[Closed] Statistics fans and parking tickets

More likely to be poorly written software than a deliberate act

Sounds like someone's done some lazy query writing to extract the data from their system, so has used a MIN([Arrival time]) and MAX([Departure Time]) to work out the offenders, probably because people were being scanned by more than one camera on arrival/departure which meant their original query failed so they bodged it without considering your case.

So I suppose (with a single camera) the first assumption is how many photos are unreadable and how many are readable? We don’t know this but as the purpose is revenue generation and ANPR is mature lets assume it’s quite a low number. Make your own assumption but 1 in 30 unreadable seems plausible.

It will probably be related to the positioning of the plate on the vehicle and other factors specific to individual vehicles (is the plate damaged, for example). If your vehicle has some problematic features then you will be much more likely to have this happen than the overall statistics would suggest.

Sounds like someone’s done some lazy query writing to extract the data from their system, so has used a MIN([Arrival time]) and MAX([Departure Time]) to work out the offenders, probably because people were being scanned by more than one camera on arrival/departure which meant their original query failed so they bodged it without considering your case.

This.

This was being discussed on radio four last week, so I'm guessing that it is pretty common.

It starts to get suspicious that both the first exit and the second entrance are missed, but I think it is fairly safe to assume there is little checking or validation even when the time of stay is way over the maximum (the case on R4 was two visits, one in the evening and one in the morning).

I love the benefit of the doubt that it is lazy, rather than deliberately poor implementation. The attitude of the guy they interviewed from the parking cartel was basically 'don't worry they can always appeal'.

Sounds like someone’s done some lazy query writing to extract the data from their system

Indeed that 🙂

But it does make me wonder if you could have T-shirt made up with your numberplate on it.  Park your car, wander out of the car park wearing said shirt.  Later you just wander back in, pick up your car and leave

I would suggest that as there are no penalties and every incentive to write the code a certain way that it's deliberate

It was probably lazy code to begin with.

But there's no incentive for the parking cos to fix it.

But it does make me wonder if you could have T-shirt made up with your numberplate on it. Park your car, wander out of the car park wearing said shirt. Later you just wander back in, pick up your car and leave

Or,just do this and leave the car OVERNIGHT....
walk back in (via the entrance) and then drive the car out...

Would be an interesting test!

DrP

What was the time between each visit? A lot of car parks have a "No return within X hours" rule that I have fallen foul of before.

Thanks all - deffo no, 'no return' rule on the signage but on a specific point, I wonder if no return is specific about its application to the car or the driver....

I'm going with Sobriety's "It was probably lazy code to begin with. But there’s no incentive for the parking cos to fix it" from a plausibility PoV.

Will see if I can dig out the R4 item but presumably the parking company keeps all photos for a period of time regardless of what the database query told them so I will be asking for all copies of our car from that date.

But it does make me wonder if you could have T-shirt made up with your numberplate on it.

There was a case a few weeks ago where someone got a ticket from a bus lane camera that turned out to be a pedestrian's jumper with a logo similar to their numberplate...

If the cameras are rolling all the time, they will have several frames where a number plate is read. So taking min(frame_timestamp(IN_Photograph)) seems like a reasonable way to program a system. Presumably though, you are not the first person to visit a car park twice, so they ought to have stable code to account for this.

To answer your actual question, assuming an innocent explanation, what are the odds of it occurring? Presumably the cameras have a 'failure rate' I.e. someone drives in or out but the camera doesn't 'clock' them, due to objects blocking line of site, rain on camera lens, mist, reflections etc etc. We don't know what this failure rate is but let's assume 1>failed detection >0. I'd guess at 10-20% (but it doesn't actually matter).

The result of these failures is almost always that they cannot issue a ticket. If they miss a typical driver on either the way in or out, they cannot issue a ticket. This is a problem for the parking company, but not for the drivers and importantly, drivers wouldn't get any notification of "We're sorry, we tried to send you a fine but we were unfortunately unable to." If you knew the number of incomplete parking detections, you could work out the likelihood of driving in and out twice and being detected on only the first and last passage.

But we don't know how often the system fails, and since failures of the system only work against the driver in very specific circumstances (such as yours), we are only notified of the system failure very rarely.

TLDR: Whilst you might think it's surprising that they can 'lose' the precise combination of photos required to ticket you, without knowing the denominator here (the failure rate of number plate capture), it's not possible to calculate how likely this eventuality is. Moreover, failure of capture usually disadvantages the parking company and benefits drivers - an erroneously-issued fine will only occur to the driver's detriment in rare circumstances like the OP's.

Or,just do this and leave the car OVERNIGHT….
walk back in (via the entrance) and then drive the car out…

Would be an interesting test!

I did that in a car park in Memmingham. Got ticket on the way in, parked van, left it a week while I went bikepacking, came back in on the bike, it gave me a new ticket on the way in, used that to get out half an hour later and it charged me one euro

It does raise the question of whether the reverse could be claimed:

"Your honour, I didn't stay in the car park for 10 hours as the plaintiff suggests, I left and came back. We know car park cameras have a failure rate. Can the plaintiff prove that the parking system did not fail to detect my first exit?"

