The quarterly review
A fixed date to reread the ledger, not the moment something breaks, is what catches a project quietly draining hours it stopped earning.
There is a budget at the top of the ledger now, a total under it, and a name against whatever has to come off. Everything below those belongs to a project: a state, a maintenance figure, a verdict with a revisit date, an effective hourly rate, a kill criterion, a sunset plan. All of it was true on the day it was typed.
None of it stays true, and the file has no way of noticing. A maintenance estimate drifts. A competitor’s price moves. A revisit date passes and nothing opens on it. The ledger is a record of decisions made once, and nothing in the course so far has said when you go back and make them again. This lesson is that mechanism, and the mechanism is a date.
A date beats a trigger
The obvious alternative is a trigger: reopen the ledger when something happens. It is cheaper, it feels responsive, and it fails in one specific way that this entire course has been walking toward.
A trigger can only fire on an event. A build breaks. A bill jumps. A customer writes in angry. A dependency gets deprecated with a deadline attached. Every one of those is real, and every one of them already gets your attention without a ritual, because that is what an event does.
What a trigger cannot fire on is the absence of anything. A project that still works, still renews, still takes its share of the maintenance budget, and quietly stopped being worth what it takes, produces no event at all. There is no morning on which it announces itself. Nothing about it is broken. It is simply costing more than it returns, and it will go on doing that for as long as nobody opens the file.
So a review that runs on triggers reviews the broken things and never reviews the draining ones — and it is worse than merely incomplete, because it inverts. The project that breaks gets your attention, gets fixed, and the fixing goes into the hours-sunk column, which is the one column the triage module ruled out as an input to any decision. Attention flows to the loudest project rather than to the least defensible one.
There is a second problem with a trigger, and it is the same one the kill-criteria work already named. A trigger puts the timing of the decision in the hands of the person least able to be neutral about it. You reopen the ledger on a day you feel like reopening it, which is not a randomly chosen day. A date set in advance was chosen by a version of you who was not yet defending the answer.
Somebody else’s fixed re-decision point
The shape of this is not original, and there is a version of it written down clearly enough to quote. Shape Up, published by 37signals and credited on the page itself to Ryan Singer, describes a scheduled meeting whose entire job is deciding what happens next:
“The betting table is a meeting held during cool-down where stakeholders decide what to do in the next cycle.” (Singer, Shape Up, chapter 8)
The part worth taking is what the meeting assumes about work that already exists. Under a heading called “Keep the slate clean”:
“The key to managing capacity is giving ourselves a clean slate with every cycle. That means only betting one cycle at a time and never carrying scraps of old work over without first shaping and considering them as a new potential bet.” (Singer, Shape Up, chapter 8)
Read that sentence precisely, because the version of it that circulates is stronger than the text. It does not say old work never comes back. It says old work never comes back automatically — it has to be shaped and considered again as a new bet before it can. That is a forced re-justification, not a bar on continuing, and the distinction is the whole of what this lesson borrows. Continuing is allowed. Continuing by default, because nobody reopened the question, is not.
The same chapter puts a hard edge on it for work that overruns: “Teams have to ship the work within the amount of time that we bet. If they don’t finish, by default the project doesn’t get an extension.” Singer calls this a circuit breaker. Note the words “by default” there too — the chapter describes it as intentionally creating a risk, not as a rule with no exceptions. Both sentences are doing the same job: making the renewal of a commitment into an act somebody has to perform.
Where people get burned
Name what you are reading. 37signals gives the digital edition of Shape Up away, sells a print edition of it from its own shop, and sells the tool for running the method; the chapter itself closes with “We built Basecamp to execute the techniques in this book.” That does not make the argument wrong, and it is quoted here rather than paraphrased for exactly that reason. It does mean the confidence in it is a company’s confidence in its own method and its own product, and it should be read at that weight.
The chapter page wears no publication date of its own — the only date on it is a site-wide copyright line covering a range of years, which is a fact about the company, not about the text. Worth knowing before repeating a year for it.
And the setting does not transfer, so do not pretend it does. Their unit is a team’s next block of new work, decided by a group of people in a room; yours is a shelf of already-shipped things that nobody is pitching and nobody is betting on. Their circuit breaker fires on work that failed to finish, and nothing on your shelf is unfinished in that sense — the problem is the opposite one. What crosses over is the shape and nothing else: a moment on the calendar at which continuing has to be argued for.
Quarterly is a choice, not a finding
Nothing establishes that a quarter is the right interval for one person’s shelf of projects. I went looking for it — a study, a standard, any measurement comparing this interval against a shorter or a longer one for a solo portfolio — and there is none. What exists is business convention, practitioner opinion, and at least one piece of arithmetic dressed up afterwards as though it had been evidence.
The chapter above is worth reading once more on exactly this point, because it does not pretend either. Its own interval was settled the same way: “After years of experimentation we arrived at six weeks.” That is a report of what worked for one company, offered as such. It is not a measurement, and the chapter never claims it is.
So take the interval as a default and not as a result. It is easy to remember and easy to put in a calendar. If you have a reason to review more often or less often, that reason is better than the convention is. The argument of this lesson is entirely about there being a date, and not at all about which one. Everything that follows works unchanged at any interval you actually keep.
The agenda
Three moves, in order. Open the ledger. Re-run the gates. Decide again, in writing.
Open the ledger — the whole thing
Not the project you have been thinking about lately. The file, from the top, including the budget and the total under it, and including the projects you have no feelings about at all. The projects you have no feelings about are the reason this is on a calendar.
