Coursera Operation Project (COP)
Post on 10-May-2015
1480 Views
Preview:
DESCRIPTION
Transcript
COP session 1Web development: reducing idle time of a Web Studio employees
This presentation follows an experiment performed for a web developer studio. The problem statement and initial analysis is described in COP 1-3, COP 4 introduces the ideas to optimize the business processes associated with the inefficiencies and COP 5 introduces an experiment of switching the production method form traditional sequential “Waterfall” to an iterative “Scrum” cycle.
Skip to COP 2Skip to COP 3Skip to COP 4Skip to COP 5
Performed for Coursera Operations Projectby Alexandra Salova
Web studio: Overview
Bitejaw is a web development studio that provides its clients with websites of various types (from e-commerce platforms to enterprise portals). The studio specializes on massive projects involving not only web-design and basic programming but rather creating high-load platforms for large communities if users that process big ammounts of data. It takes usually 6-12 months to produce each website and each project team involves: a project manager, an analyst, a designer, a front-end (UI) programmer, a system engineer, a web-developer and a tester (Quality Assurance specialist – later on QA). Typically each project involves only one project manager and one analyst but may attract multiple designers, QAs, web-developers, system engineers and front-end (UI) programmers.
All the work of the project team begins after the analyst has interviewed the client, confirmed all the deliverables and broken down the project into dozens (even hundreds sometimes) tasks. Then the project manager appoints each task to one of the team members.
NB: Though this is not a typical cyclical synchronized production (as a factory) the Studio has a certain reputation and takes projects that are alike and involve around the same amount of workforce. So it is fair to assume that there may be a standard project production time in this particular case.
COP 1
Web studio : Problem Key dimension: Time. The majority of the processes are consecutive which means that apart from the project manager who overviews the whole project from the beginning to the end all the other team members pick up their line of work after the previous one is done with his. Everything begins with an analyst who prepares the requirements specifications; then a designer creates the mock-ups, a front-end (UI) programmer builds the layout and user interface, a web-developer programs the website, a QA specialists checks it for bugs and inconsistencies and finally an engineer produces the finished product to the web (uploads and administers the website on the server side). Which means that at some point almost every person remains idle since his team-mate hasn’t yet provided him with work. Performance measures: -Activity time: analysis, design, html-layout, programming, testing, product upload; -Website production flow time (the time it takes to finish one project); -Waiting time (for the team members); -Percentage of projects failing the suggests deadlines; Trade-off: - Quality (much of the idle time is caused by the inefficiency of the processes and poor quality).
COP 1
Web studio : Success
Target project flow time: 6 months. Target analysis activity time: 1 month (22 days). Target design activity time: 10 days. Target user interface programming activity time: 10 days. Target programming activity time: 3,5 months (79 days). Target testing activity time: 10 days; Target product upload activity time: 1 day; Reducing the idle time (waiting time) for the team members to become max.20%; All projects meeting the deadlines appointed during initial analysis; All while maintaining reasonable labor costs. NB: 1 month = 22 working days.
COP 1
COP Session 2Web development: reducing idle time of a Web Studio employees Performed for Coursera Operations Projectby Alexandra Salova
Skip to COP 1Skip to COP 3Skip to COP 4Skip to COP 5
Web studio : Defining the problem
Problem definition• How might we increase the speed of producing each project?• Alternative #1 (narrower): How might we reduce analytics and design times?• Alternative #2: How might we reduce quality assurance time?•Alternative #3 (broader): How might we reduce the percentage of idle time our team members spend waiting for their colleagues to finish their work?
COP 2
Web studio: The Customer’s PerspectiveInitial interview with an analyst
20 daysWait for an analyst to formalize the project deliverables.
2 daysReview and approve the project deliverables (otherwise repeat the previous step with necessary adjustments).
Pay an advance.
Wait for the analyst to create multiple tasks corresponding to the project deliverables in a project management tool.
7 daysWait for the project manager to assign the tasks.
1 dayWait for the designer to create the mock-ups.
12 daysReview the mock-ups and approve (otherwise repeat the previous step with adjustments to the designs
Wait for the UI12 days
Wait for the programming80 days
Wait for the quality assurance10 days
Test the preliminary project
Wait for the project upload1 day
Pay the full invoice.
