Email the Author

About the Book

Ever since writing computer software became a profession, people have asked how long it will take; And for this entire time, the answers given haven't matched reality as much as hoped. Due to the demand to improve this skill, a myriad of processes have been developed trying to put some rigor into getting accurate date and cost forecasts, and this book describes my thoughts on such a process called Lean Forecasting.

Assuming estimation of software projects upfront will resemble reality is like expecting a long haul flight with multiple connections will only take as long as the in-air flying time. By ignoring the time taken to travel to and from the airport, checking in, passing through multiple security checkpoints, dealing with immigration officers, waiting baggage arrival we would severely underestimate the total travel time. Throw in a few weather delays, headwinds, equipment failure, or other mishap causing a missed connection, and the concept of “just the flying time” for travel time is obviously an underestimate. Given we can understand there is more to flying than just air-time, why do we persist in thinking software “development and testing time” equates to software deliver time? We make this assumption every time we estimate software upfront using story points or any other process that largely ignores the complexity of the underlying delivery system. Lean Forecasting doesn't suffer from this assumption.

What is Lean Forecasting? Lean Forecasting is a set of techniques that help you perform rapid and accurate modeling of a software delivery system in order to answer tough questions about the odds of certain delivery dates, risk, staff size and skill ratios. This is a lightweight approach that adapts to any team and project size.

Traditional agile forecasting techniques are time consuming, erroneous (especially early estimates) and don't offer any substantial guidance into staffing levels and skill set balance. Any measure of certainty is also lacking in current methods. Either there is a single date with the assumption of 100% certainty, or a range that is so large (with the high date so far into the future) that projects get cancelled before they begin. The process described here outperforms standard forecasting in accuracy and level of detail with less effort. Lean Forecasting not only gives delivery dates and cost, it provides the likelihood of hitting those targets, and the impact of various staff level and skill set balance on these forecasts. Finding a staffing and budget forecast solution is easier than ever before. Tracking progress and departure from plan is easier than before.

Lean Forecasting

Lean Forecasting uses simple models and historical data (when available) to forecast likely future outcomes. These models allow rapid what-if experiments to be performed in order to judge the impact of changes to staffing, requirements and quality. I decided to call this technique Lean Forecasting after reading The Lean Startup: How Today's Entrepreneurs Use Continuous Innovation to Create Radically Successful Businesses by Eric Ries. His methodical process for performing experiments and altering course (what he calls pivoting) a business based on the quantitative results and customer feedback (data) resonated with my own thoughts. It was clear the same techniques applied to running a software project and process where the experiments were based on modeling would also yield significant benefit. In Lean Forecasting, I take the same principles and apply them to the planning, staffing and forecasting aspects of software development.

The concept of Lean Forecasting is simple: through early and continual small experiments you make validated changes to a process and team with confidence. Small experiments allow you to reduce the risk and initial pain of change. Early improvements have more impact because they gather benefit for a longer period of time. Pretty simple concepts, and I hope you can quickly grasp, benefit and expand my initial ideas.

What does Lean forecasting offer -

1. Rapid early life cycle forecasting of project date and cost (determine viability, compare options)
2. Ongoing tracking of plan versus actual project status (determine likelihood of on-time, prioritize problem areas)
3. Match staffing levels and skill set to a project at multiple phases and stages of a project (balance skill sets, ramp up and down at the right time, who to hire next)
4. Risk management integration with forecasting likelihood (is risk growing or shrinking, what is its potential impact of delay, what is the cost of that delay)

Lean Forecasting offers these benefits by using a variety of proven mathematical and statistical techniques, although much of this complexity is hidden from you. It achieves accuracy by eliminating conversion errors from various units of size measure (points, ideal hours, etc) to calendar time, and by modeling often missed effort. There is an emphasis on modeling the entire software delivery system rather than just the development and test time. Software delivery systems are complex systems, and perform in chaotic ways (small changes of inputs can have a non-linear impact on result). In an attempt to make software estimation easier and less time consuming a variety of techniques have been invented. However, in reducing the time to estimate, they also lost a significant portion of their accuracy.

Book Structure

This book has three parts -

1. The forecasting problems to be solved (Chapter 1)
2. The current forecasting techniques and where they fall short (Chapter 2)
3. All about Lean Forecasting (Chapter 3 onwards)

Chapter 1 introduces the need for forecasting. A lot has been written on estimation effort and waste, and this chapter outlines the broader business need and use of forecasts. A lot of passion surrounds this subject, and the thrust of this chapter is: If you are saying estimation is waste because it's hard and we are bad at it, then keep an open mind for a process that makes it easier and gives usable results.

Chapter 2 looks at the current methods used for estimating and forecasting software projects. It isn’t intended to bash current processes needlessly. In order to understand why current forecasts are inadequate, the cause of errors and inaccuracies in need to be understood and contrasted.

Chapter 3 defines Lean Forecasting and the benefits (cost and certainty) by employing it in your company. By the end of this chapter you will understand the goals and how to quantify the benefits of Lean Forecasting to your peers and managers.

Chapter 4 demonstrates in detail how to use Lean Forecasting for delivery date and cost forecasting. This is the entry level of the Lean Forecasting techniques, and is a great starting point to understand implementation and use.

Chapter 5 shows how more detail about projects can be captured and incorporated into forecasts. Whilst the initial forecasts built in Chapter 4 are accurate, there is little guidance as to what project factors/levers have the most impact. This chapter shows you how to model in more detail and build an ordered list of factors from most impacting to least impacting and to quantify these in time and cost.

Chapter 6 introduces how to make important decisions on staffing size and skill sets. Modeling the impact of work throughput (flow) because of staff skill imbalance within a team offers major schedule and cost improvements. By the end of this chapter you will understand how to analyze this and quantify the impact of different staffing solutions over different phases of a project.

About the Author

Troy Magennis

@t_magennis

Troy is an experienced executive who has been involved in many leading software organizations over 20 years. Most recently, Troy founded Focused Objective to build tools and training for simulating and forecasting software development projects, including the Monte Carlo techniques as described in his book Forecasting and Simulating Software Development Projects – Effective modeling of Kanban and Scrum projects using Monte-carlo simulation.