GS vs Routing wiki

Some comments I saw recently raised a question. Where the guidance in two different wiki’s overlaps, what do we do?

The gas station wiki says: The stop point for gas stations should be placed on a parking lot road that accesses the area.

The routing server wiki says: if there is a chance that traffic can go in different directions at a junction and that junction can experience different amounts of congestion, you should keep the segment before that junction long.

So if a gas station is near an intersection (as most seem to be), do we break up the road to add PLR (maybe fouling up routing but preventing speed data contamination from the gs) or forgo the PLR to keep the road segment long (which would do the opposite, and considering that fewer people use the gs than the road, so speed contamination should be minimal)?

If a gs is accessible from two different roads (like the example in the GS wiki), then a PLR may be more important. But if not, which wiki do we favor? I would tend to favor the routing wiki, but I can see both sides, so I thought I’d ask other opinions.

Russ has brought up an excellent discussion point with conflicting wiki guidance. As a heavy Waze user, I’ve navigated many segments with AGC JN’s, Turning lane JN’s and PLR JN’s placed near an intersecting road Junction. These closely placed JN’s appear to have a minimal impact on navigation from my experience. (50,000 Waze navigation miles / year).

I’ve navigated roadways that for some odd reason, Waze navigation will send me off on an odd detour (Welsh Rd / SR-63 in Maple Glen, PA is a good example). The segment consistently bypassed had NO PLR’s until recently to capture the heavy in and out traffic along the segment from the adjacent businesses. I believe speed data pollution from Wazers stopping at the local businesses caused Waze navigation to suggest a longer “around the block” route.

Since adding PLR’s to this segment a few months ago- I have not been instructed by navigation to take the around the block route. This was a weekly occurrence previously. It is my belief the PLR’s now capture the stopped in & out traffic and prevent data pollution of the adjacent road segments.

The WME routing server wiki does not provide any specific guidance regarding adding or not adding JN’s prior to an intersection other than state - “At a minimum this causes an incorrect ETA for routing, and it might actually cause traffic to be rerouted unnecessarily, and less optimally, through another route. Hence, if there is a chance that traffic can go in different directions at a junction and that junction can experience different amounts of congestion, you should keep the segment before that junction long.”

The keyword here is “might”, suggesting that it is unknown whether or not following this guidance actually causes navigation issues?

Gas Station Places are of the highest value to Waze & are locked at the highest lock level (L4) of any WME Place. The potential downside of navigation issues placing a JN close to an intersection - “what might happen” verses the potential upsides are;

1.) reduced speed data pollution by having the PLR.
2) The extremely high value of Gas Station Places to the App. Waze wants door to front door navigation to gas stations.

It is my belief the Gas Station Place wiki guidance should take precidence in this instance over the Routing server wiki guidance, which is not specific enough to warrant changing GS Place guidance.

Although I differ in opinion with Russ on this topic, he should be commended for finding these two WME wiki pages that appear to offer conflicting guidance! Nice work!

I have noticed a similar improvement, although maybe on a slightly smaller scale for the small town I drive through daily. It used to route me around back streets on my way home…which I know from experience is slower… Because everyone I’m following up to that point, is waiting for me to go past them around the corner.
Now it routes me via the main road. The only difference there has been is the addition of a PLR at a Sheetz on the corner. Both exits in/out are within a few hundred feet of a traffic light. What difference does it matter if there are 2-200’ segments at 4 second + 4 second = 8s vs 1-400’ segment at 8s. The time across the distance is the same, but all the junk stationary data is off the road. How many 0 mph users does it take to drop an average from 25 mph to 10 mph consistently? I think adding PLRs has a benefit by keeping actually roadway data ‘cleaner’. All the 0 MPH speeds on a PLR should help deter routing through it also.

For through times, it doesn’t matter. But if a node has turns, that’s where the difference comes in. As the example in that routing wiki shows, you break a segment that has turns at the end of it, now you have two segments, one with turns and one without. The segment that no longer has any turns therefore looses all the data associated with turns, it now just has straight through times. Only the original segment with the turns will keep the turn time data, but that segment would be much shorter and therefore the turning times will be affected (since traffic is only measured over 50’ or whatever length the segment is), possibly leading to inaccurate data, which then affects routing decisions.

I’m certainly no expert, I just see both sides of this. But if your experiences show improvements due to adding PLR, then it’s hard to argue, as long as those PLR are the reason.

This is a case where there is no perfect answer. It is always going to be a judgement call. At some intersections, neither choice (with or without PLR) is particularly good. Sometimes we scratch our heads, look at what we know about the intersection, and guess what the least-worst choice is. If routing at that intersection seems to improve somewhat after we make a change, we give ourselves a pat on the back. If not, we might change things back to the way they were before. Sometimes, we wind up requesting a high lock to prevent the same segment getting edited every other week.

This is simply because of the way Waze works - it keeps track of transitions from one segment to the next, and it uses that information in the routing algorithm.

Waze team is currently working on an improvement that is being tested in intersections where it is necessary to have junctions, but where two-segment transition data at these junctions is not enough for good routing.

