suppose a web server is running at 50% of its full capacity. Browser traffic doesn’t arrive in regular, smooth amounts; it comes in spurts and occasionally large spikes. Because the server is under-utilized, when a spike arrives there are spare resource to deal with the increase. If the spike is sufficiently large, performance will degrade, and if larger still, many of the requests will have to be rejected rather than answered; after all, there’s some limit past which the server cannot do any additional work.

Now suppose someone walks up and sees “50% utilization” and says “Hey now, this is a server, not a person, it costs the same whether we drive it at 50% utilization or 90% or 99%. So let’s get our money’s worth and drive it into the 90s!”

What happens? Even normal variations in traffic will drive the server into the red zone; the average time to respond to a request will skyrocket, and often requests will be dropped altogether. Not due to an unusual event, but all the time. The system is now brittle — not good for costs, not good for the quality of the product or customer experience — just bad all around.

Maybe we can drive high utilization by having multiple servers work as a team. Suppose we have three servers, all serving traffic for the same website, all at 70% capacity. That sounds like a happy medium between 50% capacity (wasting money) and 90% (brittle). The total amount of utilization is 2.1 servers (3 x 70%), so we’re nicely over-powered for traffic spikes.

But what happens when one server runs into problems? Suppose it crashes, or the power in its data center cuts out, or someone else breaks the network with a glut of garbage traffic. The 2 remaining servers now have to deal with 2.1 servers’ worth of traffic. Both are at 105% capacity, and we’re back to broken and brittle.

This isn’t about servers, it’s about you. It’s about how your “busy” life is not only diminishing your own productivity, but how your whole team is bringing each other down.

We all have a capacity, whether you want to measure it in hours, in energy, in focussed attention, or if you don’t want to measure it at all. Instead of web-requests, we have life-requests, whether those are inbound emails, JIRA tasks, ZenDesk tickets, Salesforce leads, requests from a friend, or families that need our time and attention even more than they need our paycheck.

90% utilization is causing more fail than you realize, not just in burn-out, but in productivity and output. Of course you’ll burn yourself up, sacrificing sleep, health, friends, family, and other things you mistakenly take for granted, but I suppose you knew that already. You’re trading that for super-human productivity, right?


But you won’t even receive outsized professional gains as a reward. This condition is a combination of frequent context-switching and interruption — the Twin Enemies of productivity. Work-completion will drag out because it’s constantly interrupted. Some will be abandoned.

Worse, in many organizations everyone is operating at 90%, which then reacts like the three-server system, where the inevitable hiccup from anyone causes a ripple effect that hurts several other people or projects. Since they are over capacity, rather than absorb the spike, they too will ripple the problem to others — a cascade reminiscent of the the run-away chain reaction of an atom bomb.

The key word there is “inevitable.” People get sick or die or leave or change. True emergencies arise that deserve to interrupt work. This is not something you can “architect out” of the universe; rather you need to build a system that assumes some amount of variation and interruption, and design your personal and team work-style to be resilient to that variation.

The ideal is probably a situation where most of the time you’re in the safe zone, with occasional surges into high gear for a short period of time and for a good cause. For example, a brand new product launch is usually attended by some extra time fixing bugs, especially post-launch where it hits real customers and a few issues are discovered that we all agree should be fixed swiftly before more customers encounter it. Or there could be a clear-and-present danger to the company that requires a special, time-bounded rally. Or you could use infrequent and brief surges in a fun way, like a Hack-a-Thon or a Bug Squash Competition or a Ticket Kill Day.

We’re all erring on the side of over-utilization, and rather than providing the benefits of competitive advantage through higher productivity, it’s creating needless churn without increased productivity.

Don’t let yourself, or your team, fall into the trap.

  • Well said.

    This is problematic at all sizes of company. The amount of burn-out in startups is appalling…wasting a lot of great talent and potentially great ideas simply because of a ridiculous maxim that dictates everyone should be working to midnight.

  • Great analogy, especially for busy people who get worried when they see their own utilization dipping *out* of the red zone and scurry to put more stuff in the queue.

  • Anthea

    This is just as true of home life – if you’re over-committed to activities and projects, the smallest jiggle in your routine can throw everything off.

  • One thing I have observed is that this problem is worse at a growing company – if you’re growing at something decent like 10%/mo, any excess capacity is decreasing at an alarming rate, so you have to get seriously ahead of it. I always joke around that a fast growing company is a black hole that sucks every available resource into it. The big internet companies have always been accused of being aloof/nonresponsive, etc during their high growth phases, and I have always thought it was because it’s super hard to keep up with the growth rate, so nonessential communications are the ones that get dropped, as in your dropped requests to the server example.

  • Peter Chester

    Oh man, this is definitely where my mind’s been at. So how do you solve for it? Plan to keep people 50% utilized all the time?

    • It’s probably not possible to be so specific as “50%,” but there’s probably a sense of when we’re “definitely over-subscribed,” both in a calendar sense and a getting-tasks-done sense. This is both scientific and difficult to accurately measure, or even determine the “unit” to measure, so it might in fact be OK to be aware and use a gut feel of how things are. Or, if you know you’re in the red, a goal like “an extra 2 hours/day without meetings,” that’s specific and sensible and moves towards a better situation.

  • This makes a good point. I worked in professional services for many years where utilization = money so the pressure was high. On the other hand nothing demotivates staff more than having too little to do. Great people love to do what they love so there is a balance.

