Change orders · Nov 4, 2026 · 11 min read

Project post-mortem: start with the hours nobody billed.

An agency's post-mortem earns its hour when it starts from the timesheet, because the split between client requests that were priced, requests done for free and plain estimating misses decides which fix goes into the next quote.

A project post-mortem is a review held after a project ends, where the team compares what was planned with what actually happened and agrees what to change next time. For an agency, start it with the money rather than the mood: put the hours you quoted beside the hours you logged, sort every client request into priced or absorbed, and work out the hourly rate the project really paid. That one number tells you which lessons belong in the next proposal and which are just stories about a hard month.

Most guides to the meeting cover the agenda, the blameless tone and the action items, and they're right about all three. Productive's guide asks whether the project stayed on budget, and Plane's sample report compares planned and actual dates, but neither puts a dollar figure on the gap, so the lessons stay general. We'll walk one website project from its estimate to its effective rate, and then through the meeting that turned its numbers into three new lines in the next quote.

What is a project post-mortem?

A project post-mortem is a structured look back at a finished project, usually a meeting plus a short written report, that records what went well, what didn't, why, and what the team will do differently. You'll also hear it called a project retrospective, a lessons-learned review or a project recap, and Productive's guide uses that last name for the document the meeting produces. The word comes from medicine, which is why some teams dislike it after a project that went perfectly well, but the meeting is just as useful after a good project as after a bad one, because a profitable job can hide its own leaks.

For an agency the review has a narrower job than it does inside a product company. Your projects end with a client, an invoice and a launch date, so the questions that matter most are whether the price was right, whether the changes the client asked for were paid for, and whether the estimate for the same kind of work should go up next time.

How is a post-mortem different from a retrospective or a premortem?

A post-mortem looks back over a whole project once it's finished, a retrospective looks back over a short stretch of work while the project is still running, and a premortem looks forward before the work starts by imagining the project has already failed and asking why. Plane's guide separates the three the same way, with the retrospective as a recurring review held during the work and the post-mortem as the one-off review at the end.

For agency work the premortem has a useful money twin. Before you sign, ask which parts of the project are most likely to grow and what each would cost if it did, then write those prices into the proposal, the way a web design proposal template that still works in week six does. The post-mortem then checks whether your guesses were right, and the next premortem starts from the answer. A short retrospective halfway through a long build earns its place too, mainly as a chance to catch free requests while they're still small enough to price.

Why should an agency post-mortem start with the money?

An agency post-mortem should start with the money because agency projects usually fail quietly, as hours rather than as disasters. A website that launches on time for a happy client can still have paid your team well under the rate you quoted, and nothing in a conversation about what went well will surface that unless somebody brings the timesheet. Atlassian's guide lists budget reports among the data to gather before the meeting, which is the right instinct, but the steps that follow compare goals with results rather than the quote with the invoice.

Starting from the numbers also takes the heat out of the room. Saying that the design phase felt chaotic invites people to defend themselves, while saying that design ran nine hours over, all of them a third homepage round nobody priced, gives everyone the same fact to work from. It also stops the meeting from treating every overrun alike, which matters because the fix depends on where the hours went. Hours lost to your own estimate are a pricing lesson, hours lost to client requests you agreed to without a price are a process lesson, and the two need different changes.

How do you put a number on a finished project?

Compare four figures, which are the hours you estimated, the hours your team logged, the total you billed and the effective hourly rate that billed total works out to across the logged hours. Then sort every hour of overrun into one of three groups: client changes that were priced and approved, client changes the team did without a price, and plain estimating misses that had nothing to do with the client.

Take a five-person web agency in Bellingham, Washington, that rebuilt the website for Northfork Montessori, a school with campuses in Bellingham and Mount Vernon, adding tour booking and an online enrollment application. The agency estimated 240 hours at its $150 blended rate, which is the rate it uses across design and development once every hour of a website is counted, and signed a fixed fee of $36,000 on Monday, January 11, 2027. Imogen Castellanos, the school's head of admissions, was named in the agreement as the person who approves changes. The site launched on Thursday, April 22, and by then the team had logged 318 hours, which is 78 more than the estimate.

Seven requests came in from the school during the build. Three went through as written change orders, each priced at the blended rate and approved by Imogen before anyone started: syncing the second campus's tour calendar at 14 hours and $2,100, sending enrollment applications straight into the system the school already uses for student records at 12 hours and $1,800, and an alumni giving page at 7 hours and $1,050. Those 33 hours and $4,950 were billed on top of the fixed fee, so they didn't pull the effective rate down at all.

The other four were done without a price, because each one felt too small to raise on the day it was asked. A third homepage design round took 9 hours, a staff bios page took 6, a newsletter feed on the homepage took 8 and changes to the enrollment confirmation emails took 6. That's 29 hours, worth $4,350 at the agency's own rate, and none of it was ever invoiced. It's the same pattern our list of scope creep examples and what each one costs describes, where every request sounds reasonable on its own and the total only shows up afterward.

That leaves 16 hours the client had nothing to do with. The tour booking feature was estimated at 30 hours and took 46, because the school's calendar tool needed a workaround nobody spotted while the project was being scoped.

