46
Scrum in the Waterfall Scrum Gathering 2009, Orlando, FL Dave Prior, PMP, CSP, MBA

D Prior Scrum In The Waterfall

Embed Size (px)

Citation preview

Page 1: D Prior Scrum In The Waterfall

Scrum in the Waterfall Scrum Gathering 2009, Orlando, FL

Dave Prior, PMP, CSP, MBA

Page 2: D Prior Scrum In The Waterfall

Once upon a time in the far off land of Oklahoma...

Page 3: D Prior Scrum In The Waterfall

There was a really big little company that had some very

old software...

Page 4: D Prior Scrum In The Waterfall

That took two years in requirements, and four years

to build... ten years ago

Page 5: D Prior Scrum In The Waterfall

that they needed rebuilt in eight months, using .Net...

Page 6: D Prior Scrum In The Waterfall

and they had no documented requirements...

Page 7: D Prior Scrum In The Waterfall

and they wanted to switch to Scrum... at least, some of

them did.

Page 8: D Prior Scrum In The Waterfall

“You can use ‘scum’ or whatever you want. Just give me a Gantt chart, tell me what day you will be done all the

work and how much it will cost.”

Page 9: D Prior Scrum In The Waterfall

Here’s the rub...  Dir. of IT/Product Owner wants a fresh start,

embracing change and believes deeply in the benefits of Scrum�

 Senior Mgmt./Project Sponsor wants results, but is not interested in changing how the company looks at work�

 We needed to find a way to keep both sides happy so we could get the work done.

Page 10: D Prior Scrum In The Waterfall

the not helping part... Sr. Sponsor: When will it be done?�

Product Owner: When you stop asking for more stuff.�

Sr. Sponsor: What will you have ready for me on Date X?

Product Owner - Whatever we’ve got completed at that time

Page 11: D Prior Scrum In The Waterfall

How we solved it...

  Both the Tech Lead and PM assigned were CSMs

  1 CSM to lead development

  1 CSM to manage communication and relationship with senior management

  Tailoring of the output of the Scrum Process for an audience looking for more traditional PM reporting

Page 12: D Prior Scrum In The Waterfall

Project Leadership Project Leadership   Product Owner - storehouse of knowledge of what the

application is supposed to do and how it was done before�

  Scrum Master - traditional CSM role...leads scrum efforts, manages technical team’s efforts�

  PM/Scrum Master - acts as the “anti-corruption layer” between senior sponsor and product owner, translates sprint reporting into something palatable by senior management

Page 13: D Prior Scrum In The Waterfall

Why do they want what they want?

 There is the “thing” being demanded�

 There is the reason it is being demanded

  If you can’t meet the demand, meet the motivation�

 Uncovering why they need what they are asking for can create options for meeting their true need instead of their demand

Page 14: D Prior Scrum In The Waterfall

Want - Need   They want a Gantt chart with milestones, cost estimates,

staffing requirements�

  They need to be able to demonstrate to the organization that the project is on track and moving in a positive manner towards completion�

  A lack of visibility can appear as chaos to those who are already worried. This drives fear, which drives stress�

  Stress is way bad�

  Scrum offered a level of transparency that was ideally suited to this situation... except for one thing

Page 15: D Prior Scrum In The Waterfall

The one thing... The one thing... The one thing...   Senior Management did not have the bandwidth to

attend the daily stand-ups or visit the war room or get involved with the project at a participant level�

  Senior Management just needed information that they could trust and they needed it provided with minimal effort required of them

Page 16: D Prior Scrum In The Waterfall

Care and feeding... Care and feeding...   While Senior Management was asking for traditional

reporting, what they really wanted was a clear picture, week to week, of how the work was progressing and some reassurance that we could meet the deadline�

 Option 1: Develop a best guess Gantt Chart that would be based only on the initial estimates provided by the Product Owner and hope we’d be able to keep pace with it.�

 Option 2: Use the output of the Scrum process to build a story, sprint by sprint, that earned trust repeatedly by showing concrete examples of output and progress

Page 17: D Prior Scrum In The Waterfall

Getting Started with Option 2 Getting Started with Option 2  We began the project with a backlog of all

functionality to be provided in the deliverable. �

  The product owner provided the team with complexity estimates on each item to be delivered and had converted those into time estimates as well. �

  The product owner assumed no reusability for the components to be produced, working under the assumption that this would provide some padding in the schedule.

