Project Management Leadership with AI · Execution, control, risk and quality · lesson 9 of 21 · 14 min

Monitoring and controlling performance

What to measure

Monitoring compares actual performance with the plan and forecasts where the project will end. Measure what helps decisions, not everything that can be counted.

| Dimension | Predictive measures | Agile measures | |---|---|---| | Schedule | Milestones hit, critical path slippage, SPI | Sprint goal achievement, release burnup, forecast range | | Cost | Actual vs budget, CPI, EAC | Burn rate vs value delivered, cost per increment | | Scope | Change requests, requirements delivered | Backlog changes, scope growth on burnup | | Quality | Defects, test pass rates, rework | Escaped defects, Definition of Done adherence | | Risk | Risk exposure trend | Same | | Benefits | Leading indicators of adoption | User metrics, feature usage | | Team | Morale, turnover | Same, plus retrospective actions completed |

Earned value in brief

Earned value management compares planned value (PV), earned value (EV, the budgeted value of work actually done) and actual cost (AC):

CPI = EV / AC     (cost efficiency; < 1 means over cost)
SPI = EV / PV     (schedule efficiency; < 1 means behind)
EAC ≈ BAC / CPI   (if current cost efficiency continues)

Illustrative. BAC = 500k. At the midpoint, PV = 250k, EV = 200k, AC = 240k. CPI = 0.83, SPI = 0.80. EAC ≈ 500/0.83 ≈ 600k. The project is over cost and behind; a forecast of about 600k is more honest than restating the 500k budget. (The course Project Controls in the AI Era covers EVM in depth.)

Burnup charts for agile

A burnup chart plots completed work against total scope over time. It shows progress and scope growth separately, which a burndown chart can hide. If the "completed" line will not reach the "scope" line by the target date, the team must reprioritise, add capacity carefully, or move the date.

Scope (points):     200 ─── 220 ─── 240 ─── 240
Completed:           40 ───  80 ─── 115 ─── 150
Sprint:               2       4       6       8
Average velocity ≈ 19 points/sprint → remaining 90 ≈ 5 more sprints

Status reporting with meaning

Define status colours with rules, not feelings. For example: Green = forecast within tolerances; Amber = forecast at risk of breaching, recovery plan in place; Red = forecast breach or no credible plan. Always include what decisions are needed.

Variance analysis

For each significant variance, capture: what happened, root cause, impact on end date, cost and benefits, action and owner. Avoid symptom-level explanations such as "development is late"; aim for "integration environment unavailable for 9 days due to licensing delay; recovered 4 days via parallel testing; remaining impact 5 days on release 2".

Forecasting, not just reporting

The most valuable output of monitoring is a credible forecast: when will we finish, at what cost, with what scope, and how confident are we? Use trends, not single data points, and state assumptions.

Worked example

Illustrative. A fictional public health agency in the UK ran a digital appointment system project. Reports were green for six months. A new PM introduced a burnup chart: it showed scope had grown 40% while velocity was flat. At the current rate, completion would take five months longer than planned. The board, now seeing the real picture, agreed to descope two features to a later phase and add one experienced developer. The new forecast was met.

Common mistakes

• Measuring activity (hours logged) rather than progress (value delivered).
• Status colours without definitions.
• Reports that describe the past but never forecast.
• Burndown charts that hide scope changes.
• Ignoring team health indicators until people leave.

Quick self-check

Can you state your project's forecast end date and cost with a confidence level, and the two assumptions that most affect them? If not, your monitoring is reporting, not controlling.

Choosing a small set of indicators

Resist the temptation to track dozens of metrics. For most projects, five to eight indicators are enough: forecast finish vs baseline, forecast cost vs budget, milestone or release confidence, scope change trend, quality (escaped defects or test pass rate), top risk exposure, benefits leading indicators and team health. Review them at a regular cadence and change the set only when a metric stops informing decisions.

Leading and lagging indicators

Lagging indicators (cost overrun, missed milestone) tell you what already happened. Leading indicators (aged issues, WIP growth, float erosion, falling test pass rates) warn you earlier. Good dashboards include both.

