Hello! Just some background, I normally don't jump into these as I do like to have an organic approach and see how the local community handles it. I can't be everywhere, but I do skim these state forums when I can, and didn't know this was one was active as it has turned out. Going through every post, I value the dialogue here. A lot of time and effort has been used here to get us all on the same page and making same editing practices from one city to county, to county to state and state to state in this region. I appreciate that.
I’ll try to use the examples that were used and give a thought based on what I have found through the public resources and what was said during the 2013 meetup in Palo Alto. In combination of these resources, I think we can get a better map product.
Some background on Waze is that it started out as a commuting app first and foremost. With the direction that crowdsourcing has provided, the organic and dynamic situation we are in, has led it to be what it is evolving to today, a navigation app. With that being said, it still lacks the completeness you would expect in the sense we have a lot of rural areas to touch up on. Many I imagine are from those who aren't into technological advancements or the worry of what happens to the price of tea in Detroit from Pellston. I experience that everyday where I live literally on the border of Urban KC and the rural country. Hence why in part I don't go poking my nose into every topic that comes into a state forum, however if some clarification is needed, I'll do my best to fill the gaps or try to get the best answer.
My first example I can produce is the Robinson Rd between US-31 and Pleasantview Rd. I have brought up two external resources, Emmet County GIS and MDOT Main NFC map products. First Emmet GIS doesn't make a distinction at all between road type much. Highway or Major Collector seem to be the same to them but they do make difference between local roads, which Robinson and Mill seem to have same level according to them. It gives us some reference, it shows that they do consider it a primary road type to connect between US-31 and Pleasantview at that spot. Now to MDOT NFC, they consider it a Major Collector today and will continue in their future NFC according to their legend. Both give distinct road class of it being more important to go from US-31 to the lake shore. If it isn't a highway perse, then I'd say it would at least classify at a primary street type.
Using that area still, Van Rd was next. One map, the GIS for the county, considers it a local road, but MDOT NFC considers it a major collector. Brutus Rd / Stutsmanville Rd, from US-31 to the shore, is considered a minor Collector according to MDOT. GIS again considers a local road. So we get two conflicts here for sure. I'd say best practices, local knowledge, and judgement are what will prevail for these scenarios.
MDOT NFC
MDOT Emmet Count NFC PDF
Emmet County GIS
Now, I'll take your routing example head on in what appears to fail for me. I place origin in livemap in Pellston, I place destination near West bank of Larks Lake, get two decent routes, one from the North along Levering, and one along the South on Robinson. Now, here is where the routing fails to provide the more direct, efficient route. Using same origin on Edgar in Pellston, destination is now near the shoreline on Trick Rd, funny how that road is there where I randomly clicked, literally. Anyways, the routing doesn't take the most direct route, it wants to go along Robinson up Pleasantview, then across Levering or up US-31 and then across Levering. Why? I assume because the road types at this point. That is why we could and should use primary streets for these roads that connect rural cities. For a local, we may know the best route, even Waze might learn our frequent route, but the passer by might be the one who ends up suffering because they were taken x amount of miles around, wasting gas and time. Rural, maybe not so much of a big deal, or a big deal anywhere. But we do have that bit of confidence that users must build to gain their trust in using Waze as a normal navigation app and not just a novelty. Based on my research above, I believe that Robinson should be a primary street from the shore to US-31 at least. Van may be disputed yet. Does GIS have a more up to date classification of their maps or does the NFC version provide the more current representation of traffic data? I don't know that much about the local map resources currently so I can't answer that. The v12 Nov 2012 on the NFC helps with a date though.
To answer some other things brought up, Waze in theory, should route based on how a user programs their device (shortest/fastest, avoid dirt/avoid long dirt, avoid tolls/highways etc) to the most direct path. In short distances, this works because as it was originally designed, it was the commuter design. It was at the meetup that they said road types in the end should not make a difference in how a route is generated and provided, no matter what. However, through practice and theory we have discovered that isn't the case, and have been doing so called, hacks, to circumvent this dilemma. Case above with Robinson for example. Another case was presented in front of the heads of Waze, Ehud, Noam, Ohad and Shirli. It was routing in rural California and how a route was used, when stopped for gas, that route became broken in the middle as it was trying to find a route from that new starting point. The roads were classified and placed in their proper types, but the routing bug wanted to route inconveniently out of the way. Now this isn't this case in direct terms, but part of placing road types is to get routing to cooperate. There are other factors, but parameters such as speed meta data and length, junctions, etc are taken into account with the routing. While in theory, road types should not matter, in some ways they do. Most of the issues that are affected are the long distance routing as we transition to this navigation app phase of development. Now granted, I am not up there driving these routes, so I don't have a clue if that is the case or I am just talking to talk at the moment. I do have a scenario where construction is going on here nearby, a direct route did exist around the construction, but it wanted to route me backwards to go on the freeway and around. It was once I past a certain point it continued to then give me the better route. I call that a bug because the map even updated the next day after I made the changes the night before.
