Chapter 16 of 32

Software Project Management

So far we have seen how software can be built (the process models, from Waterfall Model to Scrum) and what should be built (Requirements Engineering, Software Requirements Specification (SRS), Feasibility Study). But a good process and a clear SRS do not deliver a product on their own. Someone has to decide how big the job is, who does what, in what order, how much it will cost, and what to do when things go wrong.

That "someone" is software project management. Studies of failed projects keep finding the same culprits: bad estimates, unrealistic deadlines, poor communication and ignored risks. Very few projects fail because the team did not know how to code. This chapter gives the big picture of project management. The next four chapters zoom in on its main tools: size estimation, COCOMO, scheduling and risk management.

Software project management is the planning, organising, staffing, monitoring and controlling of a software project so that it delivers the required product on time, within budget and with acceptable quality.


What Is a Project?

A project is a temporary effort, with a definite start and end, undertaken to create a unique product, service or result.

Two words matter here. Temporary: a project ends when its goal is met (or abandoned), unlike routine operations such as running a server farm. Unique: no two software projects are exactly the same, so past experience helps but never gives the exact answer. That uniqueness is why estimation is hard.

Why Software Projects Are Harder to Manage

Building a bridge and building a banking app are both projects, but software has some special difficulties (recall Software Characteristics & Software Crisis):

Property of softwareWhat it means for the manager
InvisibilityYou cannot walk on site and see that the "first floor" is 60% done. Progress must be measured through documents, demos and metrics.
ChangeabilityBecause software looks easy to change, customers keep asking for changes, and scope keeps growing ("scope creep").
ComplexityA medium-sized system has thousands of interacting parts; small errors anywhere can stop the whole thing.
UniquenessEach project uses a new mix of requirements, people and technology, so historical data only partly applies.
Human-intensiveThe main cost is people's time. Productivity between developers can differ several-fold, so staffing decisions matter enormously.

The Triple Constraint

Every project is squeezed between three limits:

                 SCOPE
                (features)
                  /\
                 /  \
                /    \
               / QUAL-\
              /  ITY   \
             /__________\
          TIME          COST
       (schedule)     (budget, people)

Change one side and at least one other side must move, or quality suffers. If the customer adds features (scope up), then either the deadline moves (time up), more money is spent (cost up), or corners get cut (quality down). A large part of the manager's job is making these trade-offs visible and deliberate instead of letting them happen silently.

Agile methods handle the triangle differently: they usually fix time and cost (a fixed-length Sprint with a stable team) and let scope flex, delivering the most valuable items first (see Agile Methodology).


The Four Ps of Project Management

Pressman organises project management around four Ps. They appear often in university exams.

PWhat it coversTypical questions
PeopleStakeholders, team leaders, developers, customers, end usersWho is on the team? Who decides? How do they communicate?
ProductThe software to be built: its scope, objectives and decompositionWhat exactly must it do? What is out of scope?
ProcessThe framework of activities used to build it (Waterfall, Spiral, Scrum, …)Which model fits this project's size, risk and requirement stability?
ProjectThe planned and controlled work: estimates, schedule, tracking, riskHow long? How much? How do we know if we are on track?

Pressman ranks people first on purpose: a motivated, skilled team can survive a mediocre process, but the best process cannot rescue a demoralised team.

The W⁵HH Principle

Barry Boehm suggested a simple checklist that any project plan should answer. It is called W⁵HH (five Ws, two Hs):

  1. Why is the system being developed? (business reason)

  2. What will be done? (tasks and functionality)

  3. When will it be done? (schedule and milestones)

  4. Who is responsible for each function?

  5. Where are they located in the organisation?

  6. How will the job be done technically and managerially?

  7. How much of each resource (people, money, hardware) is needed?


Activities of a Software Project Manager

Project management activities fall into two broad groups: planning (done mainly at the start, then refined) and monitoring and control (done continuously until the end).

   Feasibility study done
            │
            ▼
   ┌───────────────────── PROJECT PLANNING ─────────────────────┐
   │  1. Estimate size        (LOC, Function Points)            │
   │  2. Estimate effort/cost (COCOMO, expert judgement)        │
   │  3. Estimate duration    (COCOMO, scheduling)              │
   │  4. Schedule tasks       (WBS, Gantt, PERT/CPM)            │
   │  5. Plan staffing        (who, how many, when)             │
   │  6. Plan for risks       (identify, analyse, mitigate)     │
   │  7. Other plans          (quality, configuration, ...)     │
   └────────────────────────────┬───────────────────────────────┘
                                ▼
   ┌──────────────── MONITORING & CONTROL (loop) ──────────────┐
   │  measure progress → compare with plan → take corrective   │
   │  action → update plan → repeat                             │
   └────────────────────────────────────────────────────────────┘

Notice the order of the estimates: size first, then effort and cost, then duration. This order is important and is a favourite exam question. Effort depends on how much there is to build, and duration depends on how much effort is needed (and how many people can usefully share it).