Hands-on: burnup forecast in Excel

Columns: A Week | B Completed (cumulative) | C Total scope | D Completed this week =B3-B2
Recent throughput (last 6 weeks)   =AVERAGE(OFFSET(D1,COUNT(D:D)-5,0,6,1))      (or reference the last six cells directly)
Slowest / fastest recent week      =MIN(...) / =MAX(...) over the same six cells
Remaining scope                    =INDEX(C:C,COUNTA(C:C))-INDEX(B:B,COUNTA(B:B))
Forecast weeks (likely)            =ROUNDUP(Remaining/Recent_throughput,0)
Forecast weeks (range)             =ROUNDUP(Remaining/Fastest,0) to =ROUNDUP(Remaining/Slowest,0)
Scope growth per week (recent)     =(INDEX(C:C,COUNTA(C:C))-INDEX(C:C,COUNTA(C:C)-6))/6

If scope is still growing, forecast with net throughput (throughput minus scope growth) as well; if net throughput is near zero, the project will not finish without a scope decision.

Hands-on: a throughput Monte Carlo forecast (Python)

import numpy as np

weekly_done = [14, 20, 17, 22, 15, 19, 18, 21]      # items or points, recent weeks (illustrative)
remaining, scope_growth = 270, 3                     # growth per week if requests keep arriving
rng = np.random.default_rng(9)
weeks = []
for _ in range(20_000):
    left, w = remaining, 0
    while left > 0 and w < 200:
        left -= rng.choice(weekly_done) - scope_growth
        w += 1
    weeks.append(w)
print(f"50%: {np.percentile(weeks, 50):.0f} weeks | 85%: {np.percentile(weeks, 85):.0f} weeks")

Sampling from real recent weeks avoids assuming a distribution; state the scope-growth assumption in the forecast.

Prompt template: a forecast-first status report (approved AI tool)

From the metrics table below, write a status report: (1) headline, 2 sentences; (2) forecast finish and cost with range
and the assumption behind it; (3) top 3 variances as cause – impact – action – owner; (4) decisions needed, from whom,
by when. Apply these RAG rules: <paste rules>. Use only the data given and cite the metric for each number.

How to measure success

• Every status report contains a forecast with a range and its assumption.
• RAG status follows written rules; no "green until suddenly red".
• Scope growth tracked and decided explicitly, not absorbed silently.

Video lecture: Monitoring and controlling performance

Lecture coming soon · 10 chapters · about 9 minutes. Read the full transcript below.

1. Green for six months, then red
2. Why it matters
3. The concept: a small set of indicators
4. EVM in brief
5. Worked example one: a midpoint check
6. Worked example two: the UK appointment system
7. Watch me do it: a burnup forecast
8. Forecasting, not just reporting
9. Status reporting and variance analysis
10. Recap and try this now

Lecture transcript

Green for six months, then red

A public health agency in the UK, fictional but typical, was building a digital appointment system. For six months, every status report was green. Then a new project manager drew a simple chart. It showed that scope had grown forty per cent while the team's delivery rate hadn't changed. At that rate, the project would finish five months late. The data had been there all along. Nobody had looked at it the right way. In this lecture you'll learn what to measure, how earned value works in brief, how burnup charts reveal what status colours hide, how to write status reports that forecast rather than just describe, and how to analyse variances. By the end, you'll be able to build a burnup chart and write a forecast statement a board can act on.

Why it matters

Why does this matter? Because reporting the past isn't the same as controlling the future. A status report that says what happened this month, without forecasting where the project will end up, gives leaders nothing to decide on. Status colours without rules are worse: people report green until it's impossible, and the board learns about problems when options have run out. The whole point of monitoring is to create options early: descope, add capacity, re-sequence or reset expectations while those choices are still cheap. And that requires forecasting, based on data, with the uncertainty stated.

The concept: a small set of indicators

Here's the key idea: choose a small set of indicators, some lagging and some leading. Lagging indicators tell you what has happened: progress against plan, cost efficiency and the resulting forecast. Leading indicators warn you about what's coming: the trend in top risks, work in progress piling up, and the age of open issues. For predictive work, earned value gives you progress and cost efficiency. For agile work, a burnup chart does the same job visually. Think of it like a car dashboard: a speedometer and fuel gauge tell you where you are, but the warning lights tell you what's about to go wrong. A dashboard with fifty indicators is like a car with fifty warning lights. Nobody reads it.

EVM in brief

Earned value, briefly, because the project controls course covers it in depth. Planned value is the budgeted value of work scheduled to be done by now. Earned value is the budgeted value of the work actually done. Actual cost is what that work cost. The cost performance index is earned value divided by actual cost; below one means you're over cost for the work done. The schedule performance index is earned value divided by planned value; below one means less work is done than planned. And a simple forecast: the estimate at completion is roughly the budget divided by CPI, if current efficiency continues. A memory aid: earned value always comes first in the formula. Divide it by what you're comparing it with.

