In this post, I will share my conclusions based on my efforts to add and edit roads on Google Maps in February and March 2023.
I expect this to be the last and fifth post in the miniseries with Tips and Tricks for Road Editors. Please see the series index in the first reply below.
In post 3 I presented the results from making and documenting 72 road edits. Only 24% of my road edits got approved. Then with excellent help from senior road editors, I set out to write a manual for road editors to increase my approval rate (you can find it in post #4 in this series). I followed all these tips and continued to make road edits to test if my approval rate would go up.
After making a total of 2847 road edits, only 833 got approved. This brings the approval rate to 29%. A few more approvals are expected in the coming days, but this is far from what I was hoping for. With 2 out of 3 new roads not getting approved, I have decided to now give up.
Let me share some detailed results before elaborating on my decision to stop adding roads.
Here is the updated “delay graph” now based on 833 approved roads:
As you can see 99% of all road edits are processed within 7 days. The notion that road edits linger in a pending queue for up to 90 days is no longer correct for edits made after January 2024. From 2 different tests, I have concluded that after 7 days all pending road edits are removed.
Here are the delay data in table format:
Approved after _ days |
Number |
Percent |
---|---|---|
0 |
160 |
19.2% |
1 |
326 |
39.1% |
2 |
151 |
18.1% |
3 |
58 |
7.0% |
4 |
71 |
8.5% |
5 |
29 |
3.5% |
6 |
7 |
0.8% |
7 |
20 |
2.4% |
9 |
9 |
1.1% |
10 |
1 |
0.1% |
11 |
1 |
0.1% |
In round numbers 20% get approved the same day, 40% the day after, and another 20% the day after. That is new, quick and highly appreciated.
Here is a table showing how many edits I made per day and how many of them got approved:
Submit date |
sum # Approved |
sum # Submitted |
Approval % |
---|---|---|---|
2/26/2024 |
25 |
100 |
25% |
2/28/2024 |
2 |
8 |
25% |
2/29/2024 |
1 |
1 |
100% |
3/1/2024 |
22 |
54 |
41% |
3/2/2024 |
16 |
26 |
62% |
3/3/2024 |
18 |
95 |
19% |
3/4/2024 |
32 |
66 |
48% |
3/5/2024 |
42 |
181 |
23% |
3/6/2024 |
76 |
249 |
31% |
3/7/2024 |
48 |
158 |
30% |
3/8/2024 |
28 |
111 |
25% |
3/9/2024 |
28 |
109 |
26% |
3/10/2024 |
54 |
175 |
31% |
3/11/2024 |
31 |
101 |
31% |
3/14/2024 |
61 |
213 |
29% |
3/15/2024 |
63 |
191 |
33% |
3/16/2024 |
27 |
112 |
24% |
3/17/2024 |
58 |
154 |
38% |
3/18/2024 |
27 |
74 |
36% |
3/19/2024 |
8 |
18 |
44% |
3/20/2024 |
6 |
0 |
#DIV/0! |
3/21/2024 |
4 |
44 |
9% |
3/22/2024 |
3 |
12 |
25% |
3/23/2024 |
2 |
4 |
50% |
3/24/2024 |
11 |
66 |
17% |
3/25/2024 |
78 |
206 |
38% |
3/26/2024 |
9 |
72 |
13% |
3/27/2024 |
25 |
100 |
25% |
3/29/2024 |
28 |
147 |
19% |
These data are based on the e-mail notifications as my list of edits is full of other stuff.
Now let us discuss some of the thoughts that formed the basis for my conclusion.
Will the approval rate go up when you repeatedly revisit areas to follow up?
After following all the tips and tricks presented in the Manual for Road Editors, and my approval rate was still not improving, I tried to follow the strategy often mentioned by @ErmesT : Split your road edits into smaller bites and revisit later to add more sections after the initial roads have been approved. I use a Maps list to keep track of such places that need more work. I have been extremely patient and revisited such places repeatedly. Often pending edits have vanished and I try again. I was hoping to reach the point where I could start removing places on my Maps list because they were completed because nothing more need to do. Sorry, but this happened less than 10 times!
The tip that new roads that are linked to other roads in both ends can explain most of the benefits associated with revisiting areas. It is not encouraging to me to see so little progress.