  • Couldn´t think of a better metaphor for engineers … :-P

  • clement_911

    I think that we need to differentiate between “being busy” and “making progress”. I heard someone say that “being busy is the worst form of laziness”. On the other hand, I do think that healthy level of pressure/stress is needed for optimal performance. The trick is finding the right level and making sure it stays at the level!

  • Well said, Jason. The way I tend to think of it is… making sure I build enough margin into my schedule and life. When I set up my week in GCal, usually there are very few gaps or very thin gaps. I’m fooling myself because a) things usually take longer than we expect them to, and b) as you say, things come up. If we don’t build margin into our days, something is going to break. I agree that ‘inevitable’ is a key word here. We can either ignore that or acknowledge it and try to plan for it.

  • Scott O

    Years ago I read a novel titled “The Goal” ( in which a key concept is that systems need a certain amount of spare capacity to function optimally; that downtime that looks like waste is quite necessary and systems perform worse or not at all without it.

  • Adrijus Guscia

    Normally, I’d agree but looking at Gary Vaynerchuk, there are some people who can pull it off. I think it has a lot to do with Personality Type – Extroverts can handle more stimulation and gain energy by meeting people (thus they don’t lose energy in meetings!). Introverts do the opposite – we need alone time to recharge and lots of simulation drains us. This is huge difference and I think an advantage for Extroverts.

    Not advocating for or against, but it’s just an observation…

  • af

    The Pure Platinum Foamposite Pro, the latest Posite to ostensibly take inspiration from an Air Yeezy 2 colorway, was originally slated to launch March 12. Well, no longer. But rather than the usual infuriating delays, this time its the opposite; the shoes moving up on the calendar.The Pure Platinum Foamposite Pro, with its paled molded foam upper, tonal leather eyestay and vibrant crimson inner-lining, will hit stores kobe 10 high top on March 5. The shoe joins an endlessly impressive 3/5 release roster that already includes Air Jordans 1, 2 and 30, as well as the KD 8 EXT and Air Presto.Update (3/1): Too impressive a roster. The Foam Pro has now been bumped to March 4.In the next week or so Kobe will be passing MJ for 3rd place on the NBA all-time points leaders. MJ ended his career with 32,292 points and Kobe is less than 100 points away from passing that mark. Due to this upcoming accomplishment, the talks have nike air max 360 come up again on which one of the two is better. Taking a look at both players, they are both VERY SIMILAR in their style of play. One thing that sets apart the two legends is the number of rings they each posses. Michael finished his career with 6 NBA Championships while Kobe sits at 5, a number he will most likely stay at since the Lakers are off to one of their worst starts in franchise history this season and their future is not looking much brighter.Back in January, Marquette new new balance Basketball asked its fans to vote for a special NIKEiD colorway of the Jordan CP3.IX Marquette PE for the team to wear against Villanova on National Marquette Day. That day, apparently, is today, and this is the winning shoe.A sort of crackled, quasi-elephant print covers the CP3.IXs mesh forefoot, capped with a white leather heel and finished with navy blue and gold for that Golden Eagle touch.The Jordan CP3.IX Marquette PE wont see a public launch but, wonder of wonders and basic purpose of NIKEiD you can build your own right now on nike huarache run shoes In fact, here it is.The Nike Kobe 9 EM GS “Deep Garnet” is predominantly made of a deep garnet mesh upper throughout the low top silhouette. Hyper Crimson hints add to the sneaker’s theme while complimenting the design. Black worked into the mix give the sneaker some dimension as a metallic gold swoosh and tongue branding finish it off. Set to release December 13th let us know what you think of the sneaker below!Just to recap, the high-top silhouette is dipped in Challenge Red, while its overall air jordan shoe composition is inspired by the Red Viper, an extremely venomous snake. As implied, the model sports snakeskin leather along the heel towards its collar, while the textured upper embodies the snake’s pointy scales.Kobe Bryant has spent his entire NBA career with one team and one team only, the Los Angeles Lakers. During the first half of his career he wore the number 8 and after the 2005-2006 season he changed to his current number of 24.According to Nike’s official release calendar this colorway is being known as the Last Emperor. Paul George Shoes As you can see the Flyknit upper features Black towards the toe which then shifts to White around the ankle and on the heel. This colorway is supposed to commemorate Kobe’s past and present as well as look into his future. Bryant’s multiple championships, All-Star accolades and MVP honors are all embodied in this Kobe 11 which what shoes do stephen curry wear he hopes will inspire future heroes for generations to come.Upon closer inspection, the signature “BHM” pattern consists of an overlay that graces the rear which leads up to the collar and also sprinkles along its lower vamp. Furthermore, complimentary embroideries are found behind its ankle support.Buttery leather and bright colors beget striking summer style on the latest from sneaker customizer Fetti DBiasi.The Air Jordan 11 Low Pink Snakeskin rocks a premium black leather upper, matching up with the equally black liner, midsole and outsole. The faux snakeskin mudguard then sharply contrasts in vibrant pink, producing a remarkably balanced look whos stunning looks belie a relatively simplicity therere truly only two colors here.The pink snakeskins been done before on this model on 2013s release edition but the prevalence of nike air max shoes black here completely reimagines the concept. Take a closer look at the Air Jordan 11 Low Pink Snakeskin below, and head to Fettis website if you want to commission your own.The holiday moniker is additionally seen with more