# How to run a weekly project margin review
Source: https://korrel.ai/blog/weekly-project-margin-review
Published: 2026-08-17
Updated: 2026-08-10
Topic: how to run a weekly project margin review
Tags: cost-tracking, margin-and-profitability
---
## Key points

- A margin review reads cost and forecast margin for every live job, not delivery status.
- Each number needs a threshold agreed in advance, or the meeting produces commentary instead of decisions.
- Cost drift is cheap to correct in week three and impossible to correct in the final fortnight.
- Every breached threshold should leave the room with one named owner and a date.
- Absorbing an overrun is a commercial choice worth making openly, not a default the delivery team falls into.

A weekly project margin review is a standing meeting where every live job is read out against its money. Five figures per job, and three that matter most: hours burned against budget, margin to date against target, and forecast margin at completion. Thirty minutes. Every job, every week. Knowing how to run one is a question of agenda rather than tooling, because the figures already sit in your timesheets and nobody reads them until the job closes.

Most delivery teams already meet weekly. They cover blockers, resourcing and what the client said on Tuesday. Money rarely appears, and when it does it arrives as a feeling rather than a figure.

## Why does a status meeting miss margin drift?

Because it was built to answer a different question. A status meeting asks whether the work is on track. A margin review asks whether the work is still worth the price you agreed, which needs different evidence and a different half hour.

Range's [project review template](https://www.range.co/templates/project-review) makes the point by omission. An opening round. Notable updates and blockers. Discussion, planning for the week ahead, and a closing round.

Not one financial line anywhere in it. Every item on that agenda is a report on progress, and progress against a plan says nothing about what the progress cost to produce. A job can be on schedule, well received, and losing money on every hour booked to it.

Timing is the rest of the argument. Cost drift caught in week three is cheap to put right. Caught in the final fortnight, it cannot be put right at all. That is the reasoning behind the [leading indicators that a job is losing money](/blog/job-losing-money-before-it-finishes) before it finishes. The weekly review is the meeting where somebody finally looks at them.

## Which numbers do you pull for each project?

Five. Any longer list stops being read aloud and becomes an attachment nobody opens.

| Number | Where it comes from | The point it becomes an action |
|---|---|---|
| Hours burned against budgeted hours | Timesheets against the estimate | Burn runs more than 10 points ahead of completion |
| Margin to date against target | Recorded cost against quoted value | Below target two weeks in a row |
| Unpriced change requests | The change log | Any request open longer than seven days |
| Days to the next milestone | The delivery plan | Milestone inside a fortnight with work outstanding |
| Forecast margin at completion | Current burn rate extended to the end date | Below the margin you would refuse the job at |

The last row carries the meeting. Margin to date is history, and history is only useful for the next quote. Forecast margin at completion is the one figure that describes a future you can still change. It is also the one most teams cannot produce without half a day in a spreadsheet.

The second row needs a target worth comparing against. A blended percentage inherited from a pricing conversation two years ago will not do. Calculating the [effective hourly rate on a fixed fee project](/blog/effective-hourly-rate-fixed-fee-projects) gives the target a rate behind it, which is what makes the comparison mean something.

## When does a number stop being a status update?

When it crosses a line you agreed before the meeting. Thresholds get set in advance precisely so they are not negotiated in the room by whoever most wants the project to look healthy.

- **Burn ahead of progress.** A job forty per cent through its hours and twenty-five per cent through its deliverables is not slightly behind. It is tracking towards well over its budget, and week four is when that arithmetic is still cheap to act on.
- **Margin below target twice.** One bad week is noise. Two consecutive weeks is a direction, and the second one is where you stop watching and start pricing.
- **An unpriced change older than seven days.** Every day it sits there the work gets closer to finished. Priced after delivery, a change request stops being commercial and becomes a request for a favour.
- **A milestone inside a fortnight with work outstanding.** This is the schedule threshold with money attached to it. A slipped milestone on a staged fee delays cash and adds hours.
- **Forecast margin below your floor.** Your floor is the margin at which you would have declined the job. Crossing it is an escalation, not a discussion point.

## How do you run a weekly project margin review in thirty minutes?

By reading exceptions, not the portfolio. Every live job gets a row in the pack. Only the jobs that breached a threshold get airtime, and the pack goes out the evening before so nobody presents numbers everyone can already see.