I have been revisiting this area in Malmö, Sweden between 15 and 20 times, but still 2 road segments are missing.
In my opinion it is too labour-intensive to add roads when the successrate is so low. I started working here on February 12th, and it is still not done at the end of March!
On some days (like March 14-16th) I focussed on followup activities only, and unfortunately the approval rated did not increase.
There needs to be a reasonal balance between input and the output: If Google Maps does not appreciate and approve my efforts I reserve my “rights” to not waste my time on adding more new roads.
Does it become easier to find missing roads with more experience?
For new road editors I think the biggest hurdle is to find missing roads on Google Maps. It is honestly very difficult and timeconsuming. You need to train your brain to scan the satellite and arial photos and try guess what might be missing. You can switch between standard maps view and satellite view to check, and still it is hard. Often I have opened the road editor because it is easier to see as roads are shown with higher contrast.
As I live in a region where most roads are already mapped I have had to resort to hunting for less important roads like roads in parking lots and bike/walking paths. I believe that when the local authorities have entered into a content partnership with Google Maps, the municipal data are trusted over my edits. This fact could very well explain much of my low approval rate.
So I trained my brain to look for parking lots and bike paths. Over time it has become easier and easier to spot such places.
But when 2 out of 3 roads are rejected it makes me feel disappointed and let down.
Is it possible to have fun while hunting for and drawing roads?
For sure it is. I would not have continued searching for and drawing 2847 roads if I was not having fun while doing so. It is almost like an addictive computer game to spot and draw missing roads on Google Maps. A computer game with a higher purpose = helping Maps users navigate the real world.
But Google made it unnecessarily cumbersome. The amount of clicks needed (or lack of shortcuts) is staggering. It becomes almost ridiculous and my wife has left the room a few times because of the excessive mouse noise.
Why is it so difficult to get a higher approval rate?
If I knew, I would not be so frustrated. I have done my very best to learn and use all available tips and tricks and still my approval rate is not increasing significantly. Here are some of the thoughts I have been through:
Earlier all approvals were made by human operators. @ErmesT describe how even complex road system got approved almost immediately in the early days of Road Editor. To save costs Google Maps have likely delegated more and more (if not all) road approvals to an AI system. So today it is necessary to only submit super simple roads and revisit the areas repeatedly.
Maybe Google Maps made the cynical decision to let the volunteers work ther buts off and only approve 1/3 of the roads to keep them motivated. As long as the volunteers keep working and there are enough volunteers this is a viable strategy.
Maybe the primary goal of Road Editor is to collect roads drawn by humans to use these for training the AI systems to draw roads automatically in the future.
Maybe Roads draw by volunteers is a temporary thing while waiting for more local authorities to become Content Parthers who upload their professional geo data.
Maybe there is strange algorithm prioritizing which suggested roads get evaluated and approved. Maybe there is a cap on how many roads a volunteer can get approved per day.
Maybe the AI approving roads is not good enough. It needs a lot more training.
This is a guessing game, but the end result is the same for me. The approval rate need to be at least 2 out of 3 to keep me motivated. 96% approvals would be wonderful.
What more can be done?
Most of the problem can only be addressed by the Google Maps team responsible for the Road editing. So I will suggest two initiatives we volunteers could entertain in the current situation.
- Better manual
I expected the manual to enable new road editors to achieve a higher approval rate. More testing is needed to figure out what is missing or wrong in the manual. I invite others to take on this task. I will be happy to delegate editing rights to fellow road adding enthusiasts.
- Let us colaborate on difficult cases
I created a second Maps list with roads that were impossible to add. Maybe we could join forces and learn from each other while trying to crack such cases together. One example is when a parking lot has new and very dark asphalt and white stripes. Such high contrast seems to confuse the AI so all of them are rejected. Likewise parking lots with no stripes can also be hard to add. Another example is parking lots near airports.
This concludes my conclusions. My intention was not to scare new volunteers from adding missing roads, but help setting realistic expectations. It is fun to hunt for and draw missing roads, and if you are okay with 2/3 of your roads not getting approved, then go for it. Secondly, I hope the responsible team can use this reflections to prioritize making the system approving roads better, make the road editor more userfirendly, and share some much needed official help and tips to volunteer road editors.
Cheers
Morten