What scope creep actually costs a small agency

Scope creep is not an anomaly. It is the base rate.

PMI's Pulse of the Profession 2018 — a survey of 5,402 project professionals — found that 52% of projects completed in the previous twelve months experienced scope creep or uncontrolled changes to scope, up from 43% five years earlier. That is not a study of agencies, and it is worth saying so plainly. It is a study of project work in general, which is the closest large, honest dataset there is.

The interesting part of that report is not the headline. It is what happens to the organisations that are demonstrably good at this. PMI splits respondents into "champions" (80% or more of projects delivered on time, on budget and to business intent) and "underperformers" (60% or fewer). Among champions, an average of 33% of projects still experienced scope creep. Among underperformers, 69%.

So the best-run organisations in a survey of five thousand professionals cannot get scope drift below roughly one project in three. If your plan for next year is "scope it better", the base rate says a third of your work will drift anyway. The question a 5-to-25-person agency actually has to answer is not how to prevent scope creep. It is what happens in the fortnight after it starts.

The loss is not the extra work. It is the invoice that never happens.

Ignition surveyed 273 agency managers and executives for The 2025 Agency Pricing & Cash Flow Report, published in May 2025. Two findings sit next to each other and, read together, describe the whole problem.

And the price of that: 57% of agencies lose between $1,000 and $5,000 every month on unbilled projects and tasks, while 30% say scope creep costs them more than $5,000 a month. At the low end of the first band that is $12,000 a year; at the top of the second it is open-ended.

Notice what those two numbers do when you put them in order. The scope changed. You did the work. You did not bill it. Three separate events, and only the third is entirely inside your control — which is exactly why it is the one worth attacking. You cannot legislate away a client changing their mind. You can decide whether the change becomes a priced, agreed line item or a gift.

The same survey suggests why the gift is so easy to give. Ignition's respondents price their work as hourly (28%), productised, subscription or tiered (28%), project-based (25%) and retainer-based (10%). Only the first of those bills extra work automatically. In the other roughly seven-in-ten shapes, unsold work has no home unless somebody makes one for it — deliberately, by name, in the middle of a delivery week.

Do the arithmetic on your own agency

Industry averages are a bad way to run a business and a good way to start a conversation. The number that matters is yours, and you can have it in an afternoon.

1. List every project you delivered last quarter.
2. Mark the ones where what you finally shipped was not what you sold. Not "the client was difficult" — literally: an extra round, a page nobody scoped, a second stakeholder who wanted a version, a platform change.
3. For each one, write down the hours that went into the difference. This is an internal measurement and it never touches a client-facing document; you are counting, not billing.
4. Price those hours at whatever rate you would have quoted, had you quoted.
5. Add it up. That is one quarter.

An illustration, with made-up inputs and real arithmetic: a twelve-person studio delivers nine projects a quarter. Five of them drift — roughly the PMI base rate. Each drift costs six hours of work nobody sold, which is a modest estimate for "one more round". That is 30 hours. At €120 an hour, €3,600 a quarter, €14,400 a year — for a shop where the drift is small and frequent rather than catastrophic and rare. Ignition's measured band ($1,000–$5,000 a month) lands in the same order of magnitude, which is mildly reassuring about the arithmetic and not at all reassuring about the situation.

Why the change order doesn't get sent

If 78% of agencies rarely charge for out-of-scope work, the explanation cannot be that 78% of agency owners lack nerve. Four mechanical reasons show up again and again, and none of them is about courage.

You noticed too late

Drift arrives as single small requests, each of which is genuinely fine on its own. The pattern only becomes obvious once the work is done — and asking to be paid for finished work is a completely different conversation from asking to be paid for work you have not started.

You can't show the change

"It feels like more than we agreed" is a feeling, and a client is entitled to disagree with a feeling. "This is the third round of revisions on a deliverable we scoped at two" is a fact with a number attached, and a reasonable client cannot really argue with it. The difference between the two sentences is whether anyone wrote the baseline down in countable units.