Page 18: D Prior Scrum In The Waterfall

Working Time Working Time Working Time

While there were some staffing fluctuations across the life of the project, we began with the assumption that we could expect to see an average of six hours of productivity, per day for each resource working on the project. We also assumed that each of them would work five days per week.

Page 19: D Prior Scrum In The Waterfall

Planning the Sprint   During Sprint planning meetings, we worked with

our ideal hour of labor for that sprint and added items, based on prioritization. Each item would be estimated for both complexity and labor required to complete.

  We added only enough work to fill the number of available hours per sprint

  Because work on the use cases was just beginning at the start of the project, the team had to rely on the product owner to understand the requirements for each deliverable item

Page 20: D Prior Scrum In The Waterfall

Watching the Variance   At the start of each sprint we had a set list of tasks

for the team to work on.

  We had complexity estimates and labor estimates for each task provided by both the Product Owner and the Development Team.

  We began tracking the variance, sprint to sprint between available hours and actuals and between planned hours and actuals

  We also began tracking the variance between the original estimates and team estimates for both complexity and labor required to complete.

Page 21: D Prior Scrum In The Waterfall

The slightly educated guess

  By tracking the variance in Fibonacci estimation across sprints, we determined that the original estimates for complexity were, on average, 155% of what the team estimated for the same work.

  By tracking the variance in labor estimation across sprints, we determined that the original estimates for labor were 125% of what the team estimated for the same work.

Page 22: D Prior Scrum In The Waterfall

The slightly educated guess

In order to get closer to an understanding of when the project was likely to actually end, based on what we knew about the deliverables we had to provide, we applied the 3 Point Pert Estimation formula:

Optimistic + Most Likely + Pessimistic� 3�

Page 23: D Prior Scrum In The Waterfall

3 Point Pert Application While not an entirely proper use of the formula, it provided a more likely estimate than just using one of the three:

Pessimistic: Original Labor Estimate

Most Likely: Original Labor Estimate/Labor Estimation Variance

Optimistic: Original Labor Estimate/Fibonacci Estimation Variance

Page 24: D Prior Scrum In The Waterfall

3 Point Pert Application

Example:

Pessimistic=Orig. labor estimate = 1,000 hours

Most Likely = 1,000 / 125% = 800

Optimistic = 1,000 / 155% = 645

(1,000 + 800 + 645) / 3 = 815 hours

Page 25: D Prior Scrum In The Waterfall

Working with the educated estimate

With an established labor capacity (# of team members * 6 hours/day * 5 days/week) and a revised estimate on how many hours of labor we expect to actually have to complete, we can determine how long it will take to get through the backlog.

4 team members * 6 hours/day * 5 hours/week = ideal work capacity of 120 hours per week

Page 26: D Prior Scrum In The Waterfall

Applying capacity to the Pert Estimate

 With an ideal capacity of 120 hours of labor per week

 And an estimated 815 hours of labor to burn down

 The project should take 6.8 weeks to complete

Page 27: D Prior Scrum In The Waterfall

Tracking Performance

  By tracking the team’s actuals against their planned work, we were able to determine, on average, how effective they were at reaching their goal each week.

  If we know the team is planning for 120 hours of work per week, but, on average, they only complete 90% of that work, we can make adjustments to how we report on the burndown.

Page 28: D Prior Scrum In The Waterfall

Tracking Performance

  120 hours/wk * 90% = 108 actual performed hours realized per week

  If we are three weeks into our 815 hour project, we will have realized 324 hours of work instead of our expected 360 hours of work

  This leaves us with 491 hours of work remaining after week three

Page 29: D Prior Scrum In The Waterfall

Tracking Performance

 Given that we are realizing only 324 hours/week, we have to adjust expectations with the client because the remaining 491 hours will now take longer than originally expected

 491/108 = 4.5 weeks will be required to complete the remaining work as opposed to the 4 weeks it would take if we were performing at 120 hours as planned

Page 30: D Prior Scrum In The Waterfall

Tracking Performance

  At the end of each sprint, adjustments were made to all of the previous calculations based on new performance data, as well as any new items added to the backlog.

