"Wrong Location" Problem
At Ahamove (and OnWheel), we are facing with many interesting problems relating to map every day. Such a typical problem is "wrong location", when the address text is not correct to the expected coordinate provided by geocode/places 3rd services. There are a lot of reasons for this, some from the nature of addresses which are confused or tricky, some from the data quality itself. This problem irritates our drivers a lot because it reduces their producity & income due to longer time to find correct address for fullfilling their delivery order. So, on business side, it worth fixing, but on technical side, it is hard.
When we saw this problem, wrong location cases were reported via driver complaints and BD colleagues proxy from customers. And we were only able to forward these information to our data providers. As you can see, with hundred cases daily, this approach is useless to make the situation better.
To be honest, we haven't got any straight-forward - immediate effective solution at that time. So we just tried many ways and some worked out.
1. We built an in-house location database. We know it complicated but it will prove in the long run in many use-cases, as we can enrich it more and more, then become more independent from 3rd party provider.
2. We built an automatic flow to collect, verify and import valid addresses into our database. Based on a well-known principle called "Crowd Sourcing", we utilize and motivate our driver fleet to report and verify wrong location cases to each other, we are just moderators to make this flow more incentive and funny to drivers by reward and gamification. The flow will begin when a driver submits a photo of correct address & relevant coordinate with expectation to get a compensation of extra distance he had to drive. After that, a game on driver app will ask other drivers to verify the photo with TRUE/FALSE confirmation and earn some points in the app. Finally, any photo matched enough "TRUE" confirmations will be updated into our system.
3. We built some internal tools to verify and update ad-hoc cases reported from our staff. Among them, I like a small tool called Map-Compare which will show a coordinate in 3 different map providers to manually check its validity by eyes.
4. We built map client services such as geocode&place apis which utilize both in-house & 3rd parties in a cost-performance balance, to make sure "wrong location" fixes will be prioritized but not miss valid addresses. As a result, we can control cost more proactively and adjust data source when a party has down-time or worsen quality.
5. We built internal metrics to make sure wrong location ratio in acceptable level and keep improving. These metrics are reviewed monthly and notify us if there are any outstanding figures in cost or complaint about wrong locations, so that we can investigate the symptom.
"Wrong location" is surely a long journey at AhaMove and not finished yet, as then more data and feedback we gather, the more challenges we find and fight with. We hope our efforts will pay off with more satisfaction from driver and clients relating to this common problem.