I am not sure of all the parameters for the auto-zoom feature when driving. Living near the urban core enough to drive through it, I don't see much clutter, even when slowing down and the grid becomes more apparent. But even then I don't think that it is overwhelming. You can certainly see blocks of areas where the subdivisions are, where the primary streets connecting to the freeways, it provides a good sense of orientation. Even at slow enough speeds, only certain roads are showing names, even when stopped. Older client versions showed every street name, that was definitely a cluttered mess. I believe a map is useful when people can relate to what they know of an area and when it matches. If I am coming up in my opinion as visitor, driving N or S on 31, I come to Robinson and I see that Mill is the only primary road, I might dismiss that path across to the shore and go by it. But if it is vetted by an editor to say this path serves a better source than a dirt road, at least they'd have the visual reference.
The one example of Mundt Rd I am not sure if that can be used as an example of direct path. The street it is on is a dirt road by aerial images. Based on the road configurations, you wouldn't want to label Mundt primary for sure to navigate through that, it comes to me as a more local road type that should only be routed to if you have an address lookup there or the like. Having the primary street like that around the curve is proper in this case. And this is where the road types theory doesn't matter, doesn't hold up because clearly it is the shorter, more direct path, to just cut across, but it would appear then that Burt takes priority in the routing. Speed data on Burt is ranging from 35-45 compared to some segments on Mundt with 25-30. But then we visibly give the cues to users so they can also see the best path. Burt according to the NFC is a major collector. Many people it looks like, would rely on that to get in and out of the rural area to head to town. The GIS doesn't go beyond considering it local road. It isn't like a subdivision road that goes off a primary street into treed streets, this is a road that everyone uses. I could see the justification in the NFC and I also see why GIS refers to it another. In the end, I can't give guidance on that one except I am open to further discussions about it because it is still vague.
I've been all over the place with this post, I'm trying to be concise in explaining this. One other area I'll try to finish this up on is what was showed to me, that I didn't know was in development but a great deal of work by Andypoms and Jasonh300 in translating and putting into practice, the DOT Functional Classification (National Functional Classification according to MDOT) for CT and LA. Essentially three experiments if you will were done to translate the NFC to Waze's road types. Of the three, two worked best, and of those two, each had their pros and cons as they related to the road types and navigation nuances. Here are the two versions:
Principal Arterial - Interstate = Freeway
Principal Arterial - Freeway = Freeway
Principal Arterial - Other = Major Highway
Minor Arterial - Minor Highway
Major Collector = Primary Street
Minor Collector = Primary Street
Local = Street
Principal Arterial - Interstate = Freeway
Principal Arterial - Freeway = Major Highway
Principal Arterial - Other = Minor Highway
Minor Arterial = Minor Highway
Major Collector = Primary Street
Minor Collector = Primary Street
Local = Street
The last version according to their results is it matches the DOT guidelines more closely. While the other example I gave before had some issues with illogical road type changes, significant disconnect between major highways, and doesn't closely match DOT guidelines. For the most part, this all relies strictly best interpretation and judgement. While the major highway segment going through the small town probably should be minor highway based on some Wiki References, it stays major highway much like US-31 through Pellston.
Well, I hope I was of help and benefit. I might have more of an urban take on it, but hope I didn't let that interfere with my rural/suburban roots. I don't think that it is necessary we road type based upon looks as more as it is based upon functionality, where the UI reaps the benefits in the end, and overall I think having both in these cases is an added plus, especially if our main goal is the aforementioned of a better experience. I hope I wasn't way off base on this, let me know and I hope we can steer this ship right. Let me know if there is anything I might not have addressed, thanks!