--- title: "Measuring progress objectively | Optimize All Academy" description: "EVM is only as good as EV Earned value depends on how progress is measured. If supervisors guess \"about 60%\", the entire EVM system rests on opinion…" url: https://optimizeall.com/learn/project-controls-with-ai/measuring-progress updated: 2026-10-05 --- Project Controls in the AI Era · Cost control and earned value management · lesson 8 of 22 · 13 min # Measuring progress objectively ## EVM is only as good as EV Earned value depends on how progress is measured. If supervisors guess "about 60%", the entire EVM system rests on opinion. Objective progress measurement is the unglamorous foundation of credible controls. ## Earning rules (progress measurement methods) | Method | How it works | Best for | |---|---|---| | **0/100** | Earn 0% at start, 100% at completion | Very short tasks (within one reporting period) | | **50/50** (or 20/80 etc.) | Earn a fixed share at start, remainder at finish | Short tasks spanning two periods | | **Weighted milestones** | Pre-defined milestones each worth a fixed % | Engineering deliverables, procurement packages | | **Units complete** | Earn per unit installed (metres, tonnes, devices) | Repetitive physical work | | **Physical % complete** | Assessed percentage, backed by evidence | Where no better method exists; use with care | | **Level of effort (LOE)** | Earns with the passage of time (EV = PV) | Support work like project management | | **Apportioned effort** | Earns in proportion to a related discrete task | Quality inspection tied to construction | The earning rule is set at baseline and should not change without approval; otherwise progress can be "improved" by changing the rule. ## Weighted milestones: a template ``` Procurement package: Chillers (budget 300,000, illustrative) Milestone Weight Earned when 1. Specification approved 10% Signed-off spec 2. Purchase order placed 15% PO issued 3. Vendor drawings approved 15% Approved drawings 4. Factory acceptance test passed 20% FAT certificate 5. Delivered to site 25% Delivery note 6. Installed and tested 15% Test record ``` Each milestone has an **objective evidence** requirement. If the delivery note does not exist, the 25% is not earned, however confident the supplier sounds. ## Beware LOE Level-of-effort work always shows SV = 0, because EV equals PV by definition. If too much budget is LOE, performance problems are diluted. A common guideline is to keep LOE a small share of the budget and restrict it to genuinely time-based work such as project management or site security. ## Units and rules of credit for construction For bulk quantities, combine units with **rules of credit**: for pipework, for example, 20% on pipe delivered to workface, 50% on fitted, 20% on welded, 10% on tested. Each step is observable, and the total always equals 100%. ## Software and digital progress Progress in software and digital projects can use similar principles: - Earn story points or features only when they meet the **definition of done**. - Use weighted milestones for releases (designed, built, tested, deployed, accepted). - Avoid earning on "in progress" tickets unless partial rules are defined. ## Worked example: subjective vs objective *Illustrative.* A fictional fibre-rollout contractor in Texas reported 70% progress on a 50 km network based on supervisors' estimates. The controls team switched to a units-with-rules-of-credit method: trenching 40%, duct laid 20%, cable pulled 25%, spliced and tested 15%. Recalculated progress was 52%. The EV dropped, CPI fell from 1.02 to 0.76, and the real overrun became visible four months earlier than it otherwise would have. Painful, but it allowed a re-plan while there was still time. ## Validating progress - Sample-check progress claims on site or in the system each period. - Require evidence attachments (photos, test records, sign-offs) for milestone credit. - Compare EV trends with physical quantities and hours: if EV jumps but hours do not, investigate. - Watch for "front-loading": earning early milestones generously to flatter the numbers. ## Common mistakes - Allowing each supervisor to choose their own method. - Changing earning rules mid-project to improve the numbers. - Too much LOE, masking poor performance. - Earning milestones without evidence. - Using hours spent as a measure of progress. ## AI assistance Computer vision on site photos, drone surveys and sensor data are increasingly used to estimate installed quantities. These can make progress more objective, but they need calibration and a human check against the agreed rules of credit before EV is recorded. ## Quick self-check Pick any work package on your project and ask: if two different people measured its progress independently today, would they arrive at the same number? If the answer is no, the earning rule is too subjective. Tighten it by defining observable steps, attaching evidence requirements and agreeing the rule with the control account manager before the next period. Also check the mix of methods across the project: note what