The billed total was $36,000 plus $4,950, or $40,950, and across 318 logged hours that comes to $128.77 an hour against the $150 the agency quoted. Had the four free requests gone through the same way as the other three, the project would have billed $45,300 and earned $142.45 an hour, and the remaining gap to $150 is the booking estimate. That split is the most useful thing the meeting will see, because it says the project lost $4,350 to a habit and $2,400 to a guess, and a habit and a guess need different fixes. Across a full year of projects the habit is usually the bigger number, which is what scope creep actually costs your agency once it's added up.

What questions should a project post-mortem ask?

The questions that earn their place are the ones you can answer from a record rather than a memory, so lead with the money and come to the softer questions once everybody has the same numbers in front of them. For an agency, these five do most of the work:

After those, the usual questions from Productive and Atlassian still apply: how the team worked together, whether the handoff from design to development was clean, and what you'd want to repeat. The third question can sound like a hunt for someone to blame, though its purpose is to find the moment each request arrived without any route to a price, which is a gap in how your agency takes requests rather than a fault in whoever answered the call.

How do you run the meeting without it turning into blame?

Hold the meeting within two weeks of launch, keep it to about an hour, share the numbers a day before, and ask someone who didn't run the project to chair it. Productive's guide suggests a moderator from inside or outside the team and a separate note-taker, which is worth copying even in a five-person agency, because the person who managed the project can't both explain their decisions and referee the discussion about them.

Northfork's review ran on Thursday, May 6, two weeks after launch, with the project manager, the designer, both developers and the account lead in the room. Plane's guide lists clients as possible attendees where it's appropriate, and we'd keep them out of the internal review and offer a separate twenty-minute call instead. The team will talk more plainly about the four free requests without the person who asked for them listening, and the client call has a different purpose, which is to hear what they'd change about working with you.

The blameless rule that every guide recommends is far easier to keep when the facts are already agreed. Nobody argued about whether the third design round happened, only about why there was no point in the process where someone stopped and priced it. The answer turned out to be dull and useful, since requests arrived by phone after the Tuesday check-in calls, the project manager agreed to them on the call to keep things friendly, and nothing written ever followed. That's a fixable process gap, and the same gap shows up in most of the advice on handling scope creep without losing the client.

What should a project post-mortem report include?

One page is enough, and it should hold the numbers, the reasons behind them and a short list of changes, each with an owner and a date. Northfork's report had five parts:

Plane's guide also recommends limiting the number of action items and naming an owner for each, and that's the part that decides whether anything actually changes. Keep the report wherever your next proposal gets written rather than in a folder nobody opens, because its real reader is the person pricing the next similar job, and the parts of the scope that went wrong are usually the ones the assumptions in the scope never covered.

Twenty-nine free hours became three lines in the next quote

The Northfork review ended with three changes, and each one was written as a price rather than a promise to try harder. Two homepage design rounds are included and a third costs $1,350, any new page the size of the staff bios page costs $900, and booking work that connects to an outside calendar tool is estimated at 46 hours instead of 30. The other two free requests, the newsletter feed and the email changes, became a rule rather than a price, so a request made on a call now gets a written change request with the price filled in the same day, sent to the named approver before anyone starts the work.

That's the reason to start with the money. The agency now knows its blended rate held on the work it priced and slipped on the work it didn't, and the next school or nonprofit site it quotes will carry those three lines from the first draft. The review took about an hour, and most of that hour went on the request log, which is why Docket keeps each request with its date, price and approval as the project runs, so the list is already there when the meeting starts.

Frequently asked questions

What is a project post-mortem?

A project post-mortem is a review held after a project ends, usually a short meeting and a one-page report, that compares what was planned with what happened, explains the gaps and agrees what to change next time. For an agency, the most useful version starts with the quoted hours, the logged hours and the effective hourly rate the project actually paid.

What is another name for a project post-mortem?

Common alternatives are project retrospective, lessons-learned review, project review and project recap. Some teams avoid the word post-mortem because it sounds like something died, though the meeting is just as useful after a project that went well, since a profitable job can still hide hours nobody billed.

What's the difference between a post-mortem and a retrospective?

A post-mortem reviews a whole project once it's finished, and a retrospective reviews a short stretch of work, often a two-week sprint, while the project is still running. A retrospective fixes how the team works this month, and a post-mortem fixes how you scope and price the next project of the same kind.

When should you hold a project post-mortem?

Hold it within about two weeks of launch or final handover, while people still remember why things happened and before the timesheets get archived. Share the hours, the billed total and the list of client requests a day before, so the meeting starts from agreed numbers.

What questions should you ask in a project post-mortem?

Start with the questions a record can answer: which phases ran over and by how much, which client requests were priced and which were done for free, and which estimate you'd change if you quoted the same project tomorrow. Then ask what worked, how the handoffs between people went and what the client said about the result.

Should the client attend an agency post-mortem?

We'd hold the internal review without the client and offer them a separate twenty-minute call. The team talks more plainly about unpriced requests when the person who asked for them isn't listening, and the client call has its own purpose, which is hearing what they'd change about working with you.

Where the figures come from The Bellingham agency, Northfork Montessori, Imogen Castellanos and every hour, rate, price and date built from them are illustrative, in US dollars. Descriptions of other guides are as published by Productive (updated May 2026), Plane (April 28, 2026) and Atlassian, read on November 4, 2026.