Why agencies lose money on projects

Running our 15-person studio, we rarely had a project that went wrong without warning signs. The warnings were there. A revision cycle that ran longer than quoted. A brief that shifted after kickoff. Hours logged against the wrong code. None of it was hidden. None of it surfaced fast enough to act on.

By the time we reconciled, the invoice was either out or about to go out. The project had already been delivered. There was nothing left to do.

This is the discovery gap: the time between when a project becomes unprofitable and when the agency finds out. Closing it is not about working harder or tracking time more carefully. It is about changing where in the process the information surfaces.

The three loss patterns and the numbers behind them

The unprofitable projects in most agencies trace back to three patterns. They are not new. The PMI Pulse of the Profession has been measuring them for over a decade. The numbers move a few points each year. The shape of the story does not.

Scope creep. PMI's 2024 report puts scope creep at 52% of projects, with an average 27% cost overrun on the affected projects. For a $40,000 fixed-fee build, that is roughly $10,800 of margin gone. Multiply across a portfolio of 25 projects a year, and the figure comfortably clears six digits before anything else has gone wrong.

Unbilled time. Industry estimates of uninvoiced billable hours vary. The most-cited figure is around 15% of worked hours never billed, with a long tail of agencies above 20%. Some of that is genuine non-billable work. A larger share is hours absorbed because nobody logged them in time, the brief shifted and the variation was never raised, or the senior on the project did the client comms quietly and never wrote it down.

Underestimated projects. Estimates get sent before the brief is fully understood. PMI's "inadequate requirements" category is the most-cited primary cause of failure in their survey, sitting at around 35% of failed projects. In agency life this is the project that came in at half the hours the team eventually delivered. By the time the gap is obvious, the price is signed and the team is already in week three.

These patterns compound. A scope shift on a project that was already underestimated, with hours not being logged in real time, is the classic loss pattern. None of the three on its own would sink the work. All three at once usually does.

Why the warnings do not surface

In every loss-making project, the warnings exist before the loss is final. They are just spread across a stack of tools that nobody is watching as one picture.

The proposal is in PandaDoc or a Google doc. The brief is in Notion. The plan is in ClickUp or Asana. The time is in Harvest or Toggl. The invoice is in Xero. None of them know what the others know.

A designer logs 12 hours against a phase that was quoted at 8. The time tracker records it correctly. The project tool shows it as "in progress". The proposal does not move. The invoice has not been raised yet. No single screen surfaces the over-burn, so nobody acts on it. By the time anyone reconciles, the project is delivered and the conversation about a variation is months too late.

This is the architecture problem at the heart of most agency margin loss. It is not a discipline failure on the team. It is a system that physically cannot show the picture until after the picture matters.

The maths on a typical agency

Take a 15-person agency on $2.5 million of annual revenue. A 20% net margin would put profit at $500,000. Now apply the loss patterns.

If 52% of projects experience scope creep at a 27% cost overrun, and project revenue makes up around 70% of the agency's book ($1.75M), the back-of-envelope drag from creep alone is roughly $245,000 a year. If 15% of worked hours are unbilled across a delivery team costing $1M loaded, the unbilled share is around $150,000. Underestimated projects sit on top of that.

The numbers are deliberately rough. The point is the order of magnitude: a profitable-looking agency can have half its theoretical margin eaten by structural leakage and not realise, because the leakage is spread across projects that each look "fine" on their own.

Closing the discovery gap

Three changes carry most of the result. None of them are about working harder.

Put margin in the room when the work is being done. A team that can see, mid-project, that they have spent 70% of the budget for 50% of the deliverable will raise it on a Wednesday. A team that finds out at month-end will not, because there is nothing to do about it then. AgencyFlo exists because we needed exactly this and could not buy it.

Make the variation the cheapest path. When raising a scope variation is a five-minute action in the same tool you are already in, more variations get raised. When it requires opening Pandadoc, finding the master template, getting the partner's sign-off and remembering to send it, fewer do. The conversation with the client is usually easier than the team thinks. The barrier is internal, not external.

Track hours against the phase, not the project. Phase-level margin catches problems weeks earlier than project-level margin. A discovery phase running 30% over does not look bad against a six-month engagement. It does look bad against the eight days it was scoped for. The earlier the unit, the earlier the warning.

What to do this week

Pick the next project you scope. Estimate the hours at phase level, not project level. Log time against the phase, not the project. Look at the burn at the end of week one. If you cannot see it on a single screen without exporting anything to a spreadsheet, that is the gap that explains most of your margin loss. Closing it is the single highest-leverage change available to most agencies.

The work is not the problem. It almost never is. The structural lag between the work and the visibility is the problem, and it is fixable.