Do you need geocoding? I ask because I am one of the makers of the OpenCage geocoder. We offer forward and reverse geocoding from open data (primarily OpenStreetMap, but also others). We're listed in the migration guide, and we'll be offering 20% off our pricing for all former Mapzen customers. Blog post about that is coming shortly.
I'm sorry, but the last time I tried OpenCage it was terrible. And by terrible, I mean in line with everything else that uses OSM-based data. Geocoding is hard and I wish you luck, but you've got a long way to go.
I've never come across a geocoder better or cheaper than Google, unfortunately. They all claim accuracy but fail in trivial cases, or interpolate to the wrong part of the street where Google is spot on (though Google has its occasional flaws). I'd love to see competition but it's probably not coming any time soon.
A comparison is challenging in that in our experience there are many different use cases. It all depends on which geographies you are interested in, and what level of accuracy, speed, data freedom, and cost you are willing to accept. I mean of course we can do query X on different services and say service A returns answer 1, service B returns answer 2, etc. But that's only useful if your queries look like X, and very often they don't. So it's difficult to say "service A is better". Much more relevant is which is better for your specific needs (and budget).
The only guaranteed way to know if a service will meet your needs is to test, hence why we offer 2,500 free queries per day for as long as you like.
sorry you had a bad experience. Feel free to contact me if you want me to take a closer look. "Geocoding" is a very broad term, it really depends if you need reverse, forward, what level of specificity, which region of the world, the quality of your queries, etc. I'm happy to concede we're not perfect for every use case, but we are definitely more than good enough for many (as proven by our many long-term clients). Thanks for trying us out, and as said happy to have a look at your needs specifically if you contact me.
I can't speak for the commenter above, but as someone who runs a paid service built on open geodata I can tell you our best customers are those who first try to self host and then realize what a pain that is. The challenge is not just keeping the software up to date, but that you also need to keep the data fresh. How fresh of course depends on your exact use case, of course, but for perspective OpenStreetMap has 3-4 million edits per day. It's totally doable - and is technically very interesting, but in much the same way that fixing your own car or baking your own bread is doable and fun. It's more cost-effective to hand it off to a reliable expert.
Building/installing is (relatively) easy. Maintaining is hard.
Even without a requirement to stay particularly up to date, the size of the datasets makes working with OpenStreetMap reasonably complicated.
I needed tiles in non-web-standard projections: equirectangular and polar (EPSG:4326, 3575 and 3031). ESRI used to provide equirectangular tiles[0], but have deprecated the service. Polarmap have an Arctic map, but only at standard resolution [1]. I couldn't find an Antarctic map.
With a fair amount of fiddling, and building on the work OpenMapTiles.org have done, it was possible to generate the tiles, but I now realise why few people have attempted this: it requires a lot of processing power, technical ability to get everything working properly, and cartography to have a map that looks good (not just the final styling, but the choice of what data is present in each layer of vector tiles).
I met the minimum of what we needed [2], but I need to find more time to do the rest -- like contours.
Well... what's next for me is rewriting this application in the next 30 days.
Not feeling too "optimistic" there about that.