Start using the project.
Flow Time Efficiency (or %VAT) = Total value add time of a unit /Total time a unit is in the process = (20+2+7+1+12+12+80+10+1) days / 270 days = 145 / 270 = 54%
COP 2
Web studio: Process flow diagramInitial interview with an analyst
Wait for an analyst to formalize the project deliverables.Primary bottleneck
Review and approve the project deliverables (otherwise repeat the previous step with necessary adjustments).
Pay an advance.
Wait for the analyst to create multiple tasks corresponding to the project deliverables in a project management tool.
Wait for the project manager to assign the tasks.
Wait for the designer to create the mock-ups.Secondary bottleneck
Review the mock-ups and approve (otherwise repeat the previous step with adjustments to the designs
Wait for the UI
Wait for the programming
Wait for the quality assurance
Test the preliminary project
Wait for the project upload
Pay the full invoice.
Start using the project.
COP 2
Web studio: KPI treeProfit
Revenue
Price
Volume
Demand
Capacity
# of workers
Task processing speed
Idle time
# of projects
Percentage of rework (bugs found during QA)
Cost
Indirect
Rent
SG&A
Insurance
Utilities
Direct
Labor
COP 2
After the experiment
COP Session 3 Web development: reducing idle time of a Web Studio employees
Skip to COP 1Skip to COP 2Skip to COP 4Skip to COP 5
Performed for Coursera Operations Projectby Alexandra Salova
Web studio: Front-line observationsThe overview of the studio’s activity connected to the two bottleneck’s (specified in COP2) – the formalization on project deliverables by the analyst and the preparation of the mock-ups by the designer (including the review process by the client).
The studio actually employs the KANBAN system for some of it’s projects. Which helps “pull” the tasks into work but since the tasks are all very different (in terms of time and difficulty) it doesn’t help a lot to plan the worker occupancy rate.
There’s little to none variability in demand – which is perfect and gives us the opportunity to focus more on what is actually bothering us (idle time of workers, waste).
The primary bottleneck is created by the analyst when he\she is writing the project deliverables according to what the client had said in the interview process.
The secondary bottleneck is the design process. The interesting thing is that in 70% of the cases the delay on this step (leading to most of the worker’s being idle) was made by the client (instead of reviewing the design right away and either approving it or making comment for the re-work the clients take 2-14 days to respond).
Many times the programmer has to rework the tasks he has already done right away (even before the task is moved to quality assurance and testing) – either because the analyst had understood the client wrong (10% of the cases) or because the client has changed his mind about his needs (90% of the cases).
The testing step (Quality Assurance) on average catches 90% of the ‘bugs’ (defects made by either UI or programmers). The other 10% are noticed by the client after the project is put into operation. Those are fixed by the team (under the contract terms) even if they are currently involved with another project.
COP 3
Web studio: Value-Add analysis (programming activity)
Tota
l paid
tim
e
Pro
gra
mm
ing
project deliverables
Design
UI
Quality assurance
RE-WORK
- Time spent waiting for other activities to be completed- Time spent doing re-work (fixing defects and bugs)- Time spent actually working and creating value)
True v
alu
e
ad
d t
ime
COP 3
See after the experiment
Web studio: InflexibilityWaste
Re-work is a huge problem for the studio: Analyst re-works the specifications and deliverables if the client changes his mind about his needs;
Designer re-makes the mock-ups if the decision about the product design is made by more than one person.
Programmer is the champion of re-work since the QA activity usually finds defect for at least two weeks of work.
Activity times for analyst bottleneck vary between 10 days and 1,5 month per project (depends on the both the project difficulty and how well the client understands his\hers needs).
Activity times for designer bottleneck vary between 5 days and one month per project (depends on the qualification of the worker and the decision-making process on the client-side).
Product mix flexibility
All the projects are made to order leading to increased activity time for all the workers.
Workers are not cross-trained mainly because almost each activity in the process involves a highly trained, well educated professional experienced in this exact activity. Sometimes a designer may understand user interface programming or a programmer may understand the principles of quality assurance, but it is not enough to provide quality service.