  These adjustments to the estimates keep the reporting as near to real time as possible and it goes a long way with respect to demonstrating the value provided of this type of reporting over the originally request Gantt chart.

Page 31: D Prior Scrum In The Waterfall

Keeping It Real Keeping It Real   You need to re-evaluate the variances and update the

forecasting based on revised averages which include the new performance data at the end of each sprint and report on this, showing how it trends as the work progresses�

  This not only helps you re-evaluate the total work remaining each Sprint it build a story for Sr. Management that can show positive or negative trends�

  The story that is created through this type of reporting is what “sells” the value to Sr. Management

Page 32: D Prior Scrum In The Waterfall

Wrapping it up for the folks upstairs... upstairs...

No matter how much information you have on work performed, or work remaining, it is only as good as your ability to communicate it to the parties who need to be able to digest it

Page 33: D Prior Scrum In The Waterfall

One Report to Bind It All

The weekly report to management  Summary Page of Issues/Accomplishments

 Summary Table of Performance Data

 Trend and Completion Information

 Breakdown of recorded man hours worked and cost of those hours

(if these can be tracked against an original set budget, even better)

Page 34: D Prior Scrum In The Waterfall

Example of a the Summary Table of Performance Data

Project Development Summary

Week Ending 7/25/08

Estimated Total Man Hours of Work Available 3118 Total of Actual Man Hours Worked To Date 1018 Percent of Total Man Hours Worked To Date 33%

Hours of Work Completed in Sprint Ending (7/25/08) 87 Hours of Work Planned for Sprint Ending (7/25/08) 94 Percent of Work Completed to Work Planned 93% Percent of Work Completed to Work Planned Last Sprint 97% Velocity Change Since Last Week -4%

Estimated Hours of Available Work Remaining 1620

Hours of Development Work Not Yet Done Done 2009

Average Estimation Accuracy (Original vs. Developer) for Open Development Work 145% Adjusted Number of Development Hours Remaining 2922 Adjusted Remaining Development Work - Remaining Available Work 1302

Weeks off based on Variance (with 150 wk Man Hours) 8.68

Page 35: D Prior Scrum In The Waterfall

Additional Data Points

We also track the following on a Sprint to Sprint basis:�

 Targeted (ideal) hours vs. Actual Hours Achieved

 Hours Achieved vs. Hours Planned�

We report on these each week to show trends here as well

Page 36: D Prior Scrum In The Waterfall

Trend Charts

Page 37: D Prior Scrum In The Waterfall

What’s Missing

 The Summary and Trend Analysis is a great starting place and provides excellent detail that you can use to measure progress, similar to earned value�

 But there is an easier way...

Page 38: D Prior Scrum In The Waterfall

The Most Important Point �In This Presentation

Page 39: D Prior Scrum In The Waterfall

Management Likes Pie!

Page 40: D Prior Scrum In The Waterfall

The Detailed Pie We track each item in the product backlog by its status

Page 41: D Prior Scrum In The Waterfall

A Simpler Pie

Break out remaining work by status and total hours

Page 42: D Prior Scrum In The Waterfall

Management’s Favorite Pie

Page 43: D Prior Scrum In The Waterfall

Management’s Favorite Pie

Break out all work by • Done • Done Done • Not Done

Page 44: D Prior Scrum In The Waterfall

Summary Summary   You can use Scrum in a Waterfall Environment, even if the

environment will not be converted

  It may work best with a pair of scrum masters, one for the team and one for management

  Understanding the needs of management, not just what they demand is the key to your success�

  You need to tailor the information you provide them with to meet their needs. If Scrum does not give them what they want, take what Scrum give you and translate it into something they can digest

  Management likes Pie

Page 45: D Prior Scrum In The Waterfall

Questions?

Page 46: D Prior Scrum In The Waterfall

Thank You!

Dave Prior [email protected]