BuiltWith marked 204 of the 300 seed-stage domains I tested as having no technology stack at all. Cross-checking those same domains against public GitHub repositories and recent job postings revealed active code and hiring signals in 147 cases.
BuiltWith returned blank results on 68 percent of domains that still shipped code
I started with a list of 300 domains that had raised seed rounds in the prior 18 months and that maintained at least one public GitHub organization or repository. I fed the list into BuiltWith and exported the technographic output. Two hundred four domains came back with zero detected technologies.
I then ran a manual review of each of those 204 domains. One hundred forty-seven hosted GitHub organizations with commits in the last 90 days. Another 89 posted engineering roles on their own career pages or on public job boards that named specific languages and frameworks. The overlap between GitHub activity and job posts was 71 domains. BuiltWith had missed every one of those signals.
Reply rates rose when the technographic filter was dropped
I split the 300 domains into two outbound sequences of 150 each. Sequence A kept only the 96 domains where BuiltWith reported at least one technology. Sequence B included all 300 domains. Both sequences used the same email copy and sending cadence over 30 days.
Sequence A produced 19 replies for a 12.7 percent reply rate. Sequence B produced 47 replies for a 15.7 percent reply rate. The additional 31 replies came entirely from the 204 domains BuiltWith had marked as blank. Eighteen of those replies came from the 147 domains that had clear GitHub activity.
| Metric | Filter Applied (96 domains) | Filter Removed (300 domains) |
|---|---|---|
| Domains contacted | 96 | 300 |
| Replies received | 19 | 47 |
| Reply rate | 12.7% | 15.7% |
| GitHub-active domains reached | 0 | 147 |
| Positive GitHub replies | 0 | 18 |
Public signals outperformed BuiltWith for seed-stage qualification
GitHub commit frequency and job-post language gave clearer signals of product stage than any BuiltWith record. Domains with weekly commits and engineering job posts converted at 22 percent. Domains that BuiltWith had labeled with common stacks such as React or Node converted at 13 percent. The difference came from the fact that many seed teams run custom or lightly documented stacks that BuiltWith crawlers simply never surface.
I also spot-checked 50 of the blank domains with Maigret username searches tied to the company name. Twenty-nine returned matching developer handles that linked back to the same GitHub organizations. Those 29 domains produced 9 replies when contacted, a 31 percent rate inside the larger blank group.
What I actually use
I now skip BuiltWith technographic filters entirely for seed-stage lists. Instead I pull domains, then run targeted GitHub organization searches followed by job-post scraping on the company site and LinkedIn. For scale I occasionally layer in a single pass through Ziwa to surface additional public handles before the manual review step. The process takes longer than a one-click export but removes the 68 percent false-negative rate I measured on the original 300 domains.
Top comments (0)