Understanding rejected road modifications: how to turn a disappointment into an opportunity

[ErmesT]

Disclaimer: I write what I’ve learned from personal experience and analysis, which will certainly be different from that of others, and cannot be considered an “Official” documentation about adding Roads in Google maps. Feel free to share your observations too.

With the increasing diffusion of Road Maker for Desktop, more and more Local Guides contribute to adding roads in Google Maps.

As one of the lucky ones who was able to contribute to the program first, (See Roads for Everyone - Helping others with the New Road Editor ) I was also lucky enough to be able to witness all the changes to the program since it was made public.

In the beginning everything was easy, almost all the new roads were approved quickly, and very few were rejected.

With the opening of the program to more Local Guides, the pain began: more and more edits were rejected, to everyone’s amazement.

Since November 2021 I have been observing and taking note of these changes, and adapting my way of contributing to the new situation.

What you find below are my personal observations which I hope will help you not only understand why an edit was rejected, but also how to turn our disappointment into an opportunity.

As I have already said, I write what I’ve learned from personal experience, which will certainly be different from that of others.

How our edits are analysed

• From what I’ve seen I can assume that the moderation of our edits takes place in different stages, with an interaction between the AI and human operators:
• The AI analyzes our edit at an early stage and assigns it a priority.
• An operator verifies the edits according to the priority proposed by the AI, and generally confirms it. The action of the operator increases/improves the ability of the AI to analyze our edits.
• Depending on the proposed priority, the edits can be verified immediately, or delayed.
• All edits that are not verified within 90 days (+/- 2 days) will be automatically rejected

From an analysis of my rejected edits I can say that:

• Over 90% are rejected for reaching the “Time Limit”, i.e. for not having been verified within 90 days
• About 5% are rejected as “duplicate”, i.e. submitted twice and already approved in Google Maps
• The remaining 5% are rejected because they belong to categories that Google Maps has more difficulty identifying, such as parking lots and cycle paths.

Look carefully at the screen below. The top edit was submitted on May 9th (90 days ago) and the bottom one on July 30th, and the road is actually already present in Maps. Furthermore, the first edit is related to a secondary road in a remote area of Brazil, while the second is related to a new urbanization in a densely populated city in Turkey

This gives us some indication of how the AI assigns priorities: Densely populated areas where the number of users who can benefit from our edits is higher have a higher priority than those with a potentially lower number of users. Also, as I already explained in “3000+, and counting. My tips for adding roads in Google Maps” simple edits have a higher priority than complex ones.

How to turn our disappointment into an opportunity

This brings us directly to the title of the post: how to turn our disappointment into an opportunity. The AI is learning day by day, and furthermore it seems that an edit rejected for Time limit increases its chance of being analyzed quickly when it is submitted again. So currently my work with the Road Editor is mainly to resend all the edits that have been rejected due to Time limit.

This also gives me the opportunity to explore the surrounding area, and generally for every rejected edit I send out two to six new edits.

Currently about 60 of my edits are approved every day.

So let’s not lose courage if our edits are rejected, but instead let’s use them as a flywheel to multiply our contributions

Haven’t started adding streets to Google Maps yet? What are you waiting for? Join us in the campaign started by @SholaIB . How do you do it? This post will give you all the explanations:What’s TRAC?Can I join??How???

By request I am tagging here @TravellerG

[Amiran]

Thanks for these valuable tips @ErmesT

I really appreciate your guides to better road mapping. For many of us who contributed to MM with those handy tools and approval process, not approving road edits is so disappointing and many contributors left it completely.

But I stand with changes and I’m coming up to have more approved road edits in this new system of mapping, and your tips are great in this way.

[MortenCopenhagen]

Interesting post, @ErmesT

I like how you systematically keep track and evaluate the outcomes.

A few comments:

You used the term Diffusion in the intro. Did you mean roll out or use?

If the 90 + 5 + 5 percent are calculated as percentages of the number of rejected road edits, I would love to also know what your rejection percentage is.

To me the term Overtime is when an employee works longer hours (with higher pay/hour). Maybe Timeout is a better term.