COP 3
COP Session 4 Web development: reducing idle time of a Web Studio employees
Skip to COP 1Skip to COP 2Skip to COP 3Skip to COP 5
Performed for Coursera Operations Projectby Alexandra Salova
Web studio: Ideas for improvement#1
Outsource the activity processes with the lowest capacity (both bottlenecks – analysts and designers):
We may employ as many workers as demanded by a project;
If we don’t have analysts and designers on our team, we don’t have the problem of them being idle after their part is done;
There’s no guarantee that we find enough qualified workers in time to start working on the signed project (they may be employed elsewhere).
We cannot control them and if they fail their deadlines or produce poor work (analysts especially) the implications for the whole project may be catastrophic.
#2
Schedule the upcoming projects so that the team doesn’t wait for the current project to end in order to begin working on another one:
The idle time of the workers forced to wait for their teammates to finish their line of work is potentially greatly decreased or reduced to zero;
Extremely difficult to execute – we may loose some clients if they need the project urgently and our studio can schedule the client for the next month.
#3
Switch from a sequential approach to a flexible and holistic method where all the work on the project is split up into short periods of time during which all the team-mates are equally occupied:
Flexibility to adjust to ever-changing client needs: the team becomes able to deliver quickly and respond to emerging requirements.
May take a long time to convert the whole work to a significantly different methodology while the projects will suffer due to uncertain control.
COP 4
Web studio: Empirical support#1
Outsource the activity processes with the lowest capacity (both bottlenecks – analysts and designers): Reduce the costs of paying to analysts and designers while they are idle and actually pay for the value-
add time only (reduce costs by ~7%).
Increase the capacity of these lines of work by employing more outsourced workers (increase overall capacity of the production line by 15%).
#2
Schedule the upcoming projects so that the team doesn’t wait for the current project to end in order to begin working on another one: The idle time of the workers forced to wait for their teammates to finish their line of work is
decreased by 65%;
#3
Switch from a sequential approach to a flexible and holistic method where all the work on the project is split up into short periods of time during which all the team-mates are equally occupied: The idle time is decreased by 80%; The re-work of the designer and analyst is decreased by 85%
COP 4
Web studio: Comparison table
COP 4
Ease of implementation
Likely magnitude of impact
Side effects
Outsource bottlenecks Medium – need to lay off employees before outsourcing the target activity.
Low – reduces idle time of target production units but doesn’t solve the quality and inflexibility problem.
Moderate – loosing direct control over crucial process (analytics)
Schedule projects Easy - appoint the next client after the assumed finish of the current project.
Medium – reduces idle time but doesn’t affect the bottlenecks capacity.
Severe – might loose clients not willing to wait.
Switch to a flexible iterative approach
Hard – train all the workers to the new methodology, introduce new procedures.
High – reduces idle time, balances the production line, reduces the rework (increasing the bottleneck capacity).
Moderate – takes a long time to implement.
COP Session 5 Web development: reducing idle time of a Web Studio employees
Skip to COP 1Skip to COP 2Skip to COP 3Skip to COP 4
Performed for Coursera Operations Projectby Alexandra Salova
Web studio: hypothesis for experimentThe premise for the experiment lies in the idea of switching the project production from a traditional sequential approach to a flexible iterative one.
(BEFORE the experiment: sequential production methodology)
This model above that is being used at the moment is flawed in the way that it allows workers to be idle for a lot of time: depending on the line of work the worker utilization may vary in between 30% and 80%. It happens because the project production steps are strictly sequential – first the analyst interviews the client and writes the project deliverables (other workers are idle). Then the tasks are planned and the designing process starts (other workers are idle). And so on. On every step of the production line only one production unit it utilized causing low value-add time for the staff. In waterfall, development teams only have one chance to get each aspect of a project right.
COP 5
Analyst (specificati
ons and deliverable
s)
Task planni
ng
Design
User Interfa
ce
Programming
Quality Assura
nce
Bug Fix (defect re-work)
Deployment
Web studio: hypothesis for experiment The idle time will be decreased by 80%;
The re-work of the designer and analyst will be decreased by 85%;
Due to more flexibility the re-work caused by changing client needs will be decreased by 30%;
The production cycle will be diminished from 6 months on average to 2 weeks which allows to continually revise every aspect of the production throughout every cycle.
