TL;DR. In PMI’s 2018 Pulse of the Profession, 52% of projects completed in the previous 12 months experienced scope creep, up from 43% five years earlier. A review of software estimation surveys found that 60-80% of projects overrun effort or schedule, with average overruns of 30-40%. The ILO’s survey of workers on online freelance platforms found average paid hourly earnings of US$11.2 falling to US$7.6 once unpaid time was counted, a 32.1% drop. On a fixed fee, the arithmetic is simple: effective hourly rate = fee / (planned hours + unbilled extra hours). Extra hours equal to 40% of the plan leave you with 71.4% of your quoted rate; extra hours equal to the plan cut it in half. The CEO move: treat every out-of-scope request as a priced decision (absorb, bill, defer or refuse) and report effective rate per project, not just revenue. The student move: log planned against actual hours on every project, compute your own overrun ratio, and use it to correct the next quote.
Rate-setting is covered elsewhere on CEOtudent: the salary-to-rate calculation sets your floor, the pricing models comparison explains when hourly, fixed, retainer or value pricing fits, and the productized service framework shows how to package a fixed scope. This piece covers what happens after the price is agreed.
This article is general information, not legal or financial advice. Contract terms and consumer or freelancer protection rules differ by jurisdiction.
Why is scope creep a pricing problem, not a boundaries problem?
Most advice on scope creep is about confidence: say no, set boundaries. For anyone who sells a fixed fee, a capped retainer or a package, it is a pricing issue first.
PMI, the Project Management Institute, defines scope creep in its 2018 report as the uncontrolled expansion of product or project scope without adjustments to time, cost and resources. When the scope grows and the fee does not, the work is still done, at a lower price per hour.
That price per hour is your effective hourly rate:
Effective hourly rate = fee received / (planned hours + unbilled extra hours)
The unbilled extra hours have four usual sources:
- Extra rounds: revisions beyond what the plan assumed.
- Additions: small new requests absorbed because each one looked minor.
- Rework: redoing work because the requirement was unclear or changed.
- Coordination overhead: unplanned calls, messages and the time it takes to get back into the work after each one.
None of these appear on the invoice. All of them appear in the denominator.
The audience is large. In the US Bureau of Labor Statistics’ July 2023 Contingent Work Supplement, 11.9 million people were independent contractors on their sole or main job, 7.4% of total employment, up from 6.9% in May 2017. Of those independent contractors, 80.3% preferred their arrangement to a traditional job.
What does the evidence say about how often scope and effort overrun?
No public dataset tracks scope creep for freelancers. The evidence comes from neighbouring fields; each has limits, set out under the table, but they point the same way.
| Finding | Figure as published | Source |
|---|---|---|
| Projects that experienced scope creep or uncontrolled scope changes (past 12 months) | 52% in 2018; 43% five years earlier | PMI, Pulse of the Profession 2018 |
| Same measure, previous edition | 49% | PMI, Pulse of the Profession 2017 |
| Scope creep at “champion” vs “underperformer” organisations | 33% vs 69% of projects | PMI, Pulse of the Profession 2018 |
| Software projects that encounter effort and/or schedule overruns | 60-80% | Moløkken and Jørgensen (2003), review of surveys |
| Most common average cost overrun across those surveys | 30-40% | Moløkken and Jørgensen (2003) |
| IT projects: mean ratio of actual to estimated cost (4,677 projects with both figures) | 1.8, with the mode around 1.0 and a fat right tail | Flyvbjerg et al. (2022) |
| Students’ best-estimate vs actual time to finish a thesis | 33.9 vs 55.5 days | Buehler, Griffin and Ross (1994), Study 1 |
| Share who finished by their best estimate | 29.7% | Buehler, Griffin and Ross (1994) |
| Freelance platforms: average hourly earnings, paid time only vs paid plus unpaid time | US$11.2 vs US$7.6 | ILO, World Employment and Social Outlook 2021 |
| Freelance platforms: unpaid work per hour of paid tasks | About 23 minutes | ILO, World Employment and Social Outlook 2021 |
| Self-rated stress after interrupted vs uninterrupted work (1-20 scale) | 9.46 (same context) and 9.13 (different context) vs 6.92 | Mark, Gudith and Klocke (2008) |
Sources: PMI Pulse of the Profession 2017 and 2018; Moløkken and Jørgensen (2003); Flyvbjerg, Budzier, Lee, Keil, Lunn and Bester (2022); Buehler, Griffin and Ross (1994); ILO (2021); Mark, Gudith and Klocke (2008). Figures as printed.
