What a Concept Study Needs to Start

Concept studies fail in a small number of repeated ways, and almost all of them trace back to a single missing document that takes twenty minutes to write and that virtually no project produces.

This guide covers what has to exist before production, ordered by how much damage its absence causes. The technical inputs are the easy part. The decisions are where the failures live.

The decided list, which prevents most failures

Write two columns. What has been decided, and what has not. That single page determines what the images can honestly show and it settles most arguments before they happen.

At concept stage the decided column typically holds massing, approximate height, orientation, the position on the site and the broad arrangement. The undecided column holds facade treatment, materials, window rhythm, entrance detail, landscape, signage and almost everything that makes an image look finished.

Without that list, a studio fills the undecided items with plausible choices, because an image cannot show nothing. Those choices become expectations, and when the developed design differs somebody asks why the building changed, and the honest answer is that it never was that building.

With the list, the same studio knows exactly where to stop, and the resulting images are honest about their own provisionality, which is the property this stage most needs.

The options, reduced before anybody quotes

How many schemes are being compared is the largest cost variable and the reduction should happen before a vendor is approached rather than after.

Six schemes produce six buildings to model and a comparison nobody can hold in their head. Three produce a comparison a group can actually make and roughly half the production.

The reduction is a judgement rather than an analysis, and it usually has to come from whoever owns the project, because a design team eliminating its own work is a difficult position to put people in.

The cheap way to do it is rough massing screenshots produced internally in an hour, circulated before anything is commissioned. Ugly, free, and entirely sufficient to eliminate the obvious failures.

The yield position, or an acknowledgement that there is not one

Before commissioning imagery, establish whether each scheme has survived the numbers, because producing images of a scheme that fails on yield is money spent decorating something already dead.

That does not require a finished model. It requires knowing roughly what each arrangement produces, what the parking implication is, and whether the result is in the neighbourhood of the target.

If none of that exists yet, that is a legitimate position and it should be stated, because it changes what the imagery is for. Images produced to explore rather than to decide can be rougher, faster and cheaper, and nobody should treat the output as a comparison of viable options.

What causes damage is the middle case: imagery produced as though the schemes were viable when nobody has checked, and a decision taken on that basis.

The site, with real ground

A concept study places buildings on a site, and the site has to be described rather than assumed.

Boundaries, setbacks and any easement or constraint that affects where a building can sit. Ground levels, from survey where any slope exists, because a scheme sitting on an assumed flat plane reads wrong on any site that is not flat.

Anything that will remain: existing structures, mature trees, retaining conditions, access points. Those constrain the schemes and they also appear in every image, so getting them wrong is visible.

This is the input most often supplied late, because it belongs to the survey scope rather than to the design scope, and it holds up production more than any other missing item.

The surrounding context, and how accurately it is known

The question at concept stage is nearly always relational, which makes the neighbours part of the subject rather than part of the background.

What is needed is height and position rather than detail. Adjacent buildings as accurate volumes, the street, and anything visible from the chosen viewpoints.

Where the surroundings are only approximately documented, say so, because the alternative is a studio estimating them and the estimate becoming the basis for a height comparison that somebody will later check.

On a sensitive site, establishing the context properly at this stage is worth the cost, since the same context carries into entitlement imagery and marketing later and it only has to be built once.

The decision date, treated as the deadline

Concept imagery expires. The decision happens whether or not the images arrive, and imagery delivered afterwards is worth nothing at any quality.

So the date is not a scheduling preference, it is the specification. And the real date is usually earlier than the meeting, because material has to be circulated, reviewed internally and sometimes translated into a board pack.

Working backwards from circulation rather than from the meeting typically removes a week from the available window, and discovering that late is the most common way a concept package arrives too late to matter.

Who is in the room

The audience determines how much legibility has to be added, and it is the input least often stated in a brief.

A design team reads massing models directly and needs very little. A mixed group with non technical members needs context, scale reference and eye level views. A board or a funder who will see the material without anybody present to explain it needs annotation as well.

Naming the audience explicitly changes the product, and it also changes the budget, because adding legibility for non technical viewers is real work and adding it unnecessarily is waste.

What each image is supposed to answer

Every view should have a job written down before it is produced, and the absence of that is why most concept packages consist of general views that answer everything vaguely.

The jobs recur: how tall does this feel from the street, how does it sit against the neighbour, how much of the site does it consume, what happens at the corner, what does arriving look like.

Naming the job also chooses the viewpoint, which is the decision with the largest effect on whether the package works and the one most often made by default.

A viewpoint chosen because it addresses the disputed question can be explained later. One chosen because the building looked good from there cannot.