All the best

Morten

[ErmesT]

Thanks for you notes, @MortenCopenhagen

I have replaced overtime with “Time Limit”. Timeout for me is a term related to sports like basketball or volleyball, when a team request to stop the timer in a match.

I believe the roll out of the editor is complete, and the editor should be already available for everyone. Of course I am talking about an increased number of contributions by more users. We ourselves here in Connect invite new people to participate in adding roads almost every day.

A correct number of rejections should be calculated on a timeframe of 90 days, but on a daily basis I can say that the number of rejection is around 30% of the number of edit submitted, and the number oof approval is around 40%. The number is quite variable because the rejected roads are related of submissions made 3 months before. In some period I was very active in remote areas of South America and this increase the number of rejected edits three months later. A three months time shift is difficult to check, and I simply don’t have time to take note of every single submission. In addition the ability of the AI to evaluate our edits is changing on the meantime, and many edits that would have been rejected six months ago are now being approved smoothly

[ErmesT]

Thank you @Amiran

Yes, this is very different from MM, and due to the larger amount of contributor’s Google do not want to delegate the approval process to volunteers.

This is also an interesting way to push the AI at work.

In any case it will be helpful to see if you have the same feeling about the approval process. I’m sure that every comment or example can help Google to improve it

[Amiran]

I agree about the approval process and How our edits are analyzed @ErmesT and you managed its description well.

As I saw the edits on pedestrians and parking roads also approve well when we only draw one road per edit. But when I change and add a complicated road, I didn’t get approval soon.

For example, I edited this part about two months ago (after seeing many complex road edits on Connect) and still, it didn’t approve. I also can’t check the edit history since it has lots of edits. So, I should create a post and ask Googlers to do it.

And about duplicates, I did it only one time. When I selected the Local Road instead of Parking and then I noticed and repeated it with the correct priority. Fortunately, the true edit was approved in one day. But such a situation caused me to think again about needing a “Revert” option (for example within 30 sec after editing) both for roads and POIs edits.

On the other hand, we should see why many road edits can’t be approved within Time Limit. I think most of them are complex ones and all edits except changing the shape/adding the road. The AI mostly needs to handle complex road edits, adding/changing names, and changing priority better.

[SholaIB]

Hello @ErmesT

Sorry I missed your post,as you know there is so much going for us at this time. I’m.quite busy but steal off a moment or 2 to reply and participate here on connect. Thanks for the mention.

Now there are lots of interesting points in this post of yours. I’ll mention just a few that concerns me personally. I always value your input and systematic presentations of facts.

The first is the automatic rejection of edits more than 90 days old.or so. That’s a real pity and loss of data. This delay/rejection of legitimate edits in such a time frame has made some local guides feel its a waste of time and data adding roads. I don’t feel bad for rejections anyway.

The second one is shuttle between AI and operator when it comes to approvals/rejections. Its quite interesting really. It certainly is a tedious job. Hopefully the AI çan get lots of training to improve before TRAC is over.

Of course I have discovered that s long time ago, that compliment and long roads don’t stand much chances.

Happy mapping

Cheers

[ErmesT]

@SholaIB , pragmatism is always better than emotional reactions, and can help to understand what’s going on and how to react.

What you said" “This delay/rejection of legitimate edits in such a time frame has made some local guides feel its a waste of time and data adding roads” is exactly the reason why I wanted to write this post. To transform disappointment in an opportunity.

Every system has some technical limitations, and in my opinion every road must be approved by an operator, even if the AI can do a great job.

A wrong road can cause more problems than a missing one, and the road network is the real pulsating heart of Google Maps.

In an ideal world every correct edit should be applied, but even Google does not have unlimited resources, so they must work based on priorities, and if they can’t review all the edit they cannot (in my opinion) to let them as pending forever.

What would be your “rational” suggestion to the team to increase the approval rate?