Worked example one: a midpoint check

Let's do the example from the lesson. Budget at completion: five hundred thousand. At the midpoint, planned value is two hundred and fifty thousand, earned value two hundred thousand, and actual cost two hundred and forty thousand. CPI: two hundred over two forty, about nought point eight three. SPI: two hundred over two fifty, nought point eight. So we're over cost and behind. The estimate at completion, if efficiency continues, is five hundred divided by nought point eight three: roughly six hundred thousand. A status report that says 'we're halfway through the timeline and have spent just under half the budget' would sound fine. It isn't. We've done forty per cent of the work for forty-eight per cent of the money, at the halfway point. The forecast statement tells the truth: likely around six hundred thousand, unless something specific changes.

Worked example two: the UK appointment system

Back to the fictional UK public health agency. The new project manager introduced a burnup chart. Two lines: total scope, measured in story points or features, and work completed. The completed line rose steadily. But the scope line had risen forty per cent since the start, because new requests had been accepted without anyone recording the trade-off. Project both lines forward at their current rates, and they meet five months after the planned date. Now the board, seeing the real picture for the first time, could make real decisions. It agreed to move two features to a later phase and to add one experienced developer. The new forecast was met. The team hadn't been slow. The project had been growing invisibly, and the burnup made it visible.

Watch me do it: a burnup forecast

Let me build a burnup forecast. One row per week, with two numbers: cumulative work completed and total scope, both in the same unit, story points or items. Chart both as lines. Then the forecast. I take the average weekly completion over the last six weeks, not since the start, because recent performance is more relevant. Say it's eighteen points a week. Remaining scope is total minus completed: two hundred and seventy points. Two hundred and seventy divided by eighteen: fifteen weeks. Then a range, using the slowest and fastest of those recent weeks: roughly thirteen to eighteen weeks. And one more line on the chart: the trend in total scope. If scope is still growing, the finish date moves even if the team speeds up. That's the conversation the board needs to have.

Forecasting, not just reporting

Let's be precise about what a forecast is. It's the trend from recent performance, plus known changes that haven't happened yet, such as approved scope changes or a team member leaving, plus the effect of the top risks, expressed as a range. And every forecast needs its assumption stated: 'assuming scope stays at today's level' or 'assuming the vendor delivers the integration by week twelve'. Then one of the most useful habits in monitoring: compare this month's forecast with last month's. If it moved, explain exactly why, naming the changes. If it hasn't moved for months despite visible problems, be suspicious. Forecasts that never move usually mean nobody is really forecasting, just repeating the plan.

Status reporting and variance analysis

A status report should forecast. Structure it: a headline in two sentences; the forecast, cost and finish date, with a range; the top variances; the top risks with their trend; and the decisions needed, from whom, by when. Give RAG status explicit rules, for example green if the forecast is within tolerance, amber if it's outside tolerance with a credible recovery plan, red if it's outside tolerance without one. Explain each significant variance with a cause, the impact, the action and the owner, rather than a symptom like 'resourcing issues'. And combine leading and lagging indicators, so the report warns as well as describes. The common mistakes: reports that describe but don't forecast, undefined colours, dozens of indicators, and variances explained by symptoms.

Recap and try this now

Let's recap. Monitor with a small set of indicators, mixing lagging measures like progress, cost efficiency and forecasts with leading ones like risk trends, work in progress and issue age. Use earned value for predictive work, with EV first in every formula, and burnup charts for agile work, which make scope growth visible. Report forecasts, not just history, with ranges and defined RAG rules, and explain variances by cause, impact, action and owner. Your try-this-now: create a burnup chart or an EVM table for a real or practice project, and write a forecast statement with its assumptions. For example: at current rates we expect to finish in fifteen weeks, range thirteen to eighteen, assuming scope stays at today's level.

Key takeaways

• Measure schedule, cost, scope, quality, risk, benefits and team health, choosing measures that aid decisions.
• CPI = EV/AC and SPI = EV/PV; EAC ≈ BAC/CPI gives an honest cost forecast.
• Burnup charts show progress and scope growth separately.
• Define status colours by rules and always forecast, with assumptions and decisions needed.

Try it

Create a burnup chart or EVM table for a real or practice project and write a forecast statement with assumptions.

• Previous: Leading execution: kickoff, cadence and RAID
• Next: Risk and issue management for project leaders
• All lessons of Project Management Leadership with AI