1. **Minutes 0 to 4. The roll-up.** Portfolio margin, and the count of projects at risk this week against last week.
2. **Minutes 4 to 20. Exceptions.** One job at a time. Breached number first, cause second, no narrative preamble.
3. **Minutes 20 to 27. Decisions.** Absorb, resequence or raise a change request. Say which one out loud.
4. **Minutes 27 to 30. Owners.** A name and a date against every decision taken.

Sixteen minutes of exceptions sounds thin until you try it. Three or four projects will breach something in a normal week, and four minutes each is enough when the cause is already on the page.

## Who owns the follow-up?

One named person per breach, with a date, recorded while everyone is still in the room. Not the team. Not the meeting.

Split ownership by what the action needs. A change request needs somebody who can hold a commercial conversation with the client: the account lead, not the delivery lead. Resequencing hours needs whoever controls the resource plan. Mis-booked time and wrong rates belong to whoever owns the timesheet data. Cheapest fix of the three, that last one.

The failure mode is easy to recognise. A number gets discussed, everyone agrees it looks bad, and the agenda moves on. It reappears the following week, worse. Two identical conversations about the same overrun mean the first one had no owner.

## When should you absorb an overrun rather than charge for it?

Sometimes, deliberately, and never by default. Over-servicing is a commercial choice with a price on it. The weekly review is where that choice gets made in the open, rather than at six o'clock on a Thursday by whoever happened to take the call.

Upsourced Accounting's guidance on [evaluating project margin](https://blog.upsourcedaccounting.com/evaluating-project-margin) puts a healthy project gross margin between 50% and 70%. Anything under 40% it treats as a red flag: the job was underpriced, over-delivered, or both. It points the over-serve judgement at a weekly delivery meeting with project management leadership. Same cadence, near enough the same room.

Retainers work the same way with a shorter fuse. Enterprise DNA's guide to [preventing retainer scope creep](https://enterprisedna.co/resources/guides/agencies-prevent-retainer-scope-creep/) locates the decision at the moment the request arrives, not at invoice time. That is the moment the account manager has three other fires burning and says yes. A weekly review will never catch that request on the day. It will catch the pattern by the following Monday, early enough to reprice the rest of the month.

## What does the review need from your data?

Granularity, mostly. Marcel Petitpas of Parakeeto frames the choice in his discussion of [how to track time in an agency](https://parakeeto.com/blog/how-to-track-time-with-marcel-petipas-episode-42/). Pick your level of detail by asking what you would want to investigate when things do not go to plan. If your time data cannot say which deliverable absorbed the hours, the meeting can establish that a job is over budget and nothing else.

Two habits make the difference:

- Book time against the line item it was quoted under, not against the project as a whole.
- Mark billable and non-billable at entry. Reconstructing it a fortnight later is guesswork wearing a suit.

Korrel's project analytics carry these figures directly: budget consumed, cost variance, planned against actual margin, forecast at completion, and billable against non-billable hours. The portfolio dashboard adds a queue of projects needing attention, with a recommended next action beside each. Views you open, not alerts that arrive. Which is exactly why the cadence matters. The rest of the [cost tracking](/blog/topics/cost-tracking) coverage deals with the mechanics underneath the meeting.

## Start with next Monday

Do not build a dashboard first. Take the five columns above into the delivery meeting you already hold, give the last ten minutes of it to money, and write down every project you cannot answer for. That list is your real data problem, and it will tell you more than any tooling decision you could make this quarter.

Then set the thresholds. Agree them once, with the person who signs off pricing, and hold them for a full quarter before revising. Thresholds that move whenever a project breaches one are just opinions with numbers attached.

## Common questions

### How long should a weekly margin review take?

Thirty minutes for a portfolio of twenty or so live projects. The time holds because only projects that breached a threshold get discussed. Everything else appears as a row in the pack and is never read aloud. If the meeting regularly overruns, the thresholds are set too tightly and every job is an exception.

### Who should attend a weekly project margin review?

The person who owns delivery, the person who owns the client relationship, and whoever can authorise a price change. Three or four people, no more. Delivery leads for individual projects can be pulled in for their own exception rather than sitting through the whole half hour.

### Is this different from month-end reconciliation?

Yes, and the difference is what you can still do about it. Month-end tells you what a job cost after the period closed, which is accounting. A weekly review looks at hours and forecast margin while the remaining work is still unbooked, so resequencing, rescoping and change requests are all still available.

### What if our timesheets are not accurate enough yet?

Start the meeting anyway and let it expose the gaps. The first few reviews will produce questions your data cannot answer, and that list is the specification for fixing time tracking. Waiting for clean data before starting the cadence delays both indefinitely.