When people estimate how long a piece of work will take — or how much it will cost — their estimates are fairly reliable if they've done lots of similar work before. But if the work is unfamiliar, or if there are significant unknowns associated with the work, estimates of cost and schedule tend to be far less accurate, even when all we asked for was a "rough estimate." And the inaccuracies are almost always of a certain kind: final costs tend to exceed estimates, and deliveries are often late.
Why does this happen?
Last time, I explored three important contributors to this pattern — the planning fallacy, self-serving bias, and organizational politics. But in addition to these sources of systematic errors, there are two conceptual errors that are a bit mathematical in nature. One is a general misunderstanding of the term "rough estimate," and the second is a misconception about the shape of probability distributions of costs and schedules for unfamiliar projects.
- Misunderstanding the term "rough estimate"
- Rough estimates can be useful if we understand the true meaning of "rough." Technically, a rough estimate is an estimate that has a relatively wide margin of error. Every estimated value has a band of values that corresponds to, say, 95% confidence, and a mean or average that corresponds to a point within that band. If the width of the 95% confidence band is W, and the mean is M, then there is a ratio R equal to W/M. The "rougher" the estimate is, the greater is R.
- To be safe, users of rough estimates need to use the value at the top end of the band for estimated values in which larger is worse, or the value at the low end of the band for estimated values in which smaller is worse. Too often, people just use M. They can be forgiven, of course, when the estimator provides M alone, without reference to a 95% confidence band.
- There are clear reasons for this. In my experience, many consumers of estimates — the decision makers, as they're often called — are uncomfortable with any estimate that includes an uncertainty band. I've actually heard reports about decision makers' responses to estimates of the form, say, 8.7 million +/- 1.5 million, that go like this: "Don't give me this +/- stuff! If you don't know how much it will cost, just admit it! Now give me an estimate! Exactly what will this cost?"
- In fairness, Rough estimates can be useful
if we understand the true
meaning of "rough"many estimators lack sufficient information needed to compute a 95% confidence band. Others aren't sure how to calculate it. Others are intimidated by decision makers who reject the "+/- stuff," and still others aren't intimidated, but just don't want to fight with resistant decision makers. Whatever the case, estimates without 95% confidence bands are common, and because the estimates lack these bands, overruns are defined as "less favorable than the so-called estimate," which makes overruns very likely.
- The shape of probability distributions of costs and schedules
- Another source of estimation error is our human intuition about the shape of the probability distributions of costs and schedules. We tend to think of these distributions as fairly symmetrical around some average, or mean. And we also like to believe that the mean of, say, the cost distribution, is fairly close to the most likely value of that distribution.
- We base this intuition on our experience with familiar efforts. And as it happens, for familiar efforts, our experience is a useful guide. That is, the distribution of cost or schedule outcomes for familiar efforts is fairly symmetric around some average value.
- But for unfamiliar efforts, our intuition is misleading — sometimes grossly misleading. And this comes about for reasons that are easy to understand intellectually, but difficult to incorporate into our intuition.
- With unfamiliar efforts, or efforts that are subject to significant unknowns, the probability distributions for costs and schedules aren't symmetric about their means. In fact, they're wildly asymmetric. While there are some pretty hard limits on how quickly or cheaply things can happen, there are no analogous limits on how slowly or expensively things can happen. For example, no matter how hard you try, you probably cannot get to the conference room in the other building in less than 45 seconds — even if you run. But without trying to dawdle, it can sometimes take you 20 minutes to get there, even though it's only a five-minute walk, depending on the weather, who you run into on the way over, and whether or not you have to go back to pick up something you forgot.
- For unfamiliar projects, or projects subject to significant unknowns, the probability distributions for costs and schedules tend to have long "tails" corresponding to unanticipated costs and delays. These tails cause the distributions to have average values that can be much greater than their most likely values. And that's what throws off our intuitions. When we estimate, we tend to estimate the most likely value. And when the average value is much greater than the most likely value, our estimates tend to fall crazily short. The situation is even worse when we stack one task or project after another in a long train, because the uncertainty bands stack up.
