Gestalt Estimation

gestalt (noun)

an organized whole that is perceived as more than the sum of its parts.

One way of comparison is to take the project as a whole and compare it with other whole projects. If you’ve built a number of starter “business card” websites with little or no active code, you can have a pretty good idea about how much work the next one will be. The more you’ve done some sort of work, the more you’re likely to be able to estimate a project that’s similar to one you’ve done before. Or, perhaps, you might estimate relative to a previous one: “This seems bigger than the Foo Project but smaller than Baz.”

Dan North makes the point that experienced people in the software development field internalize this sense of comparison and don’t have to make explicit comparisons to past projects. He calls this blink estimation. He tells a couple of wonderful stories^[6] about estimating project staffing and duration. When asked how he did it, he replied, “I got a group of really smart people in a room, with at least 10 years’ experience each, and asked them.” In effect, he ran one round of Planning Poker (see ​Planning Poker​) on the project as a whole, based on the gut experience of people who had studied the problem and knew the development context. Humans can be quite good at balancing a whole lot of poorly quantified and tentative knowledge in a useful way.

Dan offers a number of cautions about this technique. He emphasizes the need for deep experience, and a diversity of backgrounds and disciplines in the group. He also cautions to be aware of cognitive biases. (See ​Cognitive Biases​.) The diversity helps fight groupthink, and everyone displaying their answer simultaneously guards against Anchoring Bias. It’s important to keep things loose and in the realm of estimation rather than calculation.

Could it be wrong? Of course it could—it’s an estimate! It’s a way to get close to an answer. If you’re nervous, you can cross-check with a second estimation technique.

One advantage of gestalt estimation is that it saves a lot of time. It also defuses the fractal nature of decomposition estimation. People pay attention to the factors that will materially shift the answer and don’t worry about the small perturbations.

A disadvantage of gestalt estimation is that you might not have a good comparison in mind. Or you might not notice some significant differences without looking in more detail. That’s why you might prefer comparison of smaller components.