Re-run the gates
The gates page is the list you run and this lesson is the calendar entry that makes you open it. Each gate is a question with a pass test attached, and the review takes every project through all of them again.
Do not skip a gate because you already know the answer. The answer you already know is last time’s answer. A gate that passed comfortably is precisely the one you will not think to re-ask, and skipping it is not a shortcut — it is an assumption that nothing behind it moved, entered into the record as though it were a reading. That is the same failure as the trigger, running one level down.
Decide again, in writing
Every project leaves the review with a verdict, and the verdict gets typed even when it is the same word as last time. This is the step that looks like bureaucracy and is not. A verdict that was reaffirmed and a verdict that nobody reopened are identical in the file — same word, same line — unless a date got written beside it. The date is the only evidence that the review happened at all, and the next review reads it to find out.
Everything here runs on what the ledger already holds: estimates, one measured maintenance figure, and a rate built on top of both. Nothing in this review needs revenue you have not earned yet, and a review you cannot run until you are earning is a review you will never run.
The ledger is a file, on purpose
learning/monetizing/PORTFOLIO.md is markdown in a repository. It is not a form on this site, and it does not live in your browser. That is a deliberate choice and it pays off precisely here, at the review.
A file has a history. The hardest question in the whole agenda — did this verdict get reaffirmed, or is it just still sitting there — is a question about a diff, and a diff is something a tracked file has and a memory does not. A file can also be handed over whole: the teaching agent reads the ledger, so “here is the file, argue with me about it” is one paste rather than a retelling in which the inconvenient rows quietly go missing.
Retrieval check
The review comes round. You open the ledger, reach a project whose verdict already reads “pursue”, and you still think pursue is right. What do you write?
Check your answer
Write it again, with today’s date beside it. Nothing else in the file can distinguish a verdict you reconsidered and kept from one nobody has looked at since it was written, because both of them read “pursue”. The re-stamped date is the whole difference, and it is what the next review reads to know whether this one actually happened.
If reaffirming feels like a formality, that is worth noticing rather than skipping past — it is the same instinct that makes a trigger-based review feel sufficient. A commitment that renews itself when you are not looking is not a commitment you made. It is one you inherited from a previous version of yourself who had less information than you do now.
Hands on
Put the review in the calendar, then run the first one now
Done when: PORTFOLIO.md carries a new top-of-file Next review field holding a specific date that also exists in a real calendar, and every project in the file leaves today’s review with a verdict dated today — reaffirmed in the same words or changed, but typed either way, and with one sentence naming any gate whose answer moved.
- Open
learning/monetizing/PORTFOLIO.mdand add Next review at the top of the file, beside the maintenance budget and above the per-project sections. The ledger does not carry this field yet — you are adding it. It goes at the top for the same reason the budget does: it is a fact about you, not about any one project. - Write a date, not an interval. “Quarterly” in that field is a description of an intention; a day you could put in a calendar is a commitment. Then put that same day into whatever calendar actually interrupts you. The ledger field records the decision; the calendar is the thing that makes it arrive.
- Now run the first one, against the ledger exactly as it stands. Start at the top of the file rather than at the project you care about most.
- Reread the maintenance budget and the summed total beneath it. If either has moved, write the new figure and leave the old one visible next to it, so the drift is readable rather than overwritten.
- Take each project through the gates, in the order the page lists them, answering each pass test yes or no. Every gate, every project. A gate you skip is a gate whose answer you are assuming has not changed.
- Where a gate that used to pass now fails, write one sentence beside that project naming which gate moved and what moved it. Those sentences are the actual output of the review — the verdicts follow from them.
- Give every project a verdict with today’s date on it. The same word as last time is a perfectly good answer. An undated one is not an answer at all.
- Check the parked and killed rows specifically, because they are the ones a review skims. A kill verdict whose sunset plan has not actually been run means the project is still alive, still reachable, and still consuming its full line in the budget — the verdict was the decision, and removing the line is separate work.
- Confirm the cap the triage module set still holds: at most two projects in pursue. If the review moved something into pursue without moving anything out, it did not finish.
- Bring the file into the chat. I’ll push hardest on two things: any project reaffirmed in the same words with nothing written about why, and any gate recorded as passing that you did not actually re-ask. The second one is easy to do accidentally and it is the failure the whole ritual exists to prevent.
What this does not cover
The loop is closed. There is a date in the ledger, a date in a calendar, and a pass through the gates that produced a dated verdict for every project on the shelf — and it will happen again whether or not anything breaks between now and then.
All of that is aimed backwards, at things that already exist and already have your hours in them. What it does not tell you is how to choose the next thing, before it exists, before there is anything sunk into it to defend. Running this framework on something you have already built is the hard version because you are arguing with yourself. Running it on something you have not started is the honest test of whether any of it was actually learned. That is what comes next.
Read this next — primary source
Shape Up, chapter 8: The Betting TableRyan Singer, 37signals LLC — the digital edition is free to read in full online; the page carries no publication date of its own, only a site-wide 1999–2026 copyright line
Read it for the parts this lesson could not use. Its section headings are the giveaway — team and project size, uninterrupted time, what about bugs — and none of those describe a situation you are in. That is the reason to read the chapter whole rather than take the summary: watching a rule get argued for in a setting that is plainly not yours is the fastest way to see which part of it was the principle and which part was the staffing. The section called “Keep the slate clean” is where the transferable idea actually lives, and it is short. Then read the line under the final paragraph, where the company that gave the digital edition away names the product it sells. Looking for that line is a habit worth having in every method you adopt.
Stuck, curious, or think this lesson is wrong? Ask your teaching agent. The lessons are the scaffold; the conversation is where the learning gets unstuck.