"Autumn School 2004: Estimating"

Event Report

The Autumn School this year was on that perennial bugbear of all projects - estimating. 34 attendees and P-G Committee members took a look at this subject over four weeks with speakers from varied backgrounds.

Don Southey of ComputaCenter, our South-West Group Chairman, led the school off with an overview of the subject and the various approaches available. He looked at the nature of the problem and its dimensions (i.e. the nature of projects, business objectives, novelty, definition, changes, risks, and assumptions), setting estimating in historical context by quoting a major over-run on a building project in 1760. Commercial pressures were obviously as strong then as they are now! He then looked at how to address the subject, including considerations of novelty in each exercise - how often is it the case that a project in the IT field is breaking new ground somewhere - either technically, in terms of the method to be used, or the business environment? He then covered the classic approaches - top-down, bottom-up, Delta (difference), Delphi and use of stored metrics. And then there's the question of presentation - do you present an estimate as a single point value, or as a range with associated probabilities? Of course, that depends on how well educated you Project Board is! He concluded with a look at uncertainty and a view on how to accommodate contingency and tolerance, and wrapped up with a short exercise to prove that bottom-up is usually better than top-down - if you have the data!

David Dunning, Director of Corporate Project Solutions, gave a guided tour of tool support for estimating in the second week's lecture. Having extended some of the points from the first week's lecture, he covered four products as illustrations of their type: Microsoft Excel; Risk Decisions' Predict Risk Analyser; Microsoft Project 2003; and CPS' P4P. Usually there's a need for input from multiple tool types - after all, an overall duration estimate for a project is not just the sum of the task durations - the dependency network is paramount and so you need scheduling support as well! He started with a Costs vs profitability estimate based on Excel - making the point that the humble spreadsheet is a good starting point for many people when estimating. His second demo looked at making use of a starting template - these could be organisation-specific or publicly-available, but can shorten dramatically the time to put an estimate together and can aid the thought process by suggesting work to be included. The third demo looked at the use of workflow technology - in this case P4P - to build whole plans and apply scaling and customisation to the task list. The fourth demo looked at the use of a Monte-Carlo simulation tool to prepare a profiled estimate for a given task network based on a number of iterations. The fifth demo looked at the facilities available in MS Project 2003 for tracking time and adjusting estimates-to-complete to take account of actual time spent to date. Given the frequency of new product launches in this area, this is a subject we may return to in future years.

The third lecture featured two Senior PMs from Oracle UK. Mark Reilly and Nigel Minett described the process of estimating for package implementations from a supplier's perspective - obviously a different angle from that of customers purchasing and implementing the package. They covered a range of techniques used in Oracle's work, with the advantages and disadvantages of each, and looked at the challenges that an estimator faces. These typically including a tendency to underestimate business complexity, and an 'unconscious bias' based on an individual's style. This can perhaps arise as a reaction to a demand to meet fixed deadlines or budgets, or by a wish to preserve organisational reputation. As in the previous lectures, they stressed the need for an informed base of information to build estimates on - 'standing on the shoulders of your predecessors'. Oracle has built such a base of data covering different organisation types and sizes, and it involves a close attention to analysis of risks. This overlays a detailed matrix of phases, processes and tasks in the Oracle Method, which has a number of 'routes' through it depending on modules chosen and the extent of customisation and changes/extensions. Finally they looked at some of the real-life situations they had encountered where things had not gone smoothly - not least including estimating for a package at Version 1 and having to implement using Version 2!

Grant Rule of Software Measurement Services Ltd, a very interesting and eloquent speaker, presented the fourth and final lecture. The topic 'Improving the Delivery of Business Value using Functional Size Measures' was covered in an entertaining and informative way. This is a method of estimating based on the number of functional requirements allowing the measurement of output value as size measure will help management and allow more accurate predictions. Grant led us through the history of functional size measures and explained how they have moved forward over the last 20 years. He covered the values that need to be measured such as duration, effort required and illustrated it with simple examples that nicely outlined his point. He also added that software developers have an additional responsibility to focus on adding value and removing waste and defects from every step of their processes. As he puts it 'the software community cannot go on delivering projects late and over budget.'

Our thanks again to Oracle UK for hosting us - we are always looking for organisations that can support us by providing a venue for our meetings! We are repeating this School in Bristol, starting on Tuesday 22 February.

Slides from the presentations given at the school will appear at the conclusion of the Spring School in March 2005.