Новую программу проектов в компании SanDisk (производство flash-накопителей) неформально назвали «гориллой» из-за ее объема (более 5000 задач). Продукт нужно вывести на рынок к жестко заданной дате. Как выполнить такую программу и как замотивировать менеджеров проектов? На этот вопрос ответил Эрик Ютеваал (Канада), автор инструмента «Критический путь 2.0», позволяющего с успехом реализовывать такие проекты.
26. Critical Path End Date
Accumulated Leveling
Delay = 72 days !!
Pre-leveled Post-leveled
Name
Finish Finish
System Qual Complete, Rev. A 10 Jul 13 20 Sep 13
27. Critical Path Analysis
MS Project 1 CCP 2
Unleveled Critical Path Yes Yes
Workload-Leveled Critical Path No Yes
Elapsed durations (ed) No Yes
Multiple Calendars No Yes
Logical Dependencies Yes Yes
Resource Dependencies No Yes
Most-Critical Path No Yes
Longest-Path No Yes
1 Critical Path 1.0 compliant
2 Critical Path 2.0 compliant
31. Determining the High-Confidence Date for
the Program
Task Duration Estimates:
– 50% Confidence durations
– 90% Confidence durations
MS Project Forecasts Finish Date (FF)
– Based on 50% confidence durations
High-Confidence Forecast Finish Date (HCFF)
– Including 90% confidence durations in Monte Carlo simulation
– HCFF - FF = Program Buffer
Only 1 buffer = Program Buffer
34. Time Simulation Results – Overview
Proof of Concept Schedules
System Qual Complete, Rev. A
Unleveled, forecast: 10 Jul 2013
Leveled, forecast: 20 Sep 2013
High Confidence: 09 Jan 2014
Program Duration: 542 Calendar Days
Leveling Delay: +72 Calendar Days
Program Buffer: +85 Work Days
Critical Handoffs 4 out of 65
PMI 2012 North-America Congress in Vancouver (Oct 2012):A = 2 B = 20C = 3 D = 14
REMARK: isn’t this a beautiful curtain of cross-project dependencies that makes all detail tasks invisible. In the 1990s, the screen resolution was so poor that the entire screen would turn black!
REMARKS: Delegate: more accurate model if scheduler trained wellMore than 2,000 tasks: a schedule with 2,000 tasks in it is a full-time job!!
QUESTION: which of these three were our biggest headaches? Raise your hand, whom of you think it was the 1st, 2nd , 3rd? Unexpectedly, it was nr 1!!!! This was because subproject schedulers scheduled for different reasons:Program delivery: conforms to checklist in my bookManaging resources: WBS not deliverable-oriented, resource vacations inserted as “tasks”, incomplete logic, logic on summary tasks. At SanDisk, the SW guys had schedules like this and were not willing to change them.Portfolio tracking: only needs to produce performance indicators (and explaining detail)
REMARK: you never know where the program Critical Path runs; we have been surprised more than once, particularly when you start simulating the IMS!
REMARK: 1. Is what most program managers are currently doing; very labor intensive, very error prone, often lack of trust in the overall program schedule
REMARK everybody is doing this: coding/labeling the handoff deliverables in the subproject schedules; “What do you need from whom?”We recommend using descriptive labels (rather than cryptic codes), but they have to be unique (create master list!!)
REMARK Each deliverable handoff becomes one line item in this spreadsheet; the spreadsheet becomes the dashboard for the program manager!
Xxxx can we find picture of Teamplanner view with optimized workloads (no empty spaces)?REMARK: You can see when resources are perfectly balanced when there are many Critical Paths that are equally critical.
DEMO resource-constrained IMS??!!
Data we are showing here is from Proof of Concept files only; obviously we could not present real-life data!PoC was realistic data but not real data.
REMARK the green line depicts the leveling delay
REMARK:MS Project does not do Resource Calendar and Task Calendar exceptions, only Standard (Project Calendar) exceptions! As per Oct 2012, CCP does the most-Critical Path but not (yet) the obscure situations of the Longest Path analysis (SNET on late deliveries without links to earlier tasks); does Primavera do these?????
REMARK Al programmed the Monte Carlo simulation in his own time in the evenings; he is a brilliant guy! We needed this because in a workload leveled IMS, the picked durations need to be entered into the schedule, the workloads need to be leveled by MS Project and the duration can then be determined. We need to add still that it calls our CCP tool to find the RCP in each scenario, so we can determine how often each task appears on program critical path.
REMARK:CSF = Critical Success FactorFor schedule reviews: several iterations were necessary!!Pilot program: Started courses in June 2012, first schedule reviews in July, first integration in mid October!!
REMARK typically, we do this every week!
Monte Carlo uses 50% and 90% range of dates!! SanDisk calls:50% = Aggressive But Achievable (ABA)90% = High Confidence (HC)NOTE: We follow Critical Chain in that:We use workload-leveled Critical PathWe try to prevent multi-taskingNOTE: We divert from Critical Chain because : We use ABA andHC estimates (not just 50% confidence level)We use Monte Carlo simulation to determine buffer we need (not just 50% of Critical Path duration)!We only use one buffer and do not insert buffers on secondary, non-critical pathsWe do no ALAP scheduling
What was the initial forecasted date in the schedule? Oct 7, 2013 (which is off-the-chart; not listed) !!
Al: does the 542 cal. days include the 72d leveling delay AND the 85d program buffer? No
NOTE these are made-up numbers from Proof of Concept files!
REMARK:A single project is “an integrated network with closed ends”A subproject is “many strands of insanity” or “many strands that are loosely coupled”
REMARK as you can see an IMS is very different from a single project
REMARK Single, Isolated Project with No Resources: whether you need Critical Path 1.0 or Critical Path 2.0 depends on if you use advanced features in the scheduling application or not (calendars, constraints, elapsed durations/lags, cross-project dependencies and workload leveling).
REMARK: you can see that this (small) program: started 2 days late (buffer went from 8 to 6: lost 2 days of buffer right at the start) and consumes the buffer faster than planned, Extrapolation forecasts the program to be finished at -4: 4 days lateNOTE: these charts are produced by CCP; we can produce one chart for entire program or multiple charts per program: one for each major milestone
REMARK:Functions = “departments”We are NOT using baseline schedules because things change too often and we needed more dynamic way to keep track of the dynamic program! Baselines require too tedious maintenance!
REMARKS: WRT:2. two-versions-of-the-truth problem: aim at One-Version-Of-The-Truth always!!3. Recommend to address the differences in scheduling styles: Common scheduling classes, central definition of major milestones, and pre-defined WBS levels across all functions4. minimize cross-project dependenciesOther lessons learned:The Critical Path can be found across multiple files and in (automatically) workload-leveled schedulesList of “Handoff” deliverables ideally needs to be defined first before scheduling effort is started (Milestones): who needs what from whom?Schedule Management:Focus on the Critical Handoffs (Milestones)Focus on the Program Finish Milestone dateDifference in MS Project planning styles. Recommend: Common scheduling classes, central definition of major milestones, and pre-defined WBS levels across all functions
FPGA = Field Programmable Gate Array
NOTE: as of Oct 2012, 5 and 6 are currently in exposure draft, but very promising documents!