The PMI figures describe organisations, not freelancers. The 2018 report draws on 4,455 project management practitioners, 447 senior executives and 800 PMO directors, and its percentages are averages of respondents’ estimates (per the question wording printed in the 2017 edition). Even the best-run organisations saw scope creep on a third of projects: good process reduces it, but does not remove it.
The overrun distribution is lopsided. Flyvbjerg and colleagues found that most IT projects land close to budget, but a minority overrun by multiples, and those projects pull the mean up to 1.8. For a freelancer that means two risks: a steady small leak on most projects, and the occasional runaway project.
Change requests are a named cause, not an inference. In the Phan survey summarised by Moløkken and Jørgensen, respondents attributed cost overruns most often to over-optimistic estimates (51%), closely followed by changes in design or implementation (50%). In Lederer and Prasad’s study, also summarised there, “frequent requests for changes by users” was the top-rated cause of inaccurate estimates. Both are cited as reported in the review.
The ILO gap is broader than scope creep. The ILO’s unpaid time includes looking for work and building a platform profile, not only unbilled project hours. It is the clearest official measure of how far unpaid time lowers real hourly pay, but it is not a scope-creep statistic.
In the Mark study, interrupted participants finished their tasks faster, with no difference in quality, but reported more stress, frustration, time pressure and effort. Some of the cost of interruption is paid in strain rather than minutes, so the worked example below treats re-entry time as an assumption.
How much does each unbilled hour actually cost?
If unbilled extra hours equal a share x of the planned hours, your effective rate becomes 1 / (1 + x) of the quoted rate. The table below applies that rule to a quoted rate of $150 per hour on a $6,000 fixed fee and marks where published overrun figures fall.
CEOtudent analysis: break-even table for unbilled hours (quoted rate and fee are illustrative assumptions)
| Unbilled extra hours, % of planned | Effective rate, % of quoted | Effective rate on a $150 quote | Fee needed to hold $150 | Published reference point |
|---|---|---|---|---|
| 5% | 95.2% | $142.86 | $6,300 | |
| 10% | 90.9% | $136.36 | $6,600 | |
| 25% | 80.0% | $120.00 | $7,500 | |
| 35% | 74.1% | $111.11 | $8,100 | Inside the 30-40% average overrun range (Moløkken and Jørgensen) |
| 50% | 66.7% | $100.00 | $9,000 | |
| 63.7% | 61.1% | $91.63 | $9,822 | Thesis completion: actual time 63.7% above best estimate (Buehler et al.) |
| 80% | 55.6% | $83.33 | $10,800 | Mean IT cost ratio of 1.8 (Flyvbjerg et al.) |
| 100% | 50.0% | $75.00 | $12,000 |
Method: effective rate = quoted rate / (1 + x); fee needed = $6,000 × (1 + x). The reference points measure different things (effort, calendar time, cost) and are shown only to place the scenarios against documented overrun sizes. Recomputed in the script kept with this article.
A 10% leak costs one dollar in eleven and goes unnoticed. A 35% leak, inside the documented average range, costs about a quarter of the rate. The “fee needed” column is the honest price of the job if you already expect that leak. Charging $6,000 for work you know tends to run 35% over is a decision to work at $111 an hour while telling yourself you charge $150.
Your rate halves only when unbilled extra hours equal the planned hours. That is not the typical case, but it sits within the tail Flyvbjerg’s team found, and a single project can reach it through a series of reasonable concessions.
How does a $150 quote become a $75 hour?
Every input below is an illustrative assumption chosen to show how the layers stack; none is a measured average.
CEOtudent analysis: illustrative effective-rate erosion on one fixed-fee project
| Layer (all inputs are illustrative assumptions) | Hours added | Cumulative hours | Effective rate | % of quoted rate |
|---|---|---|---|---|
| Planned work: $6,000 fee, 40 hours | 0 | 40 | $150.00 | 100.0% |
| Three extra revision rounds at 4 hours each | 12 | 52 | $115.38 | 76.9% |
| Eight “quick” additions at 1.5 hours each | 12 | 64 | $93.75 | 62.5% |
| Rework after a requirement changed | 6 | 70 | $85.71 | 57.1% |
| Unplanned calls, email and chat | 6 | 76 | $78.95 | 52.6% |
| Re-entry time: 16 context switches at 15 minutes | 4 | 80 | $75.00 | 50.0% |
Method: effective rate = $6,000 / cumulative hours. All hour inputs are illustrative assumptions; the arithmetic is reproduced in the script kept with this article.