If you have a routing problem at an intersectio and cannot decide which solution (PLR or no PLR) is better, or if both seem lousy solutions, please discuss it with someone who has more experience testing this kind of thing out at similar intersections.

And if you can think of a simple way to clarify this a little bit for the wiki pages, please propose some replacement text.

From a users perspective - if adding a junction note for a gas station forces routing to send me an additional three or 4 miles down the road to take the next primary Road – then it’s a problem. I have never received the navigation then sent me several miles out of my way because of such a set up.

In a heavy urban environment – avoiding roads with lots of In and out traffic is most a good thing, avoiding congested segments.

As I read the routing wiki – it appears the problem stems from taking a longer segment and cutting it in half. The data becomes distorted somewhat. That would beg the question whether or not simply deleting the entire segment first before adding the parking lot Road wouldn’t be a better option. The data would be erased and the new replacement segments with the PLR JN would get a fresh start and not contain any distorted data ? That to me seems like it would solve the issue raised in the wiki ?

You cannot fix “distortion of traffic data from junctions too close together” by “starting from scratch” on the segments. It is a natural result of how waze does its routing calculations, and is described (more or less) in this wiki page:
https://wiki.waze.com/wiki/Routing_server#Turn_delays

In some cases, the “traffic pollution” from traffic on an unmapped PLR is WORSE THAN the “turn wait pollution” from an extra junction “stealing” the wait data from the large intersection. In some cases, it is the opposite. Sometimes it is hard for us to predict which way it will go.

If there are a lot of questions about what this really means, I’ll think about putting together a lesson with lots of pictures and as few words as possible. I find it pretty time-consuming to put together that kind of material – it is much easier to just type lots of explanations very quickly and hope someone understands. :slight_smile:

But in this discussion - there is ALAWAYS a gas station on the corner, which is why we are discussing following the gas station wiki versus the routing wiki. Knowing there is a gas station close to the intersection – the speed data pollution from the gas station would have a far greater detrimental impact.

That’s why I’m in favor of following the USA guidance for gas station places.

Of course, that depends. What works in the majority of cases won’t work in all cases.

That is usually a good plan - that guidance was developed because it seems to work in the majority of cases. ONLY IF that does not bring good results (and there will be cases where it will not, because of the turn-wait-time issues) do we need to consider an alternate approach. And at that point, you’ll want to work with someone who has experience with making “difficult” intersections work.

Cutting down a segment that has turns at the end of it has to mess with the turn timing data. How accurate are turn times measured in a 25’ segment vs a 250’ segment? I would think not too accurate. More cars pass through the segment than use a gas station, which is why I was putting priority on keeping the segment long rather than adding PLR.

That is the key issue that determines the answer to my question. But like V said, there may not be a one-size-fits-all answer, which would leave the question open. But in that case, do we need to ask someone for every intersection that may be affected? I’m thinking more so for larger intersections that have multiple turn lanes where the turn time data is more important than for smaller intersections.

The big issue, f you really have a 25’ segment, is that, when traffic backs up, most of the waiting is NOT EVEN ON THE SEGMENT WITH THE TURNS YOU CARE ABOUT. So Waze does not know that it is, on average, a 3 minute wait at that light to turn left but a 15 second wait, on average, to turn right.
As far as Waze knows, EVERYONE has to wait the same length of time before the dunkin donuts, then there is virtually no difference in timing between going left, right, or straight at the light.

This is one of the reasons for the guideline on the speed limits page that advises not creating new junctions within 200’ of an intersection, if it can be avoided.

https://wiki.waze.com/wiki/Speed_limits/USA#Creating_a_new_junction_for_a_speed_limit_change

So, yes, it depends.

Agreed. Typically when I edit parking lot RHOADES on gas station places – I create the JN at the driveway furthest from the intersection. I then drawl the road directly through the pumps where the stopped vehicles will be. Many times the parking lot Road will go around behind the building which may not capture the polluted speed data.

Yes, that is exactly my point and my reason for not wanting to add PLR near an intersection. You want that segment with the turns to be as long as possible in order to get as accurate a turn time measurement as possible.

well – what’s the evidence?
What’s the intersection?
What’s the evidence re: differential turn-wait times without the PLR?
Do they look reasonable? Is there any evidence cross-times are contaminated by parking lot traffic?
What does traffic contamination look like, anyway?
How does traffic in that parking lot behave? Does it come out o to the same segment it just left? How does this affect transition times left, right, and straight?

How close to the intersection would the PLR go, if it were added?
Does the traffic routinely back up past this PLR location? How far? With what affect on timing?

I’m sorry if I strayed from my original question, but I didn’t have any specific location in mind, it was just a general question of what to do when two wiki guidances overlap, such as when a gas station is near an intersection. Those issues you mentioned are all part of the question. I think the answer came down to it is taken on a case by case basis.

Yep. Sorry if I misunderstood the question. I was trying to get down to brass tacks.
I think there are a few places where wiki pages contradict each other, or at least seem to, and they might do with a simple note that guidelines try to cover MOST situations, there are always exceptions, etc.