• Give more decisional power to the AI
• Increase the trained operators for verifying more edits?
• Limit the amount of daily added roads for every contributor?
• Set a limit of global daily added roads?
• Give the power to trained volunteers to approve roads (like it was for MM)?
• Other

In this post I don’t have a suggestion, I am simply analyzing the data, and explaining what I am doing every day to use every single rejected edits as an opportunity to edit more. In this way I am adding much more roads per day in a shorter time, and many of my roads are approved very quickly when submitted for the second time.

I’m actually not so far to 8k approved roads, and I am now proceeding faster than the previous month.

I was thinking about to reach 8k for the end of August but with the actual approval rate o will probably reach the goal in 6/7 days, (half of the expected time) and with less time invested in this activity

[SholaIB]

My suggestion will be

• Increase the trained operators for verifying more edits?

The AI is good but can be quite limited. I wouldn’t advocate for mass approval as well. If it’s not correct why approve it?

One thing I’ll love to know is how the Googlers on road feel about TRAC. Has it increased their work load or not? Thanks again.

Cheers

[TravellerG]

Firstly, I thank you for accepting my request & tagging me in your valuable post, my dear friend @ErmesT Ji.

Yes, a lot of information… Needs real analysis and processing…

I too appreciate (& thank you) the comment by our @MortenCopenhagen … How meticulously you maintained the data & used in your process…Great!

Shall go deeper, after my Kerala trip…

Most sincerely

[Lukas_B]

Hello @ErmesT ,

Great post and a very serious one for me right now. I’ve been contributing roads with @SholaIB 's TRAC team since April and It’s all coming to a close for me. Not because I want to but because the 90 day time limit will most likely consider me a spammer. In the last almost 4 months, I’ve added on average over 200 roads a day and the 90-day rejections have started catching up to me a few days ago. I’m getting tens and up to hundreds of rejections a day.

My approval for the whole time was about 49% and increased in the last 3 weeks up to 57%. But 57% out of almost 20.000 roads is about 8.000 pending roads which are about to get rejected in the next three months…

What I can do now is to just watch my account get suspended because the AI will think that I’m spamming the system. What’s even worse is that when I go have a look at those rejected roads - as you mentioned out of the whole count about 90 % are perfectly good roads which just didn’t go through the system to be approved.

It’s very frustrating to have to deal with this. But there’s no help in sight.

Best of luck to everyone having the same problem as me.

[Hope]

Thank you for taking the time to write this, @ErmesT .

I felt like your perspective and our perspective was not entirely aligned. To test something, I went to your Google Maps profile and started adding roads in the Province of Treviso, Italy—using satellite imagery as a reference.

I am amazed that 14 out of almost 25 road additions were accepted within a day. Also how easy it is to add roads in your area, I could see dirt on the road from satellite imagery, and the roads are also mostly symmetrical. This would not happen where I live and add roads as the satellite imagery is of significantly lower quality. Only a handful of 5-10 edits would be accepted out of the hundred roads I add.

I can see where you are coming from now. Neither your perspective nor mine is inherently incorrect—we simply perceive and encounter distinct challenges. It reminds me of the 96-image meme I found online stating: “Just because you’re right, doesn’t mean I’m wrong. Life is about perspective.”

But I am not mad at Google. I hold a genuine appreciation for their initiatives, such as Road Mapper, which enables us to contribute to countries like ours, where the task of adding roads presents a unique set of challenges. But it is not perfect and I hope they improve these aspects with time.

[ErmesT]

Hi @Hope

I am happy that you were able to add road in my area. As most of the road here are already in Maps my effort is to add roads in country where (especially in the country side or in the mountains) most of the roads are not in Google Maps. I think that less than 1% of the 8500+ roads added by me are in my country. The areas of the world where I contribute more are:

• South America, with most of the contributions in the Andes range and in the central part of Brazil
• Central America, and especially Mexico
• North and central Africa
• Middle east, especially Jordan and Turkey
• Asia, especially india, Pakistan and Bangladesh

What I mean is that I am used to dirt road, or roads that disappear in the middle of a forest. I add roads because I think that everyone has the right to be reached through Google Maps, and I find not acceptable to see remote villages that are not connected…

So maybe our perspective are not so differents