Night photos were requested for hoardings with no lights

You pay a crew to photograph your boards as proof that the advertiser’s flex is up. For a lit board that includes a night shot, to show it is lit. The crew’s phone offered a night shot for every board, including boards with no lights at all. A photograph of an unlit hoarding at night is a photograph of the dark.

So staff were being invited to capture proof that no report would ever count. Nobody wins from that: the visit is wasted, the photo sits in storage, and the record looks fuller than it is.

The second half of the problem was the screens. For an LED or digital board, a still photograph cannot tell you whether the screen is playing, frozen on one frame, or has dropped your client’s slot from the loop. A frozen screen and a working screen look the same in a picture.

What was actually going on

The completeness rule was already right. It only required a night shot when the board was lit, and the task list only raised a night task for lit boards. The fault was on the capture side. The screen offered Night to everyone, and the campaign data the phone received did not even carry the board’s lighting, so the phone could not have known.

Before changing anything we checked the production data. In the production data, 62 boards were lit, 55 were not lit and 3 had no lighting recorded. Lighting and the lit flag agreed in every case, and there was not one existing night photo on a non-lit board, so nothing in the past would break.

For screens, the rule was missing altogether. A printed hoarding fails slowly and visibly, so two visits a week catch it. A screen can go dark or freeze between visits, and a photograph cannot show it playing.

What we changed

The Night option now disappears for non-lit boards, on the phone and in the office portal, and a night answer left over from a previously selected board is dropped if the board changes. A board with no lighting recorded is treated as lit, because hiding the option on an unclassified board would quietly lose a proof shot, and the mistake would only appear at invoice time.

LED and digital boards moved to a daily check that must include a short video, because only a clip shows the screen playing. The rule follows what the board is, so a screen added tomorrow is covered on day one, and it outranks any setting on a single site. The classifier has to read three descriptive fields that disagree with each other, in any casing and on whole words only, so a “ledger board” is not a screen.

Running it found two defects that the automated checks had missed. A daily schedule expanded to zero windows, and a board with no windows never shows anything due, so the strictest rule would have become the most permissive. And a video could satisfy nothing, because a clip carries no shot type. Both are fixed.

Against the copy of the data, the 2 LED boards got 7 daily windows, each needing a wide shot, a close-up and a video. The other 126 stayed on the twice-a-week rule.

What it did not fix

The log does not say how many night shots were requested in vain before the fix, so the waste is not counted.

A video proves that the screen was playing while somebody filmed it. It does not prove the client’s slot appeared in the loop at every hour. That still takes a person watching.

The pattern, for anyone who asks crews for proof

For each kind of board, write down what proof actually shows the thing you care about. Then check the crew’s screen asks for that and nothing else.

If your report ignores a kind of evidence, the app should not ask for it, and if your board type cannot be proved by a picture, the requirement is wrong rather than the crew.

Where this ends up

The one rule that decides what counts as proof for each kind of board is read by both the capture screen and the report in AdBoard, the system that runs for Gold Sign Media, an outdoor media operator.