Next time you make or receive an estimate without a 95% confidence band, ask yourself, "What are the lowest and highest values that wouldn't shock me totally senseless beyond all reason?" That should give you some idea of what 95% confidence means. Top Next Issue
Projects never go quite as planned. We expect that, but we don't expect disaster. How can we get better at spotting disaster when there's still time to prevent it? How to Spot a Troubled Project Before the Trouble Starts is filled with tips for executives, senior managers, managers of project managers, and sponsors of projects in project-oriented organizations. It helps readers learn the subtle cues that indicate that a project is at risk for wreckage in time to do something about it. It's an ebook, but it's about 15% larger than "Who Moved My Cheese?" Just . Order Now! .
Your comments are welcomeWould you like to see your comments posted here? rbrensDaBMTItJCwaKsgNner@ChacCrQTBGMzBwhIqYTXoCanyon.comSend me your comments by email, or by Web form.
About Point Lookout
Thank you for reading this article. I hope you enjoyed it and found it useful, and that you'll consider recommending it to a friend.
Support Point Lookout by joining the Friends of Point Lookout, as an individual or as an organization.
Do you face a complex interpersonal situation? Send it in, anonymously if you like, and I'll give you my two cents.
More articles on Project Management:
- Films Not About Project Teams: II
- Here's Part II of a list of films and videos about project teams that weren't necessarily meant to be
about project teams. Most are available to borrow from the public library, and all are great fun.
- Personnel-Sensitive Risks: I
- Some risks and the plans for managing them are personnel-sensitive in the sense that disclosure can
harm the enterprise or its people. Since most risk management plans are available to a broad internal
audience, personnel-sensitive risks cannot be managed in the customary way. Why not?
- False Summits: I
- Mountaineers often experience "false summits," when just as they thought they were nearing
the summit, it turns out that there is much more climbing to do. So it is in project work.
- On the Risk of Undetected Issues: I
- In complex projects, things might have gone wrong long before we notice them. Noticing them as early
as possible — and addressing them — is almost always advantageous. How can we reduce the
incidence of undetected issues?
- Should We Do This?
- Answering the question, "Should we do this?" is among the more difficult decisions organizational
leaders must make. Weinberger's Six Tests provide a framework for making these decisions. Careful application
of the framework can prevent disasters.
Forthcoming issues of Point Lookout
- Coming October 12: Downscoping Under Pressure: II
- We sometimes "downscope" projects to bring them back on budget and schedule when they're headed for overruns. Downscoping doesn't always work. Cognitive biases like the sunk cost effect and confirmation bias can distort decisions about how to downscope. Available here and by RSS on October 12.
- And on October 19: Bullying by Proxy: I
- The form of workplace bullying perhaps most often observed involves a bully and a target. Other forms are less obvious. One of these, bullying by proxy, is especially difficult to control, because it so easily evades most anti-bullying policies. Available here and by RSS on October 19.
I offer email and telephone coaching at both corporate and individual rates. Contact Rick for details at rbrensDaBMTItJCwaKsgNner@ChacCrQTBGMzBwhIqYTXoCanyon.com or (650) 787-6475, or toll-free in the continental US at (866) 378-5470.
Get the ebook!
Past issues of Point Lookout are available in six ebooks:
- Get 2001-2 in Geese Don't Land on Twigs (PDF, )
- Get 2003-4 in Why Dogs Wag (PDF, )
- Get 2005-6 in Loopy Things We Do (PDF, )
- Get 2007-8 in Things We Believe That Maybe Aren't So True (PDF, )
- Get 2009-10 in The Questions Not Asked (PDF, )
- Get all of the first twelve years (2001-2012) in The Collected Issues of Point Lookout (PDF, )
Are you a writer, editor or publisher on deadline? Are you looking for an article that will get people talking and get compliments flying your way? You can have 500-1000 words in your inbox in one hour. License any article from this Web site. More info