Three or four. That is the number, and the reasoning is arithmetic rather than taste, which we will get to in a moment. The thing worth understanding before the number is this: nobody scores your portfolio on its best project. A portfolio is a set, and a set gets judged at its weakest point — the moment a reader hits a thin spread with one render and a paragraph of studio-brief prose, every page that came before it is quietly revalued downwards.
So the default plan — everything, chronologically, first year onwards — is not a cautious plan. It is the riskiest thing you can do. It takes five or six years of work and guarantees that the reader's abiding impression is of your third-worst project.
Everybody arguing about how many projects to include is arguing about the wrong unit. The real constraint in an architecture portfolio is pages, not projects, and once you fix the page count the project count falls out of it.
Our own format guide lands on fifteen to twenty-five pages for the file you actually send. Take twenty-two as a working figure. Knock off a cover, a contents page, a CV or about page, and you are down to something like eighteen or nineteen pages of project content.
Now ask what one project needs to be believable. A context and concept spread. A page or two of development — sketches, massing iterations, the model you photographed badly at midnight. A resolved plan and section at a scale you can read. Something technical: a detail, a bay study, a construction sequence. Maybe a render or a model shot to close it. That is five pages at a squeeze, six if you want it to breathe.
Eighteen pages divided by six is three. Add a fourth at four pages if it is genuinely earning its place.
Put eight projects into the same eighteen pages and each one gets two pages. Two pages is one image and a caption. Do that across a whole portfolio and you have told the reader, very clearly, that you are a renderer — someone who makes attractive pictures and has never had to resolve a building. That may not be true. It will be the impression.
There is a blunt test for when a drawing has been shrunk past usefulness: if the door swings on the plan have disappeared into grey fuzz, the drawing is decoration. Decoration is fine on a cover. It is not a project page.
Two Is Not a Brave Choice
Everybody writing about this assumes your problem is trimming down. Sometimes it is the other way, and the advice out there is silent on it.
One project, however spectacular, reads as a fluke. Two reads as a small sample. Three is the first number where a reader can see a pattern — that your drawings are consistently legible, that you consistently think in section, that you consistently take something to a scale where it has to work. In practice, that pattern is what actually gets you shortlisted, far more than any single hero image.
There is a structural reason too. With two projects you have nothing to close on. You open strong and then it stops.
The Three-Question Test
Run every candidate project through these. All three, not two of three.
Does it demonstrate a skill the job advert actually asks for? Read the advert properly, then read the practice's project list. If they are a housing practice, your strongest housing scheme outranks your more beautiful cultural project. We teach selection as two passes: first order your projects by how good they are, then reorder by how relevant they are to that specific studio, and reconcile the two. If you have nothing in their sector, do not pretend — pick your strongest work and be explicit about the transferable part. A residential unit study and a hotel room study are the same skill wearing different clothes, but you have to say so.
Do you have process material? Not just the final render. Sketches, the rejected option, the massing study that went wrong, the site analysis that actually changed the plan. If the only surviving assets are three renders and a beautifully typeset title, you do not have a project. You have a picture. Reviewers are looking for how you think, and you cannot show thinking with an output.
Can you talk for five minutes about a decision you made? Not the concept — the concept is usually the tutor's brief with your adjectives on it. A decision: why the core moved to the north side, why the structure changed from frame to cross-wall, what you gave up to get the courtyard. If you cannot name the option you rejected, you will not survive the interview on that project, and the project you cannot defend is the project they will ask about.
Two out of three and it is your fourth project at best, sitting quietly in the middle. One out of three and it goes.
Running Order: Strongest First, Second-Strongest Last
Chronological is a filing system, not an argument. Nobody has time for your character arc.
Open with your strongest project. Close with your second strongest. Everything else lives in the middle. What we have found from reviewing graduate portfolios — and from watching what people actually do with a PDF on a screen — is that the front gets read, the back gets a look, and the middle gets scrolled. That is not a law of psychology, it is just how a busy person handles a document they did not ask for. Design for it.
The middle slot is where your range project goes — the one that is a different scale or a different sector, included to show you are not a one-trick studio. In the course scenario we build around, that is exactly the shape: a large residential scheme, a small residential project, and one non-residential piece that deliberately takes a back seat. It is in there. It is not leading.
One caveat that matters more than sequence: the set has to look like one document. Different type, different line weights and different render styles across three projects reads as three portfolios stapled together, and it undoes the argument that you have a consistent hand. Redraw the key drawings in one graphic language if you have to. It is a weekend, and it is the highest-value weekend in the whole exercise.
Whose Drawing Is It, Exactly?
Group work is the thing graduates either hide or fudge, and both are worse than the truth. A collaborative project is the closest thing in your degree to how an office actually runs, so it belongs in the portfolio — with a credit line that leaves nothing ambiguous.
Something like: Group project with A. Patel, R. Okafor and S. Lind. My contribution: envelope strategy, all technical sections, the 1:20 bay study on page 14. That is one line, it is honest, and it converts a liability into evidence that you can work in a team and still own a package.
Live office work from a placement year follows the same rule, but with two extra hurdles. First, permission — ask the practice before it goes anywhere, and expect a no on anything unbuilt or contentious. Second, be exact about your role rather than vague about it. I produced the door schedule and the 1:20 bathroom details under the job architect is more impressive than an implied claim to the whole scheme, because anyone who has worked in an office knows a Part 1 did not design the building, and a precise small claim reads as professional maturity. If the client is sensitive, anonymise properly: no client name, no address, no north point, no street names on the site plan.
As for unfinished, unbuilt or plain failed studio projects — include them if the thinking is the point, and label them as what they are. A live project shown honestly as a work in progress is interesting. A half-resolved project dressed up as a finished one is the version that hurts you: in my experience the reader spots the missing resolution, and then starts wondering what else has been oversold.
One File, Three Cuts
The last mistake is building one portfolio and firing it at everyone. You do not need four portfolios. You need one master file, organised in sections, from which you can produce variants in twenty minutes.
Swapping the third project and reordering the first two covers most of what tailoring means. A housing practice gets the housing scheme at the front. A practice doing schools gets the one with the best public-space section. Part 2 admissions are a different audience again — they want development, iteration and evidence that you can sustain an enquiry, so the process pages get longer and the technical pages get shorter. An employer usually wants the reverse: can you produce information that a contractor could build from.
Same projects, different weighting. That is the whole job, and it is the reason to keep the source file tidy rather than flattening everything into a PDF you can never edit again.
The Content and Projects lesson in our Perfect Portfolio course walks through this selection with a real set of projects, including how the tailoring decision changes the running order — and the rest of the course builds the InDesign file that makes swapping a project a ten-minute job rather than a rebuild.
One last thing. If you have a project you keep arguing yourself into keeping — the one that needs a paragraph of explanation before anyone can see why it is good, the one where you are hoping the reader skims — that is your answer. It is out. Nothing in a portfolio should need defending in its absence.
