Why Do We Estimate…Anything?

In nearly every one of the free, monthly webinars that I give each month, I start off by explaining why my freely licensed tool, Statistical PERT®, exists. It’s the same reason why we estimate, because Statistical PERT (SPERT® for short) is an estimation tool and technique.

As I thought about it, I decided to abstract and consolidate the reasons into two.

The first reason we estimate is to align expectations. You and I have to understand the nature of something that’s uncertain. If you ask me for an estimate, I have to do the legwork to prepare an estimate for you. Once I’ve done that, it’s important that you and I align on the estimate we’re both now looking at.

You need to understand the sense of uncertainty I’ve gained by preparing the estimate for you. You need to understand what my process was, what my inputs were, and to what extent my personal judgment comes into play (am I an expert in this particular field? or am I estimating something far outside my field of experience?).

This alignment process is advantageous to both of us. If done properly, I am not going to sugarcoat the sense of uncertainty I have. If done properly, you are going to explore the source(s) of my uncertainty, both for the purposes of understanding them and to question them. It’s okay to question how I prepared an estimate if you first take the trouble to understand my estimate.

The second reason we estimate is to make better decisions in the face of uncertainty. Why bother having that great alignment discussion about my estimate if you (and I, and/or other people) aren’t going to do something with that information? Usually, I create an estimate so some decision-maker can make a better, more informed decision in the midst of uncertainty.

If I don’t prepare an estimate for you, and you make a decision about something that’s uncertain, on what basis are you making your decision? How do you even know whether or not your decision is a reasonable, even rational, one?

When I do conference talks on this, or when I teach in classroom, I usually give the example of estimating at what time should my wife and I leave our house and get to the airport to board a flight somewhere.

I prepare an estimate that takes into account the factors of the problem, and an assessment of what’s uncertain. The fact is I live about 20 miles north and south of two different, international airports (I usually fly out of FLL but sometimes PBI). In both cases, the most likely route to the airport is by driving on I-95, the major north-south corridor that stretches from Miami to somewhere in the northeastern United States.

Other factors that come into play is the date and time of our flight (will we be driving on I-95 during rush hour? Is it a holiday weekend?). Who’s traveling with us (the more travelers, slower progress is made)? How long are my wife and I willing to sit at the airport gate waiting to board (I’m more tolerant of long waits at the gate, my wife less so). How willing are we to miss our flight and rebook on another flight?

As I assess all these factors, I can build a probabilistic estimate for what time to leave our house and arrive at the airport gate. But wait–I’m missing something. I haven’t assessed other uncertainties involved in traveling to the airport. It’s not just the traffic on I-95. How long will it take to drop-off our car at the off-airport parking lot and catch a shuttle to the airport? How long will it take to check-in our bags? How long will it take to go through the TSA security check? All of these are other, decomposed uncertainties underneath the parent uncertainty which is, “How long will it take to get to the airport gate?”

If I share my estimate of how long it may take to leave our house and arrive at the airport gate, my wife and I can align on that travel time estimate. What were all the factors that I took into account? What are all the uncertainties involved? Maybe during the alignment process, my wife will discover some important factor that I’ve overlooked, or she’ll question something about how I’ve assessed the situation. That’s actually a good thing; we both want to optimize the next step, which is deciding on what time we want to leave our house and head to the airport.

Once she and I align on the uncertain travel time estimate that I produced and together we refined, now we’re in position to optimize our decision making. Fortunately, we’re both risk-averse so choosing a house departure time that gives us a 98-99% chance of successfully boarding the airport gate on-time is a goal we both share. In fact, although we have had a few close-calls in nearly 30 years of marriage, we’ve never missed an airport flight…yet. I’ve probably jinxed our next trip.

This is why estimating is an important skill to develop. Estimating helps people align expectations so they can make better decisions in the face of uncertainty.

Leave a Reply

Your email address will not be published.