I am not joking. IP geolocation is not hard. The issue is the qualitative nature of IP geolocation, which is where things get squirrelly.
Get this. If you come across any IP geolocation and say, “oh, that IP address is located on Earth,” you will be 100% accurate. We have not reached a point on the Internet where we are doing interplanetary IP address allocation. Simply stating that any IP address is located on Earth is 100% correct.
Then the issue is granularity. Okay, we have found the planet. Now, what about the continent? Trust me when I say this, that is where the differences start to flesh out. Finding continents is super hard. Now, you can point to a random continent. Sure, the data gets granular, but it is not accurate. We have seen many IP geolocation providers repeatedly fail to reliably determine the continent where an IP address is geolocated.
We have literally divided our accuracy page by continent, and you will see cases where our legacy IP geolocation providers fail to locate the continent. Now, add the issues with VPN IP addresses, and things get really messy.
Now, there we have the issue with granularity and accuracy. The more granular the data gets, the more the accuracy is in classical IP geolocation.
To be granular and accurate, you must invest a significant amount of energy in collecting location data hints, building extensive data pipelines, and conducting thorough research to select the most effective hints and discover additional methods for collecting location data hints.
IP geolocation is not hard. Trust me. I have seen several, if not dozens, of side projects that aimed to provide IP geolocation services. In fact, next to the todo app project, IP geolocation might be the most popular real-life side project anyone can do. You find a free IP database, slap on an API on top of it and 
profit . Which is not a problem and we do encourage that as we have our IPinfo Lite database service commercially permissible.
Then, if you can scale it up, you can collect geofeed data from the RFC 8805 system and fall back to the HQ location of the RIR when the data is not present.
That is about the extent that young IP geolocation providers could reach.
The issue is that legacy IP geolocation providers, which have been operating for multiple years, have an edge because network providers reach out to them to have them fix (or repeat) the location they want for their range.
The fact is that whatever the industry anyone else can think of doing, we are already doing and doing a much better job then they do. But it represents a fraction of our effort to do that.
The challenge is this. To transition from barely accurate and somewhat granular to being extremely reliable, accurate, and precise, you must make a substantial investment.
The ProbeNet is a self-operated massive investment in network hardware infrastructure that requires active effort in maintenance, expansion, and management. That is just hardware rent and the human hours invested to keep this thing running. The ProbeNet is not crowdsourced. It is operated solely by us, so the investment we make is to ensure the data is better.
ProbeNet is a physical attribute of the scale we operate at for our data. But we process hundreds of location hint sources—petabytes of data from everywhere. In fact, our data processing, machine learning, and data modeling one of our operational investment.
Then we take this even further by investing in our R&D program to innovate and experiment. Also, look at our team, multiple folks with PhDs, veteran data professionals and SWEs, people who are extremely smart and passionate about their work and a visionary leadership team.
The goal is not to be marginally better than the IP geolocation industry. It is about being as good as we can get.
Currently, for us the benchmark is the Global Positioning System. Yep, that is the level of accuracy and granularity we are aiming for. We are past the IP geolocation industry when it comes to the qualitative nature of the data. GPS program launched satellites, plural, for that. Perhaps someday we can do something similar, but for now, we are committed to doing everything possible to invest in the ground, from hardware and data to software and research.