Forecasting Projects and Customer Deliveries

Forecasting Projects and Customer Deliveries
No estimates, No speculation, No crystal ball gazing! 
A mathematically robust approach to fact-based planning for the enterprise scale projects and programs 

Traditional deterministic planning – analysis, decomposition and fine-grained, often time-based estimation – is part of the reason the Agile movement happened. Yet, in the pragmatic real-world, customers and other stakeholders still need plans and forecast delivery dates and commitments. Agile guidance to address these real-world concerns haven’t been fit-for-purpose: overly simplistic techniques based on metrics such as velocity have been failing product owners and scrummasters the world over.

In response to this gap, two alternative paths have emerged: do it harder, deeper, and with greater intensity – get better at estimating; and at the opposite end, the #noEstimates movement. A lot of mumbo-jumbo, hockus-pockus, and pseudo-mathematical black magic and witchcraft has entered the Agile planning domain.

In this 1-day class developed by one of Kanbanland’s expert mathematician, Alexei Zheglov, we offer you a fact-based, scientific approach to planning projects and customer deliverables in a lightweight and agile fashion using the mathematics of probability and risk management. You will learn to produce robust, mathematically sound, defensible forecasts and plans, using the minimal of effort. You’ll leverage real historical data to make better plans. You’ll learn just how many historical data points you need to forecast with confidence. You’ll learn the mathematical and scientific basis for every practice you’ll use.

Learn to plan like pro, make promises you can keep and lead projects with confidence!

Learning Objectives:

• There are 3 key objectives to this class:
  • Learning to forecast individual item delivery, setting appropriate expectations and creating Service Level Agreements (SLAs) for lead times
  • Learning to forecast projects or batches of work and monitor them while in progress
  • Learning how to validate forecasting models and modify them appropriately based upon feedback and real-world experience
• Understanding that there are two different sets of techniques needed for forecasting delivery of individual items versus aggregations – batches of work items such as projects and product MMFs (minimum marketable features)
  • Learning to forecast individual items:
  • Understanding lead time distributions and how delay contributes to the tail of the distribution and hence the lack of predictability on individual delivery
  • Understanding how to simulate delay and forensic analysis of lead time distributions
  • Understanding patterns and shapes of lead time distributions
  • Based on the approximate shape of a distribution, how many data points are required to have a sufficient sample and confidence in a forecasting model
  • Using Kanban “dive charts” to determine SLAs (service level agreements)
• Learning to forecast aggregations of work such as projects and minimal marketable product features (MMFs):
  • Understanding mode, median and mean and how the shape and tail of the distribution affect it – known mathematical ratios/patterns of mode or median to tail
  • Using sanity checks and “napkin forecasts” with techniques such as Little’s Law
  • Knowing when it is appropriate to use regression to the mean and how many data points are required for forecasts made using the mean to be meaningful
  • Learning the difference between “mediocristan” and “extemistan”
  • Understanding the exponential function as the inflection point in distribution functions
  • Understanding the three distribution domain areas of Pareto (“extremistan”), super-exponential, and Gaussian (both “mediocristan”) and knowing why the super-exponential region is a project forecasting sweetspot
  • Learning how to manage to keep real world results in the sweetspot, super-exponential lead time range
  • Understanding Cumulative Flow Diagrams (CFDs) and monitoring project health using them
  • Understanding the concept of “dark matter” – scope expansion
  • Understanding that dark matter is the biggest risk to any forecast and project schedule – typical ratios of dark matter mapped to project uncertainty and risk
• Learning when and how to use the Kanban Cadences as feedback mechanisms to validate forecasting models and adjust them as necessary

Who Should Attend?

This class is for anyone who is held accountable and responsible for customer deliveries and deliverables. If you have to make promises and commitments on scope, schedule and budget then this class is for you. It is appropriate for anyone with the following role: project manager; delivery manager; program manager; portfolio manager; product owner; scrummaster; or release manager.

This class is useful for anyone who wants to understand how to use existing historical data to make better plans and forecasts and for people who need to make promises they can keep and commitments in which they can have full confidence.

Prerequisites:

There are no specific prerequisites for this class. It is recommended that attendees have a project management or service delivery background or a role in planning, estimating, forecasting or risk management.

Course Outline:

Agenda

• Two modes: individual items and aggregate collections such as projects
• The morning will be dedicated to forecasting and expectation setting for individual items
• The afternoon to aggregated to forecasting deliveries of collections of work such as projects
• Understanding lead time distributions
• Using the delay simulator to reverse-engineer fake charts
• Developing forensic analysis skills for real charts

Break

• Patterns and shapes of lead time distributions
• How many data points do you need to have confidence in a model?
• Using Kanban “diver charts” to set expectations and define Service Level Agreements (SLAs)

Lunch

• Understanding mode, median and mean
• Mediocristan versus Extremistan
• The importance on the exponential function
• Three regions of interest: Pareto; Super-exponential; Gaussian
• Napkin forecasts and sanity checks: using Little’s Law (and when it isn’t appropriate)
• The super-exponential zone: how to manage your projects to keep you in the forecasting sweetspot (and reduce risk)

Break

• Monitoring project health using Cumulative Flow Diagrams (CFDs) and lead time distribution functions“Dark matter” and why it is your biggest risk
• Rules of thumb: observed ratios of dark matter against observable project risks
• Using Kanban Cadences as feedback mechanisms to validate forecasting models and adjust them based on observed real-world data
• Discussion: Insights from today’s lessons. How will you use what you’ve learned today? What will you change? How might you integrate the ideas with things you already do? What are the actionable, pragmatic ideas you are taking away and intend to implement? What impediments do you see to adoption? What will the benefits be, if you can switch to this mathematically robust approach to forecasting? What impediments do you foresee to introduction and gaining traction for this style of forecasting?

Class Materials:
All students will receive a read only student edit of the classroom Powerpoint, a copy of the 2nd Edition of the book, Fit for Purpose – How Modern Businesses Find, Satisfy & Keep Customers by David J. Anderson & Alexei Zheglov, and a printed copy of the Fit for Purpose infographic poster.