Galorath Incorporated 2004 AccuScope – A Tool For Relative Comparisons Lee Fischman & Kenneth Fong Galorath Incorporated SCEA 2004 Manhattan Beach, CA
2 Ó Galorath Incorporated 2004 Presentation Outline AccuScope’s Role Key Features Four Step Sizing Process Accuracy Risk-driven estimates Size Anything Key Enabling Features Shareable Repository of Known Items Cut and Paste with Other Applications
3 Ó Galorath Incorporated 2004 AccuScope’s Role Scope Software and Other Items Earlier – provide reasonable size ranges prior to a detailed understanding of project More Accurately – relative knowledge often results in more accurate up front estimates vs. other methods More Efficiently – assist the sizing process with features making it easier and possible under more circumstances
4 Ó Galorath Incorporated 2004 Key Features Mathematically Sound Incorporates proven methods plus Galorath implementation Utilizes pair-wise comparisons against a number of known and unknown items Uses these comparisons to estimate unknown items Repository of Known Items Provides access to a repository of enterprise information on completed items Can reside on your desktop or be deployed across an organization Analogies Provide “logical patterns” for sizing new items from typical functionality Based on in-house experience, provides a common basis for characterizing scope Smorgasbord of Quantitative Metrics Source Lines of Code (SLOC) Function Points Function-Based Sizing And many more customizable metrics…
5 Ó Galorath Incorporated 2004 Four Step Sizing Process 3. Perform various expert judgment comparisons 1. Identify Items to size 2. Choose references or analogies for comparison 4. View / Use results
6 Ó Galorath Incorporated 2004 Step 1: Identify Items To Size Estimated items are those which will be sized.
7 Ó Galorath Incorporated 2004 Step 2: Choose References To Size Reference items are those from which comparisons are made. These items can be entered manually or obtained from a repository.
8 Ó Galorath Incorporated 2004 And (Or) Choose an Analogy Analogies are optional. Analogies are reference items based on past patterns. Choose from various sets of analogies, each representing different sizing perspectives, or create your own manually.
9 Ó Galorath Incorporated 2004 Step 3: Ready For Comparisons
10 Ó Galorath Incorporated 2004 Step 3a: Compare Known & Unknown Items Expert judgment establishes the relative size between ‘Tristate Net’ and ‘New Backup Site’.
11 Ó Galorath Incorporated 2004 Step 4: View / Use Results AccuScope lets you view sizing information and estimates using a variety of reports and charts. Reports are presented using different metrics depending on the size metric (source lines of code, function points, etc.) you are working with.
12 Ó Galorath Incorporated 2004 AccuScope’s Accuracy For above test: Perfect information: no errant comparisons Each bar above represents 250 randomized trials 5 reference items compared to 3 estimated items 7 stop slider range, 3 above ‘equal’ and 3 below
13 Ó Galorath Incorporated 2004 Factors Affecting Accuracy Number of items being compared – more is better, since that places fewer constraints on information available to the algorithm. Number of reference items – 1 can be enough, > 1 will improve results when comparisons are imperfect. Appropriateness of reference items – a sensible range of items spanning the same range as estimated items is helpful although not required. User’s assessments – the more accurate, the better. This is by far the most significant factor. Granularity of comparisons – more granular is more accurate BUT this results in increasing ‘cognitive overload’ for the user. Mapping between qualitative ratings and underlying ratios – the user’s conception of ‘Big’, ‘Bigger’, etc. should match the ratios they map to.
14 Ó Galorath Incorporated 2004 Least / Likely / Most Ranges Ratios have a range – so that estimates capture the edge of intuition between (for example) ‘Slightly Bigger’ and ‘Bigger’. User entries can have a range – so that the user can establish bounds on level of understanding. Reference sizes can be entered as a range – in case the size count is imperfect.
15 Ó Galorath Incorporated 2004 Range of Inputs Creates Estimates Across A Range Size By Comparison Inputs * reference values * slider entries * comparison ratios Lower Likely Highest Lower Likely Highest Outputs Estimated items as a range
16 Ó Galorath Incorporated 2004 Configurable For Many Estimating Challenges Configured out of the box for software estimating, AccuScope may also be used to: rank risk calculate hardware weight determine colleagues’ ages and so on…
17 Ó Galorath Incorporated 2004 Size Anything Using This Approach Estimates based on past known items, analogies, manually created reference items, etc. Can size ANYTHING: Hardware weight Software size Program office staff size AccuScope can access a repository of reference items, analogies can be customized Estimates are output as a range, for use in risk-based estimating Anything can be sized by relative analysis Percentage or qualifier (“Smaller”, “Larger”, est.) ratings
18 Ó Galorath Incorporated 2004 Repository Uses a variety of database types, including SQL Server or simpler, client-side solutions Repository may be shared across the enterprise, so many users can access the same information Online / offline modes Easy editing of repository entries
19 Ó Galorath Incorporated 2004 Cut and Paste With Other Applications Copy to and from Microsoft Excel and other applications Copy / paste integration with SEER-SEM Microsoft Excel AccuScope SEER-SEM
20 Ó Galorath Incorporated 2004 Summary AccuScope enables: Reasonable size ranges prior to a detailed understanding of the project Sharing of corporate knowledge Customization to a wide variety of estimating challenges