Criterion 1 — Light behaviour
Light is the strongest determinant of whether a render reads as realistic, and it is where amateur work gives itself away.
What to look at: are shadow edges hard or soft? In the real world shadow edges soften with distance; uniformly sharp shadows read as artificial.
Second, indirect illumination: when light strikes a surface and bounces, it illuminates its surroundings and carries their colour. A white ceiling beside a red wall should show a faint red bounce. Without it, the scene reads flat and lifeless.
Criterion 2 — Material accuracy
A render can look attractive while showing a material that will not be used on the project — which sets a false expectation and creates problems at handover.
The check is simple: place a real product image of the specified façade cladding next to the render. Do texture scale, colour tone and reflection behaviour match?
A common error is applying material texture at the wrong scale: brick or timber patterns appear larger or smaller than reality, and the building reads like a toy.
Criterion 3 — Scale and proportion
Human figures, door heights, balustrade heights and window proportions determine the building's perceived scale.
Quick check: are figure heights consistent with doors and windows? Does the balustrade reach an adult's waist?
Scale error is the most common flaw that leaves viewers with a sense that "something is wrong" without being able to name it.
Criterion 4 — Context completeness
Renders showing the building on empty ground with no surroundings look clean but do not serve sales — because what buyers most want to understand is the relationship to context, not the building alone.
Questions to ask: are neighbouring buildings modelled? Is the site gradient represented correctly? Is landscape at realistic scale, or is every tree identical?
The view matters especially: is the outlook from an upper floor generated from real context data, or is a representative photograph placed behind the window?
Criterion 5 — Consistency
A single image can be impressive; the real question is whether quality holds across a full set.
If the same façade appears in different colours across images, if light direction is inconsistent, or if some images are noticeably weaker, that usually indicates the work was split between people without coordination.
When requesting proposals, ask for a complete project set rather than one sample image.
The effect of real-time engines on quality
Classic offline renderers compute a single frame at very high accuracy. Real-time engines such as Unreal Engine 5 produce dozens of frames per second.
That used to mean a clear quality penalty on the real-time side. Lumen (dynamic indirect illumination) and Nanite (high-polygon geometry) have largely closed that gap, particularly for interior lighting.
The practical consequence: once a real-time environment exists, both interactive presentation and high-resolution stills come from the same source.