One weak photo sinks the shoot: how VQS scores a set.

VQS scores a photo set by its lowest image, not its average, and applies set-level gates that look across the whole run. A certification run of 12 to 20 original JPEGs is measured check by check, and the weakest frame governs the SET score because a listing is only as good as its worst photo. A set can even fail with every image passing: in one 19-image run with a mean of 98, two soft frames tripped the focus gate and the set failed.

Every image is measured, and the weakest one decides

VQS measures every image in a set against a fixed set of objective checks. The checks cover exposure, white balance, colour, geometry, artifacts and semantic checks such as visible cables and camera reflections, plus set consistency. Each check returns a measured value and a status: PASS at 90 or above, FLAG from 70 to 89, FAIL at 69 and below. Statuses apply per check and per image, and every image receives a score out of 100.

The SET score works differently. It is governed by the lowest image in the run, not the mean. Eighteen strong frames do not offset one weak one. The full method is written up in plain English at VQS explained.

Set gates catch what averages hide

The lowest-image rule is only half the story. VQS also applies set-level gates, which look across the whole run rather than at any single frame. A set can fail a gate even when every individual image passes.

Here is a real case. A 19-image set where every image passed. Mean score 98. Lowest image 93. On a composite basis it looks excellent. But two soft frames tripped the focus gate, and the set failed. No badge was issued. A failing set never certifies, regardless of its composite.

That outcome is deliberate. A mean of 98 tells a marketing manager nothing about the two frames a buyer will actually pause on. The gate does. You can see how gates and per-image results are presented on the sample report page, where every image and every measured value is shown.

A listing is only as good as its worst photo

Buyers do not experience a photo set as an average. They scroll every frame, and the weak one is the one they remember. A soft hero shot, a colour cast in the kitchen, leaning verticals on the facade: any of these undercuts the frames around it.

Scoring the set by its lowest image puts the measurement where the risk is. It also removes the temptation to bury a doubtful frame in the middle of a long carousel. Under VQS there is no middle of the carousel. Every image carries the same weight, and the worst one carries the score.

How many photos a listing needs

A VQS certification run takes 12 to 20 original JPEGs from a daytime shoot. Twilight and dusk shoots arrive with v2, which is tracked on the releases page. That range fits how most listings are actually shot, and it gives the set-level checks enough frames to be meaningful.

Padding a run does not help. Every image is fingerprinted with SHA-256, and duplicates within a set or against earlier runs are removed automatically, so every test measures fresh work. The practical answer to how many photos per listing is simple: between 12 and 20, and only frames you would stand behind individually.

Cull before you deliver

The lowest-image rule changes the economics of culling. Under a composite-only score, an extra frame is nearly free. Under VQS, a weak frame is a liability the strong frames cannot repay. For working photographers, that points to a few habits.

• Deliver fewer, stronger frames. Anything from 12 to 20 images can certify, so there is no penalty for cutting a doubtful frame and a real cost for keeping it.
• Check sharpness at full size. In the example above, the two frames that failed the set were soft, not badly exposed. Softness hides at thumbnail size and shows up in a measurement.
• Treat the worst frame as the deliverable. Before sending a set, ask which image scores lowest. That image is the set score.
• Measure before you send. The runner is deterministic: the same image gets the same score, byte-identical. You can run a shoot and see the result before a client does.

None of this requires guesswork. Because the runner is deterministic, the same files produce the same result every time. A pre-delivery check is not an estimate. It is the answer.

What a certifiable set earns

When a set passes, certification is automatic. The composite determines the tier: Platinum at 100, Certified from 90 to 99, Provisional from 70 to 89. Below 70 there is no mark, and a failing set never certifies regardless of composite.

The badge is a live embeddable widget served by SVL, one line of HTML. Anyone can click through to the public verification register and see the tier, the composite, the covered files and the dates. Every certificate opens the full annotated report, and reports remain a permanent part of the certification record. The mark expires 12 months after the last certified run, and an expired or suspended mark says so wherever it is embedded.

One weak photo can sink a shoot. That is not a flaw in the method. It is the method, because it reflects how listings are actually judged. To see it applied, start at the Standard Vision Lab home page or put a set of your own through the demo.