An existing model, if there is one

Ask the design team what exists before commissioning anything, because a usable massing model is common and it substantially changes both the timeline and the price.

What matters is what tool it is in, how current it is, and whether it covers all the options or only the preferred one, which is frequently the case and worth discovering early.

It is also worth asking whether real time rendering is already available in their workflow, because that is often the cheapest legibility upgrade and it may already be paid for.

The material position, stated as categories

Materials are almost never decided at concept stage and images have to show something, so the useful input is a category rather than a specification.

Predominantly glazed, predominantly masonry, a mixed base and upper treatment, light or dark overall. That is enough to produce an image that reads correctly without claiming a product nobody has selected.

Supplying a specific product at this stage causes the exact problem the decided list exists to prevent, because it appears in the image, becomes an expectation, and is then value engineered out later.

Whether generated imagery is acceptable

This should be settled explicitly, because generative tools are now common in early workflows and their output is indistinguishable from rendered output in a file.

Generated images interpret rather than depict. They add facade rhythms, materials and treatments because invention is what makes them convincing, which is acceptable for conveying mood and corrupting for comparing options.

The workable position is to permit generative output for atmosphere, prohibit it for the comparison set, and require anything generated to be labelled so nobody downstream mistakes one for the other.

Who approves, and at which point

Concept work benefits from two small approvals rather than one large one, and specifying that up front changes the outcome.

The first is the setup, reviewed on a single option before the others are produced: viewpoint, camera height, light, level of detail. If any of those are wrong, every subsequent image inherits the error, and changing them later means reproducing the whole set.

The second is the full set, reviewed side by side rather than sequentially, ideally by somebody who does not have a preferred scheme, looking specifically for any way one option has been advantaged.

What happens when the design changes mid study

Concept schemes move while they are being visualised more often than any other stage, because the design is by definition unresolved and the act of looking at it produces changes.

That is not a problem to be eliminated, it is the process working. Somebody sees the massing against the neighbour for the first time and realises the top floor should step back, which is exactly the insight the study was commissioned to produce.

What matters is having agreed in advance how a mid production change is handled. A change to one scheme is contained and cheap. A change that affects the shared setup, the context or the level of detail affects every option and means reproducing the set.

Agreeing a small allowance for one round of scheme level adjustment, and treating setup level changes as a separate conversation, keeps the process moving without producing an argument about scope in the middle of a live decision.

Ownership, agreed before production

The surviving scheme becomes the developed design, which makes the concept model the first asset of the project rather than a disposable expense.

If the project owns the model, the context and the site setup, then entitlement imagery, marketing renderings and sales material all start from something substantial. If it does not, each later stage rebuilds the same site from nothing.

Across a full development that is a larger sum than the concept package itself, and the leverage to secure it exists only before the work is delivered.

What each missing item costs

No decided list: the studio invents the undecided items, and the invention becomes an expectation the developed design then breaks.

No reduction: production cost multiplies across schemes that were never going to be chosen.

No yield position: attractive images of a scheme the model later rejects.

No ground levels: buildings sitting visibly wrong on any site that is not flat.

No context: images that cannot answer the relational question the stage exists for.

No decision date: a package that arrives after the decision and is worth nothing.

No stated audience: either illegible images or unnecessary polish, both of which are paid for.

A realistic sequence

Write the decided list and reduce the options, which together take an afternoon and determine most of the outcome.

Confirm the site information and the context sources while the vendor is being selected, since those are the inputs most likely to be late.

Approve the setup on one option before the rest are produced.

Produce the remaining options against that approved setup, which is fast because the context and the environment already exist.

Deliver as a side by side comparison, before the circulation deadline rather than before the meeting.

Frequently asked questions

What is the single most useful document to prepare?

A two column list of what has been decided and what has not. Without it a studio fills the undecided items with plausible choices, those choices become expectations, and the developed design then appears to have changed when it never was that building.

Should the options be reduced before or after commissioning?

Before. It is the largest cost variable and the reduction is free. Rough massing screenshots produced internally in an hour are sufficient to eliminate obvious failures, and going from six schemes to three roughly halves production.

What site information is needed?

Boundaries, setbacks, constraints, and ground levels from survey wherever any slope exists, plus anything that will remain such as existing structures, mature trees and access points. Ground levels are the input most often supplied late.

Why is the decision date the specification?

Because concept imagery expires. The decision happens whether or not it arrives, and late imagery is worth nothing at any quality. The real deadline is circulation rather than the meeting, which usually removes about a week from the window.

Should generated imagery be allowed?

Settle it explicitly. Permit it for atmosphere, prohibit it for the comparison set, and require labelling. Generated images invent facade rhythms and materials to look convincing, which is fine for mood and corrupting when comparing specific options.