One of my responsibilities in producing software at Epic is planning for projects. This involves considering:

• Man-hours required to complete a given project spread across stages of design, development, code review and testing
• A projected deadline based on the project’s scope and number of full time employees

This is not something that my company has historically been successful at. Projects often require more man-hours than estimated or surpass a deadline due to unforeseen demands on our developer’s time (customer-escalations, bug fixes, etc.).

Although stochastic, these unexpected demands are measurable and can be forecasted to a reasonable degree of accuracy based on nothing more than historical averages.

After three decades as a software company, project estimation is a process that has yet to be perfected. This is a result of not effectively using the data available to us in order to determine the required man-hours for a project.

Project estimation is not the only area within my company that suffers from poor statistical review. Another area that would benefit from statistical analysis is project selection/prioritization within a given release. The success or failure of a particular set of projects (release) is not measured in any meaningful way.

Because project planning is a decentralized process performed by the lead or management group of each product, prioritizing resources over the course of an eighteen month release based on gut feel or intuition allows for widely
varied levels of success.

Intuition is the ability to acquire knowledge without consciously defining proof, evidence, or conscious reasoning¹. It is after all, nothing more than analysis performed in the mind of an individual, subconsciously drawn upon anecdotes or datasets that individual has perceived.

The longer I pursue my professional career, the more convinced I am in the value of data-driven decision making, and the benefit of painstaking diligence required to collect it.

By defining a product’s KPIs (Key Performance Indicators) and then measuring the change in those KPIs of customers that use those projects, the team can then measure the success of the recently released projects in impacting the measures they desired and evaluate their own ability to prioritize. In the next prioritization process, the lead can use that assessment to iterate on and plan for a future release. In addition to allowing for the measurement of the rate of continuous improvement, defining a set of measurable KPIs also allows for goal setting and transparency of the success of a team.

Take for example two development teams; one team defines their mission as,

The other team defines theirs as,

Which are you more inspired to work for? Which do you think will be more successful?

I’d choose the latter.

While collection of data can sometimes be tedious, it is immensely valuable, and the benefits of such a perspective
are not limited to one’s profession.

Friends of mine most successful at losing weight are not those that have followed trendy diets, but rather those that measure their caloric intake and determine what caloric deficiency is required to exhibit the desired weight loss.

This simple, data-driven algorithm is used to drive which foods are eaten and how much; it can be relied upon to provide consistent weight loss results.

In sum, I repeat a quote from William Edwards Deming,

Taylor White, August 19, 2017