Nobody owns it

On a small team, the person who sees the drift first is the person doing the work — a designer, a developer, an editor. They are the least well placed and the least inclined to open a money conversation, and by the time it reaches whoever does own money, the week has moved on.

The vehicle is too heavy

If raising a change means drafting a new document, getting it reviewed and chasing a signature, it will lose every time to "we'll just do it and sort it out later". Later does not arrive. Nothing about that is irrational: the cost of raising the change was higher than the value of the change.

None of these four is a discipline problem, which is why the advice to "learn to say no" has not moved the number in the years people have been giving it. They are problems of evidence, ownership, timing and friction — and those are fixable.

The window is roughly a week, and it is not a data point

Here is a claim without a citation, flagged as such: the difference between an easy change order and an impossible one is mostly when you send it. We could not find a study that puts a number on that window, so treat what follows as a mechanism rather than a measurement.

Sent while the request is still fresh, a change order is a quote. The client is in the middle of deciding whether they want the thing; you are answering the question they are already asking, and "how much?" is a normal part of that answer. Sent after the work has shipped, the same message is a bill for a favour that has already been done. The client has to revise their own sense of what they were getting, and revise it downward. That is why the conversation feels bad — not because the money is unfair, but because you are asking someone to un-receive something.

Everything practical below is downstream of that one observation: the value of a change order decays fast, so the whole game is noticing early enough to send it while it is still a quote.

What actually moves the number

1. Write the baseline in countable units. Not "a website" — five page templates, two revision rounds each, one round of copy edits, one stakeholder review. A baseline that cannot be counted cannot be shown to have moved, and a baseline nobody can show is a baseline that will not survive a disagreement.
2. Log the drift the day it happens, in one line, where the project lives. Not in order to bill it — in order to have it. Most drift never becomes a change order and that is fine; you cannot decide about something you did not record.
3. Send one line and a price, not a document. "That's a third round on the homepage — €480, and it pushes delivery by two days. Want me to book it?" A change order that takes ninety seconds to raise gets raised.
4. Make it the standard vehicle, not the exception. The first change order in a relationship is awkward. The third is administrative. What makes it awkward is being unusual, and the only cure for unusual is frequency.
5. Don't discount it. A discounted change order teaches a client that scope is negotiable and that pushing works. If the work is worth doing for free, give it away openly and say so; that buys goodwill. A quiet discount buys nothing.

None of these five needs software. A shared document and a habit will do it, and for a five-person shop with one project at a time they probably should. What software changes is whether the noticing survives a bad week — the week where two projects go long, someone is ill, and the person who would have written the line is the person putting out the fire.

There is a second loss standing behind the first

The same Ignition survey found that 63% of agencies suffer from unpredictable cash flow, that 71% say at least one in every four invoices is paid late, and that 84% spend at least three to ten or more hours a month chasing late payments.

Unbilled work and late invoices look like separate problems — one is a pricing failure, one is a collections failure. They are the same failure at two stages: work that was done but never became a dated, agreed, countersigned obligation with a number on it. Work in that state is not an asset. It is a hope, and hopes have very poor payment terms.

Which reframes what a change order is for. It is not a way to extract more money from a client who is already spending. It is the moment a piece of work stops being a favour and becomes an obligation — with a date, a price and an approval attached. Everything downstream, including whether you get paid on time, depends on that transition happening at all.

The sequence, and where we come into it

The whole argument compresses into four steps: sell a baseline that can be counted, notice drift against it while it is still fresh, turn the finding into a priced line the client can approve in one click, and let the approval become the invoice. Scope creep stops being a leak and becomes a revenue line — the same event, differently handled.

That sequence is what we are building Tave around: the deliverable as the atomic unit, deterministic watchers that compare what is actually happening against the baseline you sold, and a finding that turns into a one-click change order rather than an awkward conversation you keep postponing. We built it because the habit above is a good habit and habits lose to busy weeks, which is the only thing software is genuinely good for.

But the habit is the part that matters. If you take one thing from this post, take the afternoon with last quarter's projects and a spreadsheet. Most agencies have never seen their own number, and it is almost always larger than the guess.