Why size comes first

Effort and time cannot be estimated directly with any accuracy; "how many months will the hospital system take?" has no obvious answer. But "how big is the hospital system?" can be broken down module by module, and then converted into effort using models built from historical data. Size is the foundation everything else rests on, which is why Software Size Estimation: LOC & Function Points comes next.

Sliding-window (rolling-wave) planning

At the start of a project, you know very little, so a detailed plan for month 12 is mostly fiction. Good managers plan in stages: a detailed plan for the near term, a rough plan for later phases, refined as each phase completes and more is known. This is called sliding-window or rolling-wave planning. It is the same idea behind Sprint Planning in Scrum.


The Software Project Management Plan (SPMP)

The results of planning are written down in a Software Project Management Plan. The classic outline was IEEE 1058; today its successor is the international standard ISO/IEC/IEEE 16326 (project management for systems and software). A typical SPMP contains:

  1. Introduction: objectives, major functions, performance issues, management and technical constraints.

  2. Project estimates: historical data used, estimation techniques, effort, cost and duration.

  3. Schedule: work breakdown structure, task network, Gantt chart, milestones.

  4. Project resources: people, hardware, software and special resources.

  5. Staff organisation: team structure and reporting.

  6. Risk management plan: risk analysis, identification, estimation, abatement.

  7. Project tracking and control plan: how progress is measured and reported.

  8. Miscellaneous plans: quality assurance, configuration management, validation and verification, installation, maintenance.

In Agile projects these items still exist, but they are lighter and spread across artefacts such as the Product Backlog, Sprint Backlog, release plan and Definition of Done.


Team Structures

How a team is organised affects communication, speed and quality. Three classic structures are taught:

StructureHow it worksGood forWeakness
Chief programmer team (controlled centralised)A highly skilled chief programmer designs and makes key decisions; others (backup programmer, librarian, programmers) support. Communication flows through the chief.Small, well-understood, routine problems; tight deadlinesChief becomes a bottleneck and single point of failure; lower morale for others
Democratic team (egoless, democratic decentralised)No permanent leader; decisions by consensus; members review each other's code ("egoless programming", Weinberg).Hard, research-like problems needing many ideas; small teamsSlow decisions; does not scale; needs mature members
Mixed (controlled decentralised)A senior leader coordinates sub-teams; each sub-team works democratically internally.Large projectsMore layers of communication

Modern Agile teams are closest to the democratic model: Scrum (per the 2020 Scrum Guide) describes a small, self-managing team of typically 10 or fewer people, with three accountabilities: Product Owner, Scrum Master and Developers. Note that Scrum has no "project manager" role; the managerial work is shared among these three.

Why team size matters: communication paths

In a team of n people where everyone may talk to everyone, the number of communication paths is:

Paths = n(n − 1) / 2

n =  3  →   3 paths
n =  5  →  10 paths
n = 10  →  45 paths
n = 20  → 190 paths

Doubling the team from 10 to 20 more than quadruples the paths. Time spent coordinating grows much faster than the number of hands doing the work.

Brooks' Law

This leads to one of the most quoted results in software engineering, from Fred Brooks's book The Mythical Man-Month (1975):

"Adding manpower to a late software project makes it later."

Why? New people need training (taking time from experienced members), the work has to be re-divided, and communication paths increase. Effort and months are not interchangeable: if a task takes one person 12 months, 12 people will not finish it in one month. Brooks's famous analogy: bearing a child takes nine months no matter how many women are assigned.


Project Monitoring and Control

A plan is only useful if you compare reality against it. Common tracking techniques:

  • Milestones: points where a measurable result is checked (e.g., "SRS signed off", "beta released"). A deliverable is a milestone result handed to the customer.

  • Status meetings and reports: in Scrum, the Daily Scrum and Sprint Review play this role.

  • Reviews: technical reviews of design and code catch problems that progress charts hide.

  • Charts: Gantt charts (see Project Scheduling: Gantt Charts, PERT & CPM) and, in Agile, burn-down / burn-up charts.

  • Earned Value Analysis: a numeric way to answer "are we ahead or behind, over or under budget?"

Earned Value Analysis (EVA)

The weakness of plain "money spent so far" is that spending less than planned could mean efficiency or simply that the work has not been done. Earned value fixes this by measuring the value of work actually completed.

QuantityMeaning
BAC (Budget at Completion)Total planned budget
PV (Planned Value)Budgeted cost of work scheduled up to today
EV (Earned Value)Budgeted cost of work actually completed up to today
AC (Actual Cost)Money actually spent up to today
Schedule Variance  SV  = EV − PV      (negative → behind schedule)
Cost Variance      CV  = EV − AC      (negative → over budget)
Schedule Perf. Idx SPI = EV / PV      (< 1 → behind schedule)
Cost Perf. Index   CPI = EV / AC      (< 1 → over budget)
Estimate at Completion EAC = BAC / CPI  (if current cost efficiency continues)

Worked Example

A project has a budget of ₹10,00,000. By the end of month 4, the plan said work worth ₹4,00,000 should be done. Only work worth ₹3,50,000 (by its budgeted cost) is actually complete, and ₹4,20,000 has been spent. Analyse.

BAC = 10,00,000   PV = 4,00,000   EV = 3,50,000   AC = 4,20,000

SV  = 3,50,000 − 4,00,000 = −50,000      → behind schedule
CV  = 3,50,000 − 4,20,000 = −70,000      → over budget
SPI = 3,50,000 / 4,00,000 = 0.875        → only 87.5% of planned progress
CPI = 3,50,000 / 4,20,000 ≈ 0.833        → ₹0.83 of work per ₹1 spent
EAC = 10,00,000 / 0.8333  = 12,00,000    → expected final cost

The project is both behind schedule and over budget, and if nothing changes it will cost about ₹12 lakh instead of ₹10 lakh. Spotting this at month 4 gives the manager time to act: cut scope, re-plan, or fix the cause of low productivity.


Measurement and Metrics

You cannot control what you do not measure. Project management uses three families of metrics:

TypeMeasuresExamples
Process metricsEffectiveness of the development process across projectsDefect removal efficiency, defects found per phase
Project metricsState of the current projectEffort spent, schedule slippage, earned value, velocity
Product metricsProperties of the software itselfSize (LOC, FP), complexity, defect density

Size metrics are covered in the next chapter. Product quality metrics come later in Software Quality & Metrics.


Common Mistakes

  • Estimating effort before size. The correct order is size → effort/cost → duration → schedule.

  • Treating person-months as interchangeable with calendar months. 24 person-months does not mean 24 people can finish in one month (Brooks' Law).

  • Reading CV/SV signs the wrong way. Negative variance and an index below 1 are bad news. Remember: EV comes first in every formula.

  • Confusing PV with AC. PV is what you planned to have done; AC is what you spent.

  • Planning once and never re-planning. A plan is a living document; use rolling-wave refinement.

  • Saying Scrum has a project manager. The Scrum Guide 2020 defines only Product Owner, Scrum Master and Developers.


Interview and Exam Questions

Q1. What are the four Ps of software project management? Which is most important and why?

People, Product, Process, Project. Pressman puts people first because software is built by humans; skilled, motivated and well-organised people make the other three work.

Q2. In what order are the main project estimates made?

Size first (LOC or function points), then effort and cost, then duration, and finally the detailed schedule and staffing. Each estimate depends on the previous one.

Q3. A team grows from 6 to 12 members. How many communication paths are there before and after?

6 × 5 / 2 = 15 before, 12 × 11 / 2 = 66 after. Doubling the team more than quadruples the paths.

Q4. State Brooks' Law and explain why it holds.

"Adding manpower to a late software project makes it later." New members need training and ramp-up time (taken from experienced members), tasks must be re-partitioned, and communication overhead grows roughly as n². In the short term productivity falls, not rises.

Q5. BAC = ₹50 lakh, PV = ₹20 lakh, EV = ₹22 lakh, AC = ₹25 lakh. Is the project ahead or behind schedule? Over or under budget?

SV = 22 − 20 = +2 lakh, SPI = 1.1 → ahead of schedule. CV = 22 − 25 = −3 lakh, CPI = 0.88 → over budget. EAC = 50 / 0.88 ≈ ₹56.8 lakh.

Q6. Compare the chief programmer team and the democratic team.

The chief programmer team centralises decisions in one expert: fast and consistent, good for routine problems, but the chief is a bottleneck and single point of failure. The democratic team decides by consensus with egoless reviews: good for hard, novel problems and morale, but slow and hard to scale.

Q7. What is sliding-window planning?

Planning in stages: making a detailed plan only for the near future and a coarse plan for later phases, and refining the plan as each stage finishes and more information becomes available. It avoids committing to detailed plans that early uncertainty makes unreliable.


Summary

  • Software project management plans, organises, staffs, monitors and controls a project so it delivers on time, within budget, with acceptable quality.

  • Software is invisible, changeable, complex and unique, which makes it harder to manage than most engineering projects.

  • The triple constraint (scope, time, cost, with quality in the middle) means every change forces a trade-off.

  • Pressman's four Ps: People, Product, Process, Project. Boehm's W⁵HH lists the questions every plan must answer.

  • Planning order: size → effort/cost → duration → schedule → staffing → risk; results go into the SPMP (ISO/IEC/IEEE 16326).

  • Team structures: chief programmer, democratic, mixed. Communication paths = n(n − 1)/2, which explains Brooks' Law.

  • Earned value: SV = EV − PV, CV = EV − AC, SPI = EV/PV, CPI = EV/AC, EAC = BAC/CPI.

  • Next, Software Size Estimation: LOC & Function Points shows how to measure the size of software before it exists.