This seems quite a plausible defence to me.

This happened to me at forfar Macdonald's.

We never parked . Not even for a minute.

Drove through the drive through for a coffee on the way to a duathlon.

And on the way back for a coffee as we were ****ing frozen.

They played the computer says no game for a bit then eventually rolled over to accept they were at it. -no apology mind.

I would suggest that as there are no penalties and every incentive to write the code a certain way that it’s deliberate

This, they could have the software updated, but why pay for that when they caould potentially gain more money from not fixing the problem.

@superficial - yeah I'm OK with failure rates declaring an infraction null and void, but what about how those likely already low failure rates might occur in such a way that I suffered two back-to-back failures leading to a falsely detected infraction? The point being it's likely a tiny possibility (successful detect - elapsed time - failed detect - longer elapsed time - failed detect - elapsed time - successful detect). Which is why I'm going for the 'poor software implementation, lost revenue to fix' argument.

Regardless I have written a sternly worded letter 😉

We had a work car park in central London that was 1 GBP per visit, this was 20 years ago it's long sold now. I had a classic car needing winter storage so I drove in say October, left it all winter till say March....kerrching 1 quid 6 months storage.

I got rumbled tho, I used to keep car polish kit in my desk and disappeared most lunchtimes to drive the car round the car park to turn it over. Parking it somewhere different to make it look like it had gone home and returned.

Quite a few cars were left in the car park as the owners were posted overseas.

I suppose my thesis depends on the failure rate being high enough to be relevant, which I'm absolutely sure it would be. The actual failure rate isn't really necessary here, but for a demonstration... If there was a failure rate of 10% of car detections (I.e. 10 out of 100 cars pass the cameras without being registered), which seems a reasonable assumption, then:

1 visit to car park:
Chance of being clocked in = 90%
Chance of being clocked out = 90%
Chance of being clocked both times* = 90% * 90% = 81%
Chance of computer making a mistake (I.e. a parking infraction going unnoticed) = 19%
Chance of driver at fault being alerted to said mistake = 0%

So of all single visits to the car park, 19% of infractions may be missed, but 0% of drivers will be aware that there was a problem.

2 visits to car park:
Chance of the cameras correctly detecting all four passages = 65.6%
Chance of computer seeing you only on first entry and last exit (I.e. YES NO NO YES)**
= 0.90 * 0.1 * 0.1 * 0.9 = 0.81%

So basically the cameras get it correct 66% of the time, and wrong 34% of the time. Out of these 34%, in over 33% of instances, drivers could be committing infractions that go undetected. It's only the 0.81% of double visits (~ 1 in 120) where the computer gets it wrong but crucially this is the ONLY time in a two-legal-visits scenario that the driver will be informed that they have committed an infraction.

That % will go up or down depending on the failure rate of the cameras, but the denominator (whether you are actually informed) also does. The other thing we don't know is what % of drivers outstay their welcome, which would affect the balance of false negative/false positive detections, but that's not strictly important here.

So although it's uncommon for an individual to suffer what you have, and FAR more common for a company to miss a chance at issuing a ticket, across all the car parks in the country, this sort of error could easily happen periodically. Basically, it's a bit of statistical quirk. If you were informed every time you narrowly dodged a valid ticket, that would happen relatively frequently. Since we are never told this, any incorrect infraction seems disproportionately more unlikely.

* Assumption: The risk of non-detection is constant.
** Maybe also in the more common YES NO YES YES scenario, depending on the algorithm used

scuttler
Full Member

Got collared for a PCN for a local Lidl where me and Mrs both made separate trips on the same day and the parking contractor conveniently only got a photo of us entering on the first visit and leaving on the second visit.

I had much the same except it was on 2 separate days. Didn't bother to collect evidence or anything, just sent them an angry letter demanding a retraction and apology, which I got. I wonder how many people pay though

I've been 'caught' visiting the Asda near to my workplace on two consecutive days. Caught going in one day, caught going out the next. didn't have the receipts but emailed the company and said I'm a regular visitor (not just ASDA) to the car park, but I park at work. They cancelled it.

It happens, cameras aren't perfect, the industry call them multi visits. They are fairly rare, as pointed out above got to miss 2 out of 4 reads and they have to be the middle two (direction of travel is recorded by the camera, it can tell if the plates reads are getting bigger or smaller), need to have an IN read followed by an OUT read for a valid PCN to generated. Rare but not vanishingly so given the millions of parking events captured each day across the UK. For example if the failure rate was 0.001% that still equates to 10 PCNs per million cars parked per day. It's why there is an appeals process that will hopefully be further regulated under the up coming government review. It's the cowboys in the industry that refuse to cancel tickets under any circumstances that give the more reputable companies a bad name. The better companies actually cancel a fairly high proportion that are appealed, no point in risking going to POPLA if you've got a dodgy case.

as I may have to FoI these **** about the images they didn’t include in the PCN.

Either the scope of the FOIA has been dramatically extended while I wasn't looking, or you won't get very far with that.

 (direction of travel is recorded by the camera, it can tell if the plates reads are getting bigger or smaller),

Didn't know that.
No point reversing in then, so that it just records two exits and you can stay as long as you like?