(AFTER the experiment: iterative production methodology)
COP 5
Analysis
Design
User Interface
Programming
Quality Assurance
Iterative client
participation
Pro
ject
managem
ent
Pla
nnin
g
Start iteration Deployment
Web studio: the experiment
Subject: To test the methodology switch we took one of the internal projects (the studio sometimes has domestic projects where the client is the studio itself) so that no client projects would suffer from the experiment.
Time: The iteration time was chosen to be 2 weeks – short enough to be manageable and adaptive to change yet long enough to produce a working unit of a project.
Order: The new production cycle now always starts with a ready set of tasks (analyst work) and mock-ups (designer work) so that the team doesn’t wait. The analyst and the designer begin the cycle with working on creating tasks and mock-ups for the next cycle.
Flexibility and feedback: The project manager now presents the client with a working unit of a project every cycle which helps him receive quick feedback. The PM also now gets the client’s thoughts on the project (which admittedly change often enough) more frequently and can make quick changes in the task priority in the upcoming cycle – that brings more flexibility to the production line and saves the team from a lot of massive re-work time.
COP 5
Web studio: experiment resultsIt is hard to tell at this point if all the predicted effects met their planned quantitative characteristics (since the production cycle before the experiment lasted for 6 moths minimum and the experiment is one week only). But preliminary data shows that:
The idle time of all the workers decreased significantly (if not gone completely). Now everyone has their own little cycles of work and while the developers are working on the tasks for the current iteration the analysts and designers are working to prepare the tasks for the next iteration.
The amount of re-work decreased significantly since the communication with the client became more frequent. The possible changes is needs are met quickly and implemented into the process right away - so there's no need to re-do the work of the previous months.
Same goes for the bottlenecks - they are sill there but their capacity increased greatly. Partly because the need in changing the specifications (re-work for both the analyst and designers) is now affecting a much less volume of work (since the needs are met more frequently now), and partly because these bottlenecks now have their separate cycles and don’t interfere much with the other working units.
COP 5
Web studio: OPE after experimentSince the primary concern of this research was the idle time of the workers and the programmers'’ OPE specifically, it would be most justified to look at the finances in terms of the value-add analysis and how it has changed after the experiment. The biggest problem was with the developers' idle time since it is both the highest paying
working unit and the unit that suffered most from the idle time forced by the sequential production cycle.
What we had in terms of programmers' value-add time in shown on this slide.
What we had before the experiment is shown on this slide.What we have now is: Idle time from waiting for the project deliverables (analyst), design (designer) and UI (UI
programming) is reduced practically to zero because these steps are moved to the previous production cycle (so that when the programmer starts the current cycle all of these steps are done during the previous cycle and the analyst, the designer and the UI developer work on the tasks for the next cycle during the current one). Saves 22% of idle time.
Idle time from QA is reduced by 80% since QA is done almost at the same time as the programming (programmers finish task - move it to QA - QA creates a bug fix ticket for the project managers to consider if it is worth re-working on).
Re-work reduced by 15% (presumably more - but that must be observed over several months in order to get exact comparison).
COP 5
Web studio: Value-Add analysis after experiment (example: programming activity)
Tota
l paid
tim
e
Pro
gra
mm
ing
project deliverables
Design UI Quality assurance
RE-WORK
- Time spent waiting for other activities to be completed- Time spent doing re-work (fixing defects and bugs)- Time spent actually working and creating value)- Increase in value-add time
True v
alu
e a
dd
tim
e
COP 3
Idle time waiting for resources decreased almost to 0.
Idle time from QA is reduced by 80%
Re-work reduced by 15%
Web studio: FinanceCOP 5
If we have a look at our KPI tree from COP2 we may see that by doing our experiment and reducing the percentage of both re-work and idle time we increased the capacity of our production line. All the other variables affecting the profit remained the same (price demand and costs) which gives us the confidence to say that the profits have increased. It is difficult to provide the exact numbers of the profit change since the change would be seen by accounting only after 6 months of the experiment. But the relation between the profit and out experiment is shown on the next slide.
Web studio: KPI tree
Profit
Revenue
Price
Volume
Demand
Capacity
# of workers
Task processing
speed
Idle time
# of projects
Percentage of rework (bugs found during
QA)
Cost
Indirect
Rent
SG&A
Insurance
Utilities
Direct Labor
COP 2
top related