share of budget is level of effort, what share is subjective percent complete, and set a target to reduce both over the next few periods. ## Hands-on: a rules-of-credit sheet ```text Row 2 (weights): C2 Trench 40% | D2 Duct 20% | E2 Cable 25% | F2 Splice & test 15% G2 Weight check =IF(ROUND(SUM(C2:F2),4)=1,"OK","WEIGHTS ≠ 100%") Rows 3+: A Segment | B Budget | C:F 1 = step complete with evidence, 0 = not G3 Earned % =SUMPRODUCT(C3:F3,$C$2:$F$2) H3 EV =G3*B3 I3 Evidence link (photo, test record, sign-off) J3 Evidence check =IF(AND(G3>0,I3=""),"MISSING EVIDENCE","") Package EV =SUM(H:H) ``` ## Hands-on: a progress sanity check in Python Flag periods where earned value jumps but labour hours do not, a common sign of over-claiming: ```python import pandas as pd d = pd.read_csv("ev_hours.csv") # account, period, EV, hours d = d.sort_values(["account", "period"]) d["dEV"] = d.groupby("account")["EV"].diff() d["dH"] = d.groupby("account")["hours"].diff() ratio = d["dEV"] / d["dH"].where(d["dH"] > 0) typical = ratio.groupby(d["account"]).transform("median") d["flag"] = ratio > 2 * typical # earned far faster than usual per hour print(d[d["flag"]][["account", "period", "dEV", "dH"]]) ``` A flag is a question, not a verdict: prefabricated deliveries, for example, legitimately earn with few site hours. ## How to measure success - Share of budget measured by objective methods (target: rising; LOE and subjective % kept small). - Sample checks passed each period. - Two independent measurers agree within a small tolerance on sampled packages. ## Video lecture: Measuring progress objectively Lecture coming soon · 9 chapters · about 8 minutes. Read the full transcript below. 1. Earned value is only as good as earned value 2. Why it matters 3. The concept: earning rules 4. Worked example one: a chiller package 5. Worked example two: fibre rollout in Texas 6. Watch me do it: rules of credit in a sheet 7. Software, LOE and validation 8. Common mistakes 9. Recap and try this now ## Lecture transcript ### Earned value is only as good as earned value Ask a supervisor how far along their work package is, and you'll often hear 'about sixty per cent'. Ask again next week, and it's 'about seventy'. Ask a month later, and it's 'about ninety', where it will stay, stubbornly, for the next three months. Every earned value metric we calculated in the last lecture depends on one input: how much work is actually done. If that input is an opinion, the whole system is built on sand. In this lecture you'll learn the main earning rules, when to use each, how to design weighted milestones and rules of credit with objective evidence, how to handle software and digital work, and why level of effort needs careful control. By the end you'll be able to design an earning scheme that two different people would measure the same way. ### Why it matters Why does this matter so much? Because subjective progress is almost always optimistic, and optimism in earned value flatters your cost performance index. If earned value is overstated, CPI looks healthier than it is, forecasts look better than they should, and the real overrun surfaces months later, when there's much less you can do about it. Objective earning rules fix that. They make earned value auditable: anyone can check it against evidence. They also remove a lot of arguments. When the rule is agreed at baseline, the monthly conversation is about facts, not about whose estimate of 'about sixty per cent' to believe. ### The concept: earning rules Think of earning rules like the marking scheme for an exam. If the marking scheme is agreed before the exam, everyone knows what earns credit. If it's invented afterwards, you get arguments. Here are the main rules. Zero-one hundred: earn nothing at the start, everything at the finish. Great for tasks within one reporting period. Fifty-fifty, or twenty-eighty: a fixed share at start, the rest at finish, for tasks spanning two periods. Weighted milestones: predefined milestones each worth a fixed percentage, ideal for engineering deliverables and procurement. Units complete: earn per unit installed, for repetitive physical work. Physical percent complete, backed by evidence, only where nothing better exists. And level of effort, which earns with the passage of time, for genuine support work like project management. One rule above all: set the earning rule at baseline and don't change it without approval. ### Worked example one: a chiller package Here's a simple weighted-milestone scheme from the lesson, with illustrative numbers. A chiller procurement package, budget three hundred thousand. Specification approved, ten per cent. Purchase order placed, fifteen. Vendor drawings approved, fifteen. Factory acceptance test passed, twenty. Delivered to site, twenty-five. Installed and tested, fifteen. Total: one hundred. Now, suppose the first four milestones are complete, with a signed spec, an issued PO, approved drawings and a test certificate. That's sixty per cent, or one hundred and eighty thousand of earned value. The supplier swears the chillers are on a ship. Can we earn the delivery milestone? No. No delivery note, no twenty-five per cent. However confident the supplier sounds. That's the discipline that makes earned value trustworthy. ### Worked example two: fibre rollout in Texas Now a realistic scenario. A fictional fibre-rollout contractor in Texas is building a fifty kilometre network. Supervisors estimate progress at seventy per cent. The controls team introduces units with rules of credit. Trenching, forty per cent. Duct laid, twenty. Cable pulled, twenty-five. Spliced and tested, fifteen. They measure kilometres at each stage along the route. Verified progress: fifty-two per cent. Earned value drops sharply. CPI falls from one point nought two to nought point seven six. Painful? Absolutely. But the real overrun became visible about four months earlier than it would have. And four months is enough time to re-plan the crews, renegotiate with a subcontractor and have an honest conversation with the client, while there are still options on the table. ### Watch me do it: rules of credit in a sheet Let me show you how simple this is to set up. Across the top, the steps: trench, duct, cable, splice and test, with their weights in row two. First thing I add is a check cell that sums the weights. It must equal one hundred per cent, or the scheme is broken. Down the side, one row per segment, say each kilometre. In each cell, a one when that step is complete and evidenced, otherwise zero. Then the earned percentage for each segment is SUMPRODUCT of that row with the weights row. Multiply by the segment's budget and you have earned value by segment. Sum it, and you have earned value for the package. The last column holds a link to the evidence: photos, test records, sign-offs. If there's no evidence link, the cell stays zero. ### Software, LOE and validation Digital work follows the same principles. Earn story points or features only when they meet the definition of done. Use weighted milestones for releases: designed, built, tested, deployed, accepted. And don't earn on tickets that are merely 'in progress' unless you've defined partial rules. Now, level of effort. Because it earns with time, earned value always equals planned value, so its schedule variance is always zero. Too much LOE dilutes real performance problems, so keep it a small share of the budget and limit it to genuinely time-based work. Finally, validate. Sample-check claims each period, require evidence for milestone credit, and compare earned value trends with hours and quantities. If earned value jumps but hours don't, investigate. ### Common mistakes Let's be explicit about the mistakes, because they're very common. First, letting each supervisor choose their own measurement method, so progress across the project isn't comparable. Second, changing earning rules mid-project to improve the numbers. If a fifty-fifty task quietly becomes a weighted milestone with a generous first step, earned value rises without a single extra thing being built. Third, earning milestones without evidence. Fourth, too much level of effort, which masks poor performance. And fifth, using hours spent as a measure of progress. Hours tell you how hard people worked, not what they achieved. A useful self-test: if two different people measured this work package independently today, would they reach the same number? If not, the rule is too subjective. ### Recap and try this now Let's recap. Earned value is only as good as the way progress is measured. Choose an earning rule that fits the work, agree it at baseline, and attach an objective evidence requirement to every step. Weighted milestones and rules of credit beat opinions every time. Keep level of effort small, and validate progress every period against evidence, hours and quantities. On AI: computer vision on site photos, drone surveys and sensor data are increasingly used to estimate installed quantities. They can make progress more objective, but they need calibration and a human check against the agreed rules before earned value is recorded. Your try-this-now: pick a work package and design a rules-of-credit or weighted-milestone scheme with four to six steps, weights that total one hundred per cent, and an evidence requirement for each. ## Key takeaways - Credible EVM depends on objective, pre-agreed earning rules. - Weighted milestones and units with rules of credit are strong methods; subjective % is weakest. - Level of effort always shows SV = 0, so keep it limited. - Require objective evidence for every milestone claimed. ## Try it Choose a work package and design a rules-of-credit or weighted-milestone scheme with 4–6 steps, weights totalling 100% and an evidence requirement for each. - [Previous: Earned value management: the core metrics](https://optimizeall.com/learn/project-controls-with-ai/earned-value-fundamentals) - [Next: EAC and ETC: forecasting the final cost](https://optimizeall.com/learn/project-controls-with-ai/eac-and-etc-methods) - [All lessons of Project Controls in the AI Era](https://optimizeall.com/learn/project-controls-with-ai)