ASN Lookup Accuracy Problems

I'm running a web tool called 'What is My ISP?' which aims to accurately identify the internet service provider for a given IP address. We primarily rely on IP-to-ASN mapping and subsequent WHOIS data to extract ISP information.

However, we're consistently running into accuracy issues, specifically around differentiating the actual last-mile ISP from transit providers, CDNs, or large cloud providers whose ASNs might show up for an end-user's IP. For instance, an IP might resolve to a major backbone provider's ASN, but the user is actually getting service from a much smaller, local ISP peering with that backbone. This is where our current ASN lookup methods fall short.

We've tried various approaches including aggregating data from multiple public IP geolocation and ASN lookup databases (MaxMind, IPinfo.io, etc.), performing reverse DNS lookups and parsing hostnames for ISP clues, cross-referencing WHOIS records for the ASN owner and IP block registrant, and attempting to deduce ISP from known IP ranges for major providers.

The problem is, these methods often give us the network operator of a larger segment or transit route, not the direct ISP providing internet access to the end-user. This is particularly challenging for residential IPs or IPs behind VPNs and proxies where the immediate ASN doesn't reflect the true service provider. Our current ASN lookup accuracy just isn't cutting it for the end-user ISP.

What advanced methodologies, data sources, or algorithms are out there to improve the accuracy of end-user ISP identification beyond the standard ASN lookup and WHOIS parsing? Are there specific BGP routing insights or less common data feeds that could help us pinpoint the actual ISP more reliably?

Help a brother out please...