Read from the top, each step looks defensible. Read from the bottom, the project paid half of what was quoted.
Extra rounds and absorbed additions account for 24 of the 40 extra hours, and they are the easiest to control by contract, through a revision cap and a change-request price. Clearer acceptance criteria reduce rework, and a meeting cadence in the statement of work limits coordination; CEOtudent’s analysis of what meetings actually cost has the time data.
What do change orders recover?
A change order is a short written agreement that adds scope, price and, where needed, time to an existing contract. Here is the same illustrative project under three policies.
CEOtudent analysis: the same illustrative project under three scope policies
| Policy (illustrative assumptions) | Revenue | Hours worked | Effective rate | % of quoted rate |
|---|---|---|---|---|
| A. Absorb everything | $6,000 | 80 | $75.00 | 50.0% |
| B. Revision cap plus change orders: the 12 hours of extra rounds and 12 hours of additions billed at $150 | $9,600 | 80 | $120.00 | 80.0% |
| C. Policy B plus a 10% scope contingency priced into the fee, and a call cadence that halves unplanned calls (6 to 3 hours) | $10,200 | 77 | $132.47 | 88.3% |
Method: Policy B revenue = $6,000 + 24 hours × $150; Policy C revenue = $6,000 × 1.10 + $3,600, hours = 80 – 3. Recomputed in the script kept with this article.
Billing changes recovers most of the loss, not all of it. Even under Policy C, 13 hours of rework, calls and re-entry time stay unbilled. The goal is a known, priced leak, not zero.
Contingency is not padding if you can show the data. A 10% contingency is a modest version of what Flyvbjerg calls an uplift in reference class forecasting: an adjustment taken from the actual performance of comparable past projects. In his 2006 example for UK road projects, accepting a 50% risk of cost overrun required a 15% uplift, while accepting only a 10% risk required 45%. A freelancer’s own project log is the reference class.
Change orders change behaviour, not just revenue. A 1.5-hour “quick” addition looks different when it is quoted at $225, and some requests will be dropped. That moves the decision to the person who benefits from the change.
Two pricing conventions are common. One is to price change hours at the quoted rate plus a small premium for re-entry and admin; at 15%, a 3-hour change at $150 becomes $517.50 instead of $450. The other is a minimum charge per change request, for example one hour, so that a 20-minute tweak is billed at $150 rather than $50. The minimum discourages micro-requests, where re-entry costs pile up.
Absorb, bill, defer or refuse: what is the decision test?
Absorbing small changes is good service; absorbing without a limit is the problem. The framework below gives each request one of four outcomes, with thresholds written down before the project starts.
CEOtudent editorial framework: scope-change triage
| Outcome | Use when | Thresholds (set your own; these are starting points) | What you say | What you log |
|---|---|---|---|---|
| Absorb | Small, inside the agreed outcome, and the goodwill budget is not used up | Under 1 hour (2.5% of a 40-hour plan) and cumulative absorbed time under a goodwill budget of 5-10% of planned hours (2-4 hours here) | “Done. I’ve logged it under the included allowance.” | Hours, date, request; remaining allowance |
| Bill | Inside the project’s purpose but beyond the cap or allowance | Over 1 hour, or any request once the allowance is spent, or any revision round beyond the cap | “Happy to. That’s a change request: X hours, $Y, adds Z days. Shall I go ahead?” | Signed or written approval before work starts |
| Defer | Useful, but not needed for the current deliverable, or it would move the deadline | Would push the delivery date, or needs a decision the client has not made yet | “Let’s put this in the phase 2 list so it doesn’t delay launch.” | Backlog item with an estimate; feeds the next proposal |
| Refuse | Outside your competence, changes the agreed outcome, or would sink the project below your floor even if billed | Effective rate on the project projected below your walk-away floor, or the change turns the project into a different project | “That’s a different project. I can quote it separately, or recommend someone.” | Reason, and whether a separate quote was offered |
CEOtudent editorial framework, not a validated instrument. Thresholds are design choices; calibrate them against your own project log.
Ask four questions in order:
- Is it inside the outcome the contract describes? If no, it is a new project: quote it separately or refuse.
- Does it fit in the remaining goodwill allowance and take under an hour? If yes, absorb and log it.
- Is it needed for the current deliverable and date? If no, defer it to a phase 2 list.
- Billed at your change rate, does the project stay above your walk-away floor? If yes, bill it. If no, refuse or renegotiate the whole project.
The fourth question links scope control to the decision to stop. A stream of change requests often signals that the original scope was wrong. CEOtudent’s guide to kill criteria explains how to set that line before you are under pressure.
The scope-control checklist
Written for fixed-fee and package work; on pure hourly billing, scope creep moves from your rate to the client’s budget.
Statement of work clauses
– Outcome and deliverables, listed. What will exist at the end, in what format, how many.
– Acceptance criteria. How both sides will know a deliverable is finished.
– Exclusions. A short “not included” list.
– Revision cap. For example, two rounds of consolidated feedback per deliverable, with the price of each extra round stated.
– Goodwill allowance. An explicit number of hours for small changes, so absorbing is planned, not a leak.
– Change-request clause. How requests are made, how they are priced (rate, premium, minimum), and that work starts only after written approval.
– Communication cadence. Number and length of included calls, and the channel for feedback.
– Client dependencies. What the client supplies and by when; their delays move the schedule, not your fee.
– Payment terms. Amounts, dates and what triggers each payment.
Some of this is law in places. In New York City, the Freelance Isn’t Free Act requires contracts worth $800 or more, including agreements that total $800 in any 120-day period, to be in writing, and the written contract must state the work to be performed, the pay and the payment date. If no payment date is set, payment is due within 30 days of completing the work. New York State added similar protections to its General Business Law (Article 44-A) on 28 August 2024. Both the city and state agencies publish model contracts. Rules elsewhere differ.
Running the project
– Keep a scope log: request, date, hours, triage outcome, approval.
– Track hours by layer (planned work, extra rounds, additions, rework, coordination), not just total hours.
– Compute the projected effective rate at the midpoint of the project, not after the final invoice.
AI-assisted scope logging
– Give an AI assistant the statement of work and each incoming request, and ask whether the request falls inside the deliverables, the exclusions, or neither. Treat it as a first sort; the decision stays with you.
– Use it to draft change-request notes from your log; edit before sending.
– Check your contract and the client’s confidentiality terms before putting client material into any AI tool. CEOtudent’s piece on billing and disclosure for AI-assisted work covers the disclosure question.
After the project
– Record planned hours, actual hours and the ratio.
– Note which layer grew most and which clause would have caught it.
– Decide whether this client’s work belongs on a retainer instead: recurring, evolving requests are often a retainer that was sold as a project.
How do you calibrate your own overrun ratio?
The planning fallacy explains why a checklist is not enough. In Buehler, Griffin and Ross’s first study, students predicted they would finish their theses in 33.9 days on average and took 55.5. Fewer than a third finished by their best estimate. Even their pessimistic estimates (48.6 days) fell short of the actual time, though not significantly. Yet students who predicted longer did take longer: the estimates were biased, not useless.
That calls for a correction. Reference class forecasting, as Flyvbjerg describes it, bases a forecast on the actual performance of a class of comparable projects. For a freelancer, the comparable class is your own past projects of the same type.
A simple version, with illustrative numbers:
- List your last five comparable projects with planned and actual hours. Suppose the ratios of actual to planned are 1.10, 1.20, 1.22, 1.30 and 1.50.
- Take the median (1.22) as your correction factor. The median resists the one runaway project.
- Multiply your next bottom-up estimate by it. A 40-hour estimate becomes 48.8 hours.
- Price from the corrected hours, or keep the price and tighten the scope until the corrected hours fit it. At $150, holding your rate on 48.8 hours needs a $7,320 fee.
Repeat quarterly. A falling ratio means your scope controls work; a rising one means a layer grew.
The CEO and the student in scope control
The CEO owns effective rate per project against the quoted rate, and asks not whether a client is difficult but whether the contract prices the behaviour the business sees: goodwill decided in advance, written down, and scope treated as stock that is sold, given away on purpose, or kept.
The student treats each project as a data point, planned against actual, layer by layer. People underestimate their own tasks even after past misses; the fix is not willpower but a written record consulted before the next quote.
What this does not mean
- It does not mean billing every small favour. The aim is a limit, not zero.
- It does not mean the “half” figure is typical. It comes from an illustrative example in which unbilled hours equal planned hours. The research points to average overruns of roughly 30-40% in software projects, with a minority far worse.
- It does not mean fixed pricing is a mistake. Fixed fees reward efficiency; scope controls make them safe. Where scope cannot be defined, another model may fit better.
- The cited studies are not freelancer studies. PMI surveys organisations, the software surveys cover development projects, the thesis study covers students, and the ILO data covers online platform workers in 2019-20. They support the direction of the argument, not a freelancer benchmark.
Frequently asked questions
What is an effective hourly rate?
The fee you actually received divided by all the hours the project took, including unbilled revisions, additions, rework and coordination.
How much scope creep is normal?
There is no freelancer-specific benchmark. In PMI’s 2018 survey, 52% of organisations’ projects experienced scope creep, and a review of software estimation surveys put typical average overruns at 30-40%. Your own project log is a better guide than any average.
Should every revision be charged?
No. Include a stated number of revision rounds in the price, plus a small goodwill allowance in hours. Charge for rounds and requests beyond those, at a price written into the contract.
How should a change order be priced?
Estimate the hours, multiply by your rate, and add a premium for re-entry and admin if you use one. State the effect on the delivery date in the same note.
Does hourly billing solve scope creep?
It protects your rate, because every extra hour is billed. It does not protect the client’s budget or the relationship, and the client carries the overrun risk. A fixed core scope with priced change requests combines both.
What if a client refuses to pay for changes?
Go back to the contract: the change clause, the revision cap and the exclusions. If the project’s projected rate falls below your walk-away floor, renegotiating or ending the engagement may cost less than continuing.
Sources
- Project Management Institute (2018). Success in Disruptive Times: Expanding the Value Delivery Landscape to Address the High Cost of Low Performance. Pulse of the Profession 2018. PMI.
- Project Management Institute (2017). Success Rates Rise: Transforming the High Cost of Low Performance. Pulse of the Profession 2017. PMI.
- Moløkken, K., and Jørgensen, M. (2003). A review of surveys on software effort estimation. Proceedings of the 2003 International Symposium on Empirical Software Engineering (ISESE 2003), IEEE. Author copy from Simula Research Laboratory. (Phan and Lederer and Prasad figures cited as summarised in this review.)
- Flyvbjerg, B., Budzier, A., Lee, J. S., Keil, M., Lunn, D., and Bester, D. W. (2022). The empirical reality of IT project cost overruns: Discovering a power-law distribution. Journal of Management Information Systems, 39(3), 607-639.
- Flyvbjerg, B. (2006). From Nobel Prize to project management: Getting risks right. Project Management Journal, 37(3), 5-15.
- Buehler, R., Griffin, D., and Ross, M. (1994). Exploring the “planning fallacy”: Why people underestimate their task completion times. Journal of Personality and Social Psychology, 67(3), 366-381.
- International Labour Organization (2021). World Employment and Social Outlook 2021: The role of digital labour platforms in transforming the world of work. ILO, Geneva. Sections 4.2.2 and 4.2.3.
- Mark, G., Gudith, D., and Klocke, U. (2008). The cost of interrupted work: More speed and stress. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI 2008), ACM, 107-110.
- US Bureau of Labor Statistics (2024, November 8). Contingent and Alternative Employment Arrangements, July 2023. News release USDL-24-2267.
- New York City Department of Consumer and Worker Protection. Freelance Workers: Your Rights under the Freelance Isn’t Free Act. Official guidance page.
- New York State Department of Labor. Freelance Isn’t Free Act. Official guidance page.
Note on CEOtudent analyses: the break-even table, the illustrative erosion table, the three-policy comparison, the change-order pricing examples and the overrun-ratio calibration example are CEOtudent calculations; every input in them that is not attributed to a source is an illustrative assumption, and all arithmetic is reproduced in a retained script. The derived figures in the evidence discussion (the 32.1% ILO drop, the 63.7% thesis overrun) are CEOtudent calculations from the published numbers. The scope-change triage table, the four-question test and the checklist are CEOtudent editorial frameworks, not validated instruments. Figures deliberately not used: the Standish Group CHAOS report’s average cost overrun of 89%, which the Moløkken and Jørgensen review found was not supported by other surveys (no primary CHAOS text was opened); the “23 minutes 15 seconds” interruption-recovery figure, which does not come from the CHI 2008 paper; and later PMI Pulse editions’ scope creep figures, because no primary copy of those editions could be opened.
This content was compiled with the support of AI following in-depth research, then written and prepared for publication by the CEOtudent